3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
5 2.6.32-131.6.1.el6 (RHEL6)
6 2.6.18-238.19.1.el5 (RHEL5)
7 * Client support for unpatched kernels:
8 2.6.32-131.17.1.el6 (RHEL6)
10 2.6.18-238.19.1.el5 (RHEL5)
11 2.6.16.60-0.69.1 (SLES10)
12 * Recommended e2fsprogs version: 1.41.90.wc3
13 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
14 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
15 stack overflow. For more information, please refer to bugzilla 17630.
16 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
17 the re-exporting nodes to be mounted with "32bitapi" mount option
19 Severity : enhancement
21 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
23 --------------------------------------------------------------------------------
25 09-30-2011 Whamcloud, Inc.
27 * Server support for kernels:
28 2.6.18-238.19.1.el5 (RHEL5)
29 2.6.32-131.6.1.el6 (RHEL6)
30 * Client support for unpatched kernels:
31 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
32 2.6.18-238.19.1.el5 (RHEL5)
33 2.6.32-131.6.1.el6 (RHEL6)
34 2.6.29.4-167.fc11 (FC11)
35 2.6.16.60-0.69.1 (SLES10)
37 * Recommended e2fsprogs version: 1.41.90.wc3
38 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
39 Lustre filesystem with 4K stack may cause a stack overflow. For more
40 information, please refer to bugzilla 17630.
41 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
42 mounted with "32bitapi" mount option
44 Severity : enhancement
45 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
47 Severity : enhancement
48 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
50 Severity : enhancement
51 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
54 Frequency : only with SLES10
56 Description: Use OFED "KMP" provided by Novell
57 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
58 Lustre is now built against this package. That means you need to
59 install the ofed-kmp package from Novell for the patchless client.
60 Note that the ofed-kmp that Novell ships may not exactly match the
61 kernel version but should still be compatible.
63 Severity : enhancement
65 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
66 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
68 --------------------------------------------------------------------------------
70 2010-07-15 Oracle, Inc.
72 * Server support for kernels:
73 2.6.18-164.11.1.el5 (RHEL 5)
74 2.6.18-164.11.1.0.1.el5 (OEL 5)
75 * Client support for unpatched kernels:
76 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
77 2.6.18-164.11.1.el5 (RHEL 5),
78 2.6.18-164.11.1.0.1.el5 (OEL 5)
79 2.6.16.60-0.42.8 (SLES 10),
81 2.6.29.4-167.fc11 (FC11)
82 * Recommended e2fsprogs version: 1.41.10-sun2
84 Severity : enhancement
86 Description: Distribute lfs_migrate script with lustre
87 Details : The lfs_migrate mentioned in the manual is now included in the
88 lustre tree in lustre/scripts and will distributed in lustre RPMs.
90 Severity : enhancement
92 Description: Failover nid list added to import proc file.
94 Severity : enhancement
96 Description: "som_preview" mount option required to use SOM
97 Details : The clients will need to use som_preview mount option
101 Description: Lustre does not do 1MB IOs to HW RAID
102 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
104 Severity : enhancement
106 Description: lfs quota output is not very convenient for awk/sed-parsing
107 Details : Some positions in lfs quota output table could be empty or
108 non-empty which made it hard to parse it with scripts, now
109 a dash is put instead of space where there is not supposed
114 Description: Update to OFED 1.5.1
115 Details : OFED 1.5.1 will be used in environments that don't use
116 vendor supplied OFED.
120 Description: Remove lnet. prefix from lctl params display
121 Details : Also renamed lnet.memused to lnet_memused to
126 Description: Use conf_param -d to remove permanant settings.
130 Description: Add list_param -R
131 Details : lctl list_param now supports recursive mode.
134 Bugzilla : 21829, 22312
135 Description: Fix llobdstat.
136 Details : 1. Now updating the write_bytes counter
137 2. Port script fixes from 1.8.
141 Description: decrease the usage of memory on clients.
142 Details : 1. On clients, recycle dentries and inodes unused.
143 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
148 Description: llobdstat fix and enhancement
149 Details : add a counter to set a limit to how many samples will be returned
150 fix a wildcard in the path to limit to obdfilter stats only
154 Description: lfs check only allowed for root
155 Details : Code cleanups and added a regresion test of non-root use of
160 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
161 Details : Fix hr->hr_index overload.
165 Description: Fix FIEMAP functionality.
169 Description: FC11 patchless client support added
173 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
174 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
178 Description: New acc-sm script and various test-framework enhancments.
182 Description: Previously landed patch was reverted.
186 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
190 Description: Add ldiskfs maxdirsize mount option.
194 Description: Fix mkfs.lustre support for 16TB LUNs.
198 Description: Print mmp_check_interval
199 Details : print mmp_check_interval and make it possible to abort
200 mount operation in case it takes too long.
204 Description: Add cascading_rw to lustre mpi tests.
208 Description: Store SOM data in EA.
209 Details : Also several SOM related cleanups.
213 Description: Add refcount to md_open_data
217 Description: Use refcount on lov_request_set to prevent memory corruption.
221 Description: Statahead: use parent dir's i_mutex for synchronization.
225 Description: loadgen improvements.
229 Description: Enable peer health detection in ksocklnd and ko2iblnd.
233 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
237 Description: Fix race on obd_devcie:md_stats
241 Description: Fix SLES 11 client builds.
242 Details : Resolved uninitalized pointer gcc warnings.
246 Description: Allow non-root access for "lfs check".
247 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
251 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
253 Severity : enhancement
255 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
256 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
260 Description: Sanity test_27p fails.
261 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
262 Sanity tests 27 were cleaned up and improved.
266 Description: Unify req format on client/servers
267 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
268 buffers and swabbers approach..
273 Description: drop join file stuff.
275 Severity : enhancement
277 Description: Add 2.6.27 and SLES11 (patchless client) support
281 Description: Fix in ptlrpc_expire_one_request() to print the signed time
284 Severity : enhancement
286 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
289 Severity : enhancement
291 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
293 Severity : enhancement
295 Description: correctly handle big reply message.
296 Details : send LNet event if reply is bigger then buffer and adjust this buffer
301 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
303 Severity : enhancement
305 Description: Add support for OEL5.
309 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
311 Severity : enhancement
313 Description: "weak-modules" support
314 Details : Implement "weak-modules" support which enables kernel modules
315 to be used with any kernel that implements the same kABI. In
316 order to achieve this modules are now installed in
317 /lib/modules/$(uname -r)/updates/kernel on all distributions.
321 Description: client could not reconnect to OST because of an active request.
322 Details : The client now retries to reconnect to the same server, if a
323 connect request failed with EBUSY or -EAGAIN. Test case 35b
324 conf-sanity.sh has been created.
328 Description: Don't allow make backward step on assignin osc next id.
329 Details : race between allocation next id and ll_sync thread can be cause
330 of set wrong osc next id and can be kill valid ost objects.
334 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
338 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
342 Description: Error handling in osc_statfs_interpret() has been improved.
343 Details : Check in osc_statfs_interpret() for EBADR.
347 Description: Do not send statfs() requests to OSTs disabled by administrator.
348 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
350 Severity : enhancement
352 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
354 Severity : enhancement
356 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
358 Severity : enhancement
360 Description: Enhancement for lfs(1) command to use numeric uid/gid.
364 Description: Improved cleanup procedure for conf-sanity test_46a.
368 Description: Do not start bulk transfer, if client tried to reconnect.
369 Details : Do not start bulk transfer, if client tried to reconnect,
370 but failed due to active rpcs. Check for exp_abort_active_req
371 has been implemented.
375 Description: Prevent inconsistences between linux and lustre mount structures.
376 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
377 Make the sleep interruptible.
379 Severity : enhancement
381 Description: provide server to client comms path
382 Details : server to client communications path via new LDLM_SET_INFO rpc
386 Description: truncate starts GFP_FS allocation under transaction causing deadlock
387 Details : ldiskfs_truncate calls grab_cache_page which may start page
388 allocation under an open transaction. This may lead to
389 calling prune_icache with consequent lustre reentrance.
393 Description: send correctly lsm on open replay
394 Details : MDS is trust to LSM size on replay open, but client can set wrong size
399 Description: enable client interface failover
400 Details : When a child reconnects from another NID, properly update export
401 nid hash position and ldlm reverse import.
403 Severity : enhancement
405 Description: Communicate OST degraded/readonly state via statfs to MDS
406 Details : Flags in the statfs returned from OSTs indicate whether the
407 OST is in a degraded RAID state, or if the filesystem has
408 turned read-only after a filesystem error is detected.
413 Description: don't panic if EPROTO was hit when reading symlink
414 Details : correctly handling request reference in error cases.
416 Severity : enhancement
418 Description: Add LustreNetLink, a kernel-userspace communcation path.
420 Severity : enhancement
422 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
427 Description: access to llog context before init.
428 Details : move handling CATALOGS file at osc layer and forbid access to llog
432 Frequency : rare, only when client is writing to OST on local node
434 Description: Avoid deadlock for local client writes
435 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
436 memory freeing context. This allows OST threads to set the
437 PF_MEMALLOC flag on task structures in order to allocate memory
438 from reserved pools and complete IO.
439 Use GFP_HIGHUSER for OST allocations for non-local client writes,
440 so that the OST threads generate memory pressure and allow
441 inactive pages to be reclaimed.
443 Severity : enhancement
445 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
450 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
451 Details : move ldlm namespace creation in setup phase to avoid grab
452 _lprocfs_lock with cli_sem held.
456 Description: Temporarily disable grant shrink.
457 Details : Disable the feature for debugging.
460 Frequency : only when formatting test filesystems
462 Description: Unable to run several mkfs.lustre on loop devices at the same time
463 Details : mkfs.lustre returns error 256 on the concurrent loop devices
464 formatting. The solution is to proper handle the error.
466 Severity : enhancement
468 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
470 Severity : enhancement
472 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
477 Description: resolve race between obd_disconnect and class_disconnect_exports
478 Details : if obd_disconnect will be called to already disconnected export he
479 forget release one reference and osc module can't unloaded.
481 Severity : enhancement
483 Description: Allow tuning service thread via /proc
484 Details : For each service a new
485 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
486 created that can be used to set min/max thread counts, and get the
487 current number of running threads.
492 Descriptoin: don't return error if have partially created objects for file.
493 Details : lov_update_create_set uses set->set_success as index for created
494 objects, so if some requests will be failed, they will have hole
495 at end of array and qos_shrink_lsm for allocate correct lsm.
497 Severity : enhancement
499 Description: Update OFED support to 1.4.1
504 Description: Reduce small size read RPC
505 Details : Set read-ahead limit for every file and only do read-ahead when
506 available read-ahead pages are bigger than 1M to avoid small size
509 Severity : enhancement
511 Description: move AT tunable parameters for more consistent usage
512 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
514 Severity : enhancement
516 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
517 Details : allow skip disconnected ost for send statfs request and hide error
523 Description: fix racy locking of mballoc block bitmaps causing BUG
524 Details : The locking of the mballoc buddy bitmap and the in-memory
525 block bitmap was using two different spin locks in some
526 cases. This made it possible to incorrectly access the
527 mballoc bitmap while another process was modifying it,
528 causing a sanity assertion to fail. While no on-disk corruption
529 was reported, there was some risk of this happening.
532 Frequency : rare, on llog test 6
534 Descriptoin: don't allow connect to already connected import
535 Details : allowing connect to already connected import is hide connecting problem.
540 Descriptoin: allow kill process which wait statahead result
541 Details : for some reasons 'ls' can stick in waiting result from statahead,
542 in this case need way for kill this process.
544 Severity : enhancement
546 Description: Add state history info file, enhance import info file
547 Details : Track import connection state changes in a new osc/mdc proc file;
548 add overview-type data to the osc/mdc import proc file.
550 Severity : enhancement
552 Description: MDS create should not wait for statfs RPC while holding DLM lock.
554 Severity : enhancement
556 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
561 Descriptoin: OST grant too much space to client even there are not enough space.
562 Details : Client will shrink its grant cache to OST if there are no write
563 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
564 this grant cache if there are already not enough avaible space
565 (left_space < total_clients * 32M).
568 Frequency : start MDS on uncleanly shutdowned MDS device
570 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
571 Details : stay in waiting mds<>ost recovery finished produce random bugs
572 due race between two ll_sync thread for one lov target. send
573 ACTIVATE event only if connect realy finished and import have
577 Frequency : rare, connect and disconnect target at same time
579 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
580 Details : don't call obd_disconnect under lov_lock. this long time
581 operation and can block ptlrpcd which answer to connect request.
586 Descriptoin: don't lose wakeup for imp_recovery_waitq
587 Details : recover_import_no_retry or invalidate_import and import_close can
588 both sleep on imp_recovery_waitq, but we was send only one wakeup
592 Frequency : always with long access acl
594 Descriptoin: mds can't pack reply with long acl.
595 Details : mds don't control size of acl but they limited by reint/getattr
598 Severity : enhancement
600 Description: Update to SLES10 kernel-2.6.16.60-0.33.
602 Severity : enhancement
604 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
607 Frequency : start MDS on uncleanly shutdowned MDS device
609 Descriptoin: aborting recovery hang on MDS
610 Details : don't throttle destroy RPCs for the MDT.
613 Frequency : on remount
615 Description: external journal device not working after the remount
616 Details : clear dev_rdonly flag for external journal devices in
622 Description: shutdown vs evict race
623 Details : client_disconnect_export vs connect request race.
624 if client will evicted at this time - we start invalidate
625 thread without referece to import and import can be freed
629 Frequency : rare, need acl's on inode.
631 Description: client can't handle ost additional correctly
632 Details : if ost was added after client connected to mds client can have
633 hit lnet_try_match_md ... to big messages to wide striped files.
634 in this case need teach client to handle config events about add
635 lov target and update client max ea size at that event.
637 Severity : enhancement
639 Description: Changelogs
640 Details : Changelogs are a lightweight mechanism to track filesystem
641 metadata and namespace changes. The changelog is recorded
642 permanently on the MDTs, and is periodically "consumed" / purged
643 when records are no longer needed.
645 Severity : enhancement
647 Description: compact fld format with extents
648 Details : Store range of seq rather than every seq in FLD. Seq
649 controller update FLD rather than clients. In Case of CMD, mdt0
650 has FLD, all other metadata server act as non persistent proxy
651 for FLD queries and cache fld entries in fld cache.
656 Description: don't skip ost target if they assigned to file
657 Details : Drop slow OSCs if we can, but not for requested start idx.
658 This means "if OSC is slow and it is not the requested
659 start OST, then it can be skipped, otherwise skip it only
660 if it is inactive/recovering/out-of-space.
664 Description: more cleanup in mds_lov
665 Details : not send LOV EA under replay, we can't know about they size at this
666 time. Don't allow client connect to mds before any ost connected,
667 for avoid problems with LOV EA size and returning EIO to client.
669 Severity : enhancement
671 Description: Interoperability at server side (Disk interoperability)
673 Severity : enhancement
675 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
677 Severity : enhancement
679 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
681 Severity : enhancement
683 Description: New client IO stack (CLIO).
685 Severity : enhancement
687 Description: Commit on sharing. Eliminate inter-client dependencies between
688 uncommitted transactions by doing transaction commits.
689 Thereby clients may recovery independently.
692 Frequency : Create a symlink file with a very long name
694 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
695 Details : If there is no extra space in the request for early cancels,
696 ldlm_req_handles_avail() returns 0 instead of a negative value.
698 Severity : enhancement
700 Description: Add /proc entry for import status
701 Details : The mdc, osc, and mgc import directories now have
702 an import directory that contains useful import data for debugging
705 Severity : enhancement
707 Description: Re-disable certain /proc logging
708 Details : Enable and disable client's offset_stats, extents_stats and
709 extents_stats_per_process stats logging on the fly.
712 Frequency : Only on FC kernels 2.6.22+
714 Description: oops in statahead
715 Details : Do not drop reference count for the dentry from VFS when lookup,
716 VFS will do that by itself.
718 Severity : enhancement
720 Description: Generic /proc file permissions
721 Details : Set /Proc file permissions in a more generic way to enable non-
722 root users operate on some /proc files.
726 Description: Hitting mdc_commit_close() ASSERTION
727 Details : Properly handle request reference release in
728 ll_release_openhandle().
732 Description: quota recovery deadlock during mds failover
733 Details : This patch includes att18982, att18236, att18237 in bz14840.
735 1. fix osts hang when mds does failover with quotaon
736 2. prevent watchdog storm when osts threads wait for the
741 Frequency : only patchless client
742 Description: add workaround for race between add/remove dentry from hash
744 Severity : enhancement
746 Description: Allow OST glimpses to return PW locks
750 Description: LBUG when llog conf file is full
751 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
755 Description: Prevent import from entering FULL state when server in recovery
759 Description: service mount cannot take device name with ":"
760 Details : Only when device name contains ":/" will mount treat it as
766 Description: replace ptlrpcd with the statahead thread to interpret the async
767 statahead RPC callback
771 Frequency : on recovery
772 Description: I/O failures after umount during fail back
773 Details : if client reconnected to restarted server we need join to recovery
774 instead of find server handler is changed and process self
775 eviction with cancel all locks.
777 Severity : enhancement
779 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
781 Severity : enhancement
783 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
785 Severity : enhancement
787 Description: Upcall on Lustre log has been dumped
788 Details : Allow for a user mode script to be called once a Lustre log has
789 been dumped. It passes the filename of the dumped log to the
790 script, the location of the script can be specified via
791 /proc/sys/lnet/debug_log_upcall.
796 Description: avoid idr_remove called for id which is not allocated.
797 Details : Move assigment s_dev for clustered nfs to end of initialization,
798 to avoid problem with error handling.
803 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
804 Details : When connection is reused this not moved from CONN_UNUSED_HASH
805 into CONN_USED_HASH and this prodice warning when put connection
806 again in unused hash.
809 Severity : enhancement
811 Description: File striping can now be set to use an arbitrary pool of OSTs.
813 Severity : enhancement
815 Description: Export bytes_read/bytes_write count on OSC/OST.
819 Description: Early reply size mismatch, MGC loses connection
820 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
821 the connect flags are properly negotiated.
826 Description: quotas are not honored with O_DIRECT
827 Details : all writes with the flag O_DIRECT will use grants which leads to
828 this problem. Now using OBD_BRW_SYNC to guard this.
832 Description: add quota statistics
833 Details : 1. sort out quota proc entries and proc code.
834 2. add quota statistics
836 Severity : enhancement
838 Description: enable quota support for HEAD.
842 Description: Properly propagate oinfo flags from lov to osc for statfs
843 Details : restore missing copy oi_flags to lov requests.
845 Severity : enhancement
847 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
848 : l_getgroups(8), lst(8), routerstat(8)
849 Details : included man pages for llobdstat(8), llstat(8),
850 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
852 Severity : enhancement
854 Description: configure's --enable-quota should check the
855 : kernel .config for CONFIG_QUOTA
856 Details : configure is terminated if --enable-quota is passed but
857 : no quota support is in kernel
861 Description: Remove portals compatibility
862 Details : Remove portals compatibility, not interoperable with releases
867 Description: Resolve device initialization race
868 Details : Prevent proc handler from accessing devices added to the
869 obd_devs array but yet be intialized.
871 Severity : enhancement
873 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
875 Severity : enhancement
877 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
882 Description: Using wrong pointer in osc_brw_prep_request
883 Details : Access to array[-1] can produce panic if kernel compiled with
884 CONFIG_PAGE_ALLOC enabled
888 Description: Client runs out of low memory
889 Details : Consider only lowmem when counting initial number of llap pages
893 Description: *optional* service tags registration
894 Details : if the "service tags" package is installed on a Lustre node
895 When the filesystem is mounted, a local-node service tag will
896 be created. See http://inventory.sun.com/ for more information
897 about the Service Tags asset management system.
901 Description: Kernel BUG tries to release flock
902 Details : Lustre does not destroy flock lock before last reference goes
903 away. So always drop flock locks when client is evicted and
904 perform unlock regardless of successfulness of speaking to MDS.
908 Description: add refcount protection for osc callbacks, avoid panic on shutdown
912 Description: sanity test 65a fails if stripecount of -1 is set
913 Details : handle -1 striping on filesystem in ll_dirstripe_verify
918 Description: ASSERTION(CheckWriteback(page,cmd)) failed
919 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
922 Severity : enhancement
924 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
928 Description: do not process already freed flock
929 Details : flock can possibly be freed by another thread before it reaches
930 to ldlm_flock_completion_ast.
934 Description: LBUG during stress test
935 Details : Need properly lock accesses the flock deadlock detection list.
939 Description: oops in page fault handler
940 Details : kernel page fault handler can return two special 'pages' in error
941 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
945 Description: timeout with invalidate import.
946 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
947 should be handled by ptlrpcd. This produce long age waiting and
948 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
950 Severity : enhancement
952 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
956 Description: enable MGS and MDT services start separately
957 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
958 a co-located MGS without starting the MGS, which is a complement
959 to 'nosvc' mount option.
963 Frequency : after recovery
964 Description: precreate to many object's after del orphan.
965 Details : del orphan st in oscc last_id == next_id and this triger growing
966 count of precreated objects. Set flag LOW to skip increase count
967 of precreated objects.
971 Frequency : rare, on clear nid stats
972 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
973 Details : when clean nid stats sometimes try destroy live entry,
974 and this produce panic in free.
978 Description: Stack overflow during MDS log replay
979 ease stack pressure by using a thread dealing llog_process.
983 Description: wait until IO finished before start new when do lock cancel.
984 Details : VM protocol want old IO finished before start new, in this case
985 need wait until PG_writeback is cleared until check dirty flag and
986 call writepages in lock cancel callback.
988 Severity : enhancement
990 Description: using special macro for print time and cleanup in includes.
994 Description: mds_mfd_close() ASSERTION(rc == 0)
995 Details : In mds_mfd_close(), we need protect inode's writecount change
996 within its orphan write semaphore to prevent possible races.
1000 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1004 Frequency : rare, on shutdown ost
1005 Description: don't hit live lock with umount ost.
1006 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1007 use shrink_dcache_sb instead.
1011 Description: don't panic with use echo client
1012 Details : echo client pass NULL as client nid pointer and this produce null
1013 pointer dereference.
1017 Description: fix build on ppc32
1018 Details : compile code with -m64 flag produce wrong object file for ppc32.
1022 Description: add message levels for liblustreapi
1026 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1027 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1028 will result in an -EAGAIN and in the exit call path we call
1029 alloc_rr() which will with increasing aggressiveness attempt to
1030 aquire precreated objects on the minimum number of required OSCs.
1034 Description: Use old size assignment to avoid deadlock
1035 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1036 the scheduling while holding a spinlock. We do not need locking
1037 for size in ll_update_inode() because size is only updated from
1038 the MDS for directories or files without objects, so there is no
1039 other place to do the update, and concurrent access to such inodes
1040 are protected by the inode lock.
1044 Description: resolve "_IOWR redefined" build error on SLES10
1048 Description: dump the memory debugging after all modules are unloaded to
1049 suppress false negative in conf_sanity test 39
1051 Severity : enhancement
1053 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1058 Description: fix SLES kernel versioning
1059 Details : the kernel version for our SLES 10 kernel did not include a "-"
1060 before the "smp" at the end. while this was not a problem in
1061 general, it did mean that software trying to use the kernel
1062 version to try to detect a vendor specific kernel would fail.
1063 this was most evident by the OFED build scripts.
1067 Description: Don't update lov_desc members until making sure they are valid
1068 Details : When updating lov_desc members via proc fs, need fix their
1069 validities before doing the real update.
1073 Description: don't put request into delay list while invalidate in flight.
1074 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1075 import in flight. this produce timeout for invalidate and sometimes
1076 can cause stale data.
1080 Frequency : on ppc only
1081 Description: not convert ost objects for directory because it's not exist.
1082 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1086 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1088 Severity : enhancement
1090 Description: Update to support 2.6.22.14 vanilla kernel.
1094 Frequency : rare, on recovery
1095 Description: read procfs can produce deadlock in some situation
1096 Details : Holding lprocfs lock which send rpc can produce block for destroy
1097 obd objects and this also block reconnect with -EALREADY. This
1098 isn't fix all lprocfs bugs - but make it rare.
1100 Severity : enhancement
1102 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1105 Frequency : frequent on X2 node
1107 Description: mdc_set_open_replay_data LBUG
1108 Details : Set replay data for requests that are eligible for replay.
1112 Description: lustre_mgs: operation 101 on unconnected MGS
1113 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1114 MGC, and late on MGC cannot successfully connect to MGS and a lot
1115 of the error messages complaining that MGS is not connected.
1119 Frequency : on network error
1120 Description: panic with double free request if network error
1121 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1122 but it's true only for synchronus enqueue, for async enqueue
1123 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1126 Severity : enhancement
1128 Description: client-side metadata stat-ahead during readdir(directory readahead)
1129 Details : perform client-side metadata stat-ahead when the client detects
1130 readdir and sequential stat of dir entries therein
1133 Frequency : on start mds
1135 Description: Implement get_info(last_id) in obdfilter.
1138 Frequency : occasional
1140 Description: Correctly check stale fid, not start epoch if ost not support SOM
1141 Details : open with flag O_CREATE need set old fid in op_fid3 because
1142 op_fid2 was overwritten with new generated fid, but mds can answer
1143 with one of these two fids and both is not stale. Setattr
1144 incorrectly started an epoch and assume will be called
1145 done_writing, but without SOM done_writing ever being called.
1148 Frequency : rare, depends on device drivers and load
1150 Description: MDS or OSS nodes crash due to stack overflow
1151 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1152 In some cases, in conjunction with device drivers that use a lot
1153 of stack the MDS (or possibly OSS) service threads could overflow
1154 the stack. One change which was identified to consume additional
1155 stack has been reworked to avoid the extra stack usage.
1158 Frequency : occasional
1160 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1161 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1162 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1163 oscc_has_objects and see OSC has no objests and call
1164 oscc_internal_create to resend create request.
1166 Severity : enhancement
1168 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1170 Severity : enhancement
1172 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1175 Frequency : very rare
1177 Description: Fix replay if there is an un-replied request and open
1178 Details : In some cases, older replay request will revert the
1179 mcd->mcd_last_xid on MDS which is used to record the client's
1180 latest sent request.
1182 Severity : enhancement
1184 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1186 Severity : enhancement
1188 Description: Add rhel5 support to HEAD.
1190 Serverity : enhancement
1192 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1197 Description: Don't allow skipping OSTs if index has been specified.
1198 Details : Don't allow skipping OSTs if index has been specified, make
1199 locking in internal create lots better.
1203 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1204 Details : access to bitfield in structure is always rounded to long
1205 and this produce problem with not atomic change any bit.
1209 Description: Lustre make rpms failed.
1210 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1211 builds Lustre rpms from tarball.
1214 Frequency : rare on shutdown ost
1216 Description: If llog cancel was not send before clean_exports phase, this can
1217 produce deadlock in llog code.
1218 Details : If llog thread has last reference to obd and call class_import_put
1219 this produce deadlock because llog_cleanup_commit_master wait when
1220 last llog_commit_thread exited, but this never success because was
1221 called from llog_commit_thread.
1225 Description: allow userland application know is lost one of stripes.
1226 Details : fill lvb_blocks with error code on ost and return it to
1227 application if error flag found.
1231 Description: NULL lov_tgts causing MDS oops
1232 Details : more safe checks for NULL lov_tgts for avoid oops.
1234 Severity : enhancement
1236 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1240 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1241 Details : If an OST is down the MDS will hang indefinitely in
1242 obd_statfs_rqset() waiting for the statfs data. While for
1243 MDS QOS usage of statfs, it should not stuck in waiting.
1245 Severity : enhancement
1247 Description: remote_acl support
1248 Details : Support ACL-based permission check for remote user.
1249 Support setfacl/getfacl for remote user with the utils
1250 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1251 the system "{s,g}etfacl" utils.
1253 Severity : enhancement
1255 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1257 Severity : enhancement
1259 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1263 Description: make mgs_setparam() handle fsname containing dash
1264 Details : fsname containing a dash does not work with lctl conf_param
1266 Severity : enhancement
1268 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1270 Severity : enhancement
1272 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1276 Frequency : rare, at shutdown
1277 Description: access already free / zero obd_namespace.
1278 Details : if client_disconnect_export was called without force flag set,
1279 and exist connect request in flight, this can produce access to
1280 NULL pointer (or already free pointer) when connect_interpret
1281 store ocd flags in obd_namespace.
1285 Frequency : only at startup
1286 Description: not alloc memory with spinlock held.
1287 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1288 if any spinlock held.
1290 Severity : enhancement
1292 Description: make lustre randomly fail allocating memory
1293 Details : Make lustre randomly failed allocating memory for testing purpose.
1295 Severity : enhancement
1297 Description: lost problems with lov objid file
1298 Details : Fixes some scability and access to not inited memory problems
1299 in work with lov objdid file.
1304 Description: lfs find does not continue on file error
1305 Details : Continue other files processing when a file/dir is absent.
1309 Description: Inconsistent usage of lustre_pack_reply()
1310 Details : Standardize the usage of lustre_pack_reply() such that it
1311 always generate a CERROR on failure.
1314 Frequency : occasional
1316 Description: MDS hang or stay in waiting lock
1317 Details : If client receive lock with CBPENDING flag ldlm need send lock
1318 cancel as separate rpc, to avoid situation when cancel request
1319 can't processed due all i/o threads stay in wait lock.
1323 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1327 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1331 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1335 Description: Update to RHEL4 latest kernel.
1337 Severity : enhancement
1339 Description: Build SLES10 patchless client fails
1340 Details : The configure was broken by run ./configure with
1341 --with-linux-obj=.... argument for patchless client. When the
1342 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1343 can't search header adequately. Use absolute path such as
1344 -I($LINUX)/include instead.
1348 Description: interrupt oig_wait produce painc on resend.
1349 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1350 private set, and this produce situation when rq_ptlrpcd_data not
1351 copyed to new allocated request and triggered LBUG on assert
1352 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1353 ptlrpcd set if request is changed and can be safety changed to use
1358 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1359 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1360 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1361 these macros and replace them with their difinitions instead.
1365 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1366 Details : Update which_patch & target file for SLES10 latest kernel.
1369 Description: Remove client patches from SLES 10 kernel.
1370 Details : This causes SLES 10 clients to behave as patchless clients
1371 even on a Lustre-patched (server) kernel.
1373 Severity : enhancement
1375 Description: self-adjustable client's lru lists
1376 Details : use adaptive algorithm for managing client cached locks lru
1377 lists according to current server load, other client's work
1378 pattern, memory activities, etc. Both, server and client
1379 side namespaces provide number of proc tunables for controlling
1382 Severity : enhancement
1384 Description: light-weight GSS support
1385 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1386 and reduce performance overhead.
1388 Severity : enhancement
1390 Description: Linux keyring support
1391 Details : Support using service of Linux keyring for Lustre GSS internal
1392 context refresh/cache mechanism.
1396 Description: Fix errors in lfs documentation
1397 Details : Fixes man pages
1401 Description: don't use GFP_* in generic Lustre code.
1402 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1406 Description: obdclass is limited by single OBD_ALLOC(idarray)
1407 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1411 Description: warnings with build patchless client with vanila 2.6.19 and up
1412 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1413 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1417 Frequency : Only for SLES
1418 Description: sanity_quota fail test_1
1419 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1420 file, which makes TSTID[2] inunique.
1424 Frequency : Only for SLES9
1425 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1426 Details : Update target/ChangeLog/which_patch .
1430 Frequency : Only for test_10 in sanity-quota.sh
1431 Description: a bug in quota test code
1432 Details : checking a test flag in a wrong place
1436 Frequency : Only for test_13 in sanity-quota.sh
1437 Description: a bug in quota test script
1438 Details : shouldn't check overfull allocated quota
1442 Description: jbd statistics
1443 Details : Port older jbd statistics patch for sles10
1445 Severity : enhancement
1447 Description: Add printing inode info into message about error in writepage.
1452 Description: reply_lock_interpret crash due to race with it and lock cancel.
1453 Details : Do not replay locks that are being cancelled. Do not reference
1454 locks by their address during replay, just by their handle.
1458 Frequency : When flocks are used.
1459 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1460 Details : Flock locks might destroy just granted lock if it could be merged
1461 with another existing flock, this is done in completion handler,
1462 so teach ldlm_cli_enquque_fini that this is a valid case for
1468 Description: Oops in read and write path when failing to allocate lock.
1469 Details : Check if lock allocation failed and return error back.
1473 Description: lstripe command fails for valid OST index
1474 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1475 instead of lov->desc.ld_active_tgt_count.
1479 Description: sanity.sh failed test 103
1480 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1485 Description: df doesn't work properly if diskfs blocksize != 4K
1486 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1490 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1491 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1493 Severity : enhancement
1495 Description: Adaptive timeouts
1496 Details : RPC timeouts adapt to changing server load and network
1497 conditions to reduce resend attempts and improve recovery time.
1501 Description: llapi_file_create() does not allow some changes
1502 Details : add llapi_file_open() that allows specifying the mode and
1503 open flags, and also returns an open file handle.
1507 Description: merge and cleanup kernel patches.
1508 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1512 Description: Add journal checksum support.(Kernel part)
1513 Details : The journal checksum feature adds two new flags i.e
1514 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1515 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1516 indicates that the commit block contains the checksum for
1517 the blocks described by the descriptor blocks. Now commit
1518 record can be sent to disk without waiting for descriptor
1519 blocks to be written to disk. This behavior is controlled
1520 using JBD2_FEATURE_ASYNC_COMMIT flag.
1524 Description: OSS needs mutliple precreate threads
1525 Details : Add ability to start more than one create thread per OSS.
1529 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1530 Details : execute lfs setstripe on client
1534 Description: mds_obd_create error creating tmp object
1535 Details : When the user sets quota on root, llog will be affected and can't
1536 create files and write files.
1539 Frequency : Always on ia64 patchless client, and possibly others.
1541 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1542 Details : This allows the patchless client to be loaded on architectures
1543 without this export.
1547 Description: RedHat Update kernel for RHEL5
1548 Details : Add the kernel config file for RHEL5.
1552 Description: RedHat Update kernel for RHEL5
1553 Details : Modify the kernel config file more closer RHEL5.
1557 Description: Build failure against Centos5 (RHEL5)
1558 Details : Define PAGE_SIZE when it isn't present.
1562 Description: test_7 on sanity-quota.sh can't be used on separate machines
1563 Details : add facet to handle it
1567 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1568 Details : with NFS, the anon dentry's parent was set to itself in
1569 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1570 corresponding dentry other than use rec->ur_name.
1572 Severity : enhancement
1574 Description: lfs setstripe enhancement
1575 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1579 Description: enable data checksumming by default
1580 Details : enable checksum by default, allow --disable-checksum
1581 configure option and "-o nochecksum" mount option. Checksums
1582 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1586 Description: lustre support for RHEL5
1587 Details : Add support for RHEL5.
1591 Description: Client eviction due to failover config
1592 Details : after a connection loss, the lustre client should attempt to
1593 reconnect to the last active server first before trying the
1594 other potential connections.
1598 Description: when mds and osts use different quota unit(32bit and 64bit),
1599 quota will be released repeatly.
1600 Details : void sending multiple quota reqs to mds, which will keep the
1601 status between the reqs.
1605 Description: osts not allocated evenly to files
1606 Details : change the condition to increase offset_idx
1609 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1611 Description: Data corruption for OSTs that are formatted larger than 2TB
1613 Details : When generating the bio request for lustre file writes the
1614 sector number would overflow a temporary variable before being
1615 used for the IO. The data reads correctly from Lustre (which
1616 will overflow in a similar manner) but other file data or
1617 filesystem metadata may be corrupted in some cases.
1621 Description: Tune the kernel for good SCSI performance.
1622 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1623 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1628 Description: rewrite ext2-derived code in obdclass/uuid.c
1629 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1630 from scratch preserving functionality.
1634 Description: "lfs find -obd UUID" prints directories
1635 Details : "lfs find -obd UUID" will return all directory names instead
1636 of just file names. It is incorrect because the directories
1637 do not reside on the OSTs.
1641 Description: use i_size_read and i_size_write in 2.6 port
1642 Details : replace inode->i_size access with i_size_read/write()
1645 Frequency : when removing large files
1647 Description: scheduling issue during removal of large Lustre files
1648 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1649 It causes scheduling issues when removing large files (17TB in the
1653 Frequency : only with liblustre clients on XT3
1655 Description: evictions taking too long
1656 Details : allow llrd to evict clients directly on OSTs
1659 Frequency : only on ppc
1661 Description: /proc/fs/lustre/devices broken on ppc
1662 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1663 We should make sure the type of 'index' is loff_t and then cast
1664 explicitly as needed below. Do not assign an explicitly cast
1668 Frequency : only for rhel5
1670 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1671 Details : Modify the target file & which_kernel.
1673 Severity : enhancement
1675 Description: omit set fsid for export NFS
1676 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1677 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1681 Frequency : Always, for kernels after 2.6.16
1682 Description: Fix warning idr_remove called for id=.. which is not allocated.
1683 Details : Last kernels save old s_dev before kill super and not allow
1684 to restore from callback - restore it before call kill_anon_super.
1688 Description: buffer overruns could theoretically occur
1689 Details : llapi_semantic_traverse() modifies the "path" argument by
1690 appending values to the end of the origin string, and a
1691 overrun may occur. Adding buffer overrun check in liblustreapi.
1695 Description: Fix error on 'ls .' at the top of the Lustre mount.
1696 Details : Don't revalidate dentry if it is a root dentry.
1700 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1701 Details : Modify vm-tunables-rhel4.patch.
1705 Description: Kernel config for 2.6.18-vanilla.
1706 Details : Modify targets/2.6-vanilla.target.in.
1707 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1708 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1709 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1710 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1714 Description: improve handling recoverable errors
1715 Details : if request processig with error which can be recoverable on server
1716 request should be resend, otherwise page released from cache and
1721 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1722 Details : Modify target file & which_patch.
1723 A flaw was found in the IA32 system call emulation provided
1724 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1725 value could be stored in the %RAX register, which could trigger an
1726 out-of-bounds system call table access. An untrusted local user
1727 could exploit this flaw to run code in the kernel
1728 (ie a root privilege escalation). (CVE-2007-4573).
1732 Description: O_DIRECT bypasses client statistics.
1733 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1734 read_ahead_stats were updated. copied the stats section from
1735 osc_send_oap_rpc() into async_internal().
1739 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1743 Description: change order of libsysio includes
1744 Details : '#include sysio.h' should always come before '#include xtio.h'
1748 Description: handle "serious error: objid * already exists" more gracefully
1749 Details : If LAST_ID value on disk is smaller than the objects existing in
1750 the O/0/d* directories, it indicates disk corruption and causes an
1751 LBUG(). If the object is 0-length, then we should use the existing
1752 object. This will help to avoid a full fsck in most cases.
1754 Severity : enhancement
1756 Description: adapt the lustre_config script to support the upgrade case
1757 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1758 server targets to 1.6 in parallel.
1762 Description: To avoid grant space > avaible space when the disk is almost
1763 full. Without this patch you might see the error "grant XXXX >
1764 available" or some LBUG about grant, when the disk is almost
1766 Details : In filter_check_grant, for non_grant cache write, we should
1767 check the left space by if (*left > ungranted + bytes), instead
1768 of (*left > ungranted), because only we are sure the left space
1769 is enough for another "bytes", then the ungrant space should be
1770 increase. In client, we should update cl_avail_grant only there
1771 is OBD_MD_FLGRANT in the reply.
1775 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1776 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1777 described in CVE-2007-4573. This problem could allow untrusted
1778 local users to gain root access.
1781 Frequency : when using O_DIRECT and quotas
1783 Description: Incorrect file ownership on O_DIRECT output files
1784 Details : block usage reported by 'lfs quota' does not take into account
1785 files that have been written with O_DIRECT.
1790 Description: touch file failed when fs is not full
1791 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1792 otherwise we can get ENOSP while fs is not full.
1796 Description: parallel lock callbacks
1797 Details : Instead of sending blocking and completion callbacks as separated
1798 requests, adding them to a set and sending in parallel.
1801 Frequency : only for Cray XT3
1802 Bugzilla : 12829/13455
1803 Description: Changing primary group doesn't change the group lustre assigns to
1805 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1806 overridden with the primary group provided by the group upcall,
1807 whereas the supplied fsgid can be trusted if it is in the list of
1808 supplementary groups returned by the group upcall.
1810 Severity : enhancement
1812 Description: Allow masking D_WARNING, D_ERROR messages from console
1813 Details : Console messages can now be disabled via lnet.printk.
1817 Description: User code with malformed file open parameter crashes client node
1818 Details : Before packing join_file req, all the related reference should be
1819 checked carefully in case some malformed flags cause fake
1820 join_file req on client.
1824 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1825 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1826 mode is equal to the granted mode. However, it should also check
1827 that LDLM_FL_CP_REQD is not set.
1831 Description: Heavy nfs access might result in deadlocks
1832 Details : After ELC code landed, it is now improper to enqueue any mds
1833 locks under och_sem, because enqueue might want to decide to
1834 cancel open locks for same inode we are holding och_sem for.
1838 Description: Client eviction while running blogbench
1839 Details : A lot of unlink operations with concurrent I/O can lead to a
1840 deadlock causing evictions. To address the problem, the number of
1841 oustanding OST_DESTROY requests is now throttled to
1842 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1843 callbacks are processed in priority.
1847 Description: enable ACLs on MDS by default
1848 Details : ACLs must be enabled on MDS by default.
1851 Frequency : PPC/PPC64 only
1853 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1854 Details : fix duplicated definitions between asm-ppc64/types.h and
1855 lustre_types.h on PPC.
1858 Frequency : PPC/PPC64 only
1860 Description: asm-ppc/segment.h does not exist
1861 Details : fix compile issue on PPC.
1865 Description: better handle error messages in extents code
1868 Frequency : RHEL4 only
1870 Description: mkfs is very slow on IA64/RHEL4
1871 Details : A performance regression has been discovered in the MPT Fusion
1872 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1873 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1874 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1876 Severity : enhancement
1878 Description: SNMP support enhancement
1879 Details : Adding total number of sampled request for an MDS node in snmp
1882 Severity : enhancement
1884 Description: Optimize ldlm waiting list processing for PR extent locks
1885 Details : When processing waiting list for read extent lock and meeting read
1886 lock that is same or wider to it that is not contended, skip
1887 processing rest of the list and immediatelly return current
1888 status of conflictness, since we are guaranteed there are no
1889 conflicting locks in the rest of the list.
1893 Description: Time out and refuse to reconnect
1894 Details : When the failover node is the primary node, it is possible
1895 to have two identical connections in imp_conn_list. We must
1896 compare not conn's pointers but NIDs, otherwise we can defeat
1897 connection throttling.
1901 Description: port llog fixes from b1_6 into HEAD
1902 Details : Port llog reference couting and some llog cleanups from b1_6
1903 (bug 10800) into HEAD, for protect from panic and access to already
1904 free llog structures.
1908 Description: Detect stride IO mode in read-ahead
1909 Details : When a client does stride read, read-ahead should detect that and
1910 read-ahead pages according to the detected stride pattern.
1914 Description: data checksumming impacts single node performance
1915 Details : add support for several checksum algorithm. Currently, only CRC32
1916 and Adler-32 are supported. The checksum type can be changed on
1917 the fly via /proc/fs/lustre/osc/*/checksum_type.
1921 Description: use adler32 for page checksums
1922 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1927 Description: build for x2 fails
1928 Details : fix compile issue on Cray systems.
1932 Description: Properly match for duplicate locks
1933 Details : Due to different lock order from skiplists code, we need to
1934 traverse entire list for now
1937 Frequency : only on PPC/SLES10
1939 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1940 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1941 be defined. Add a hack in mkfs_lustre.c to work around this compile
1946 Description: LASSERT on MDS when client holding flock lock dies
1947 Details : ldlm pool logic depends on number of granted locks equal to
1948 number of released locks which is not true for flock locks, so
1949 just exclude such locks from consideration.
1953 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1954 Details : Use fsfilt_sync() for both the whole filesystem sync and
1955 individual file sync to eliminate dangerous inode locking
1956 with I_LOCK that can lead to a deadlock.
1960 Description: performance in 1.6.3
1961 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1962 sure that 1MB requests are not fragmented by the block layer.
1964 Severity : enhancement
1966 Description: organize the server-side client stats on per-nid basis
1967 Details : Change the structure of stats under obdfilter and mds to
1975 The "uuid"s file would list the uuids of _active_ exports.
1976 And the clear entry is to clear all stats and stale nids.
1978 Severity : enhancement
1980 Description: eliminate client locks in face of contention
1981 Details : file contention detection and lockless i/o implementation
1982 for contended files.
1986 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1989 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1991 Description: MDS may lose file striping (and hence file data) in some cases
1992 Details : If there are additional extended attributes stored on the MDS,
1993 in particular ACLs, SELinux, or user attributes (if user_xattr
1994 is specified for the client mount options) then there is a risk
1995 of attribute loss. Additionally, the Lustre file striping
1996 needs to be larger than default (e.g. striped over all OSTs),
1997 and an additional attribute must be stored initially in the
1998 inode and then increase in size enough to be moved to the
1999 external attribute block (e.g. ACL growing in size) for file
2004 Description: skiplist implementation simplification
2005 Details : skiplists are used to group compatible locks on granted list
2006 that was implemented as tracking first and last lock of each
2007 lock group the patch changes that to using doubly linked lists
2011 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2012 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2013 because LASSERT on both the data supplied by a client, and the
2014 data on disk is dangerous and incorrect.
2016 Severity : enhancement
2018 Description: Slow truncate/writes to huge files at high offsets.
2019 Details : Directly associate cached pages to lock that protect those pages,
2020 this allows us to quickly find what pages to write and remove
2021 once lock callback is received.
2025 Description: more ldlm soft lockups
2026 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2027 starve other threads from the resource lock for a long time in
2028 case of long waiting queue, so change the debug level from
2029 D_OTHER to the less frequently used D_INFO.
2031 Severity : enhancement
2033 Description: add -gid, -group, -uid, -user options to lfs find
2037 Description: Hung threads in invalidate_inode_pages2_range
2038 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2039 once one is completed. As a result, some RPCs are stuck in the
2040 queue and are never sent.
2044 Description: filter threads hungs on waiting journal commit
2045 Details : Cleanup filter group llog code, then only filter group llog will
2046 be only created in the MDS/OST syncing process.
2050 Description: Procfs and llog threads access destoryed import sometimes.
2051 Details : Sync the import destoryed process with procfs and llog threads by
2052 the import refcount and semaphore.
2054 Severity : enhancement
2056 Description: openlock cache of b1_6 port to HEAD
2061 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2062 Details : If insertion of an extent fails, then discard the inode
2063 preallocation and free data blocks else it can lead to duplicate
2068 Description: don't always update ctime in ext3_xattr_set_handle()
2069 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2070 In some cases the ctime should not be updated, for example for
2071 2.0->1.8 compatibility it is necessary to delete an xattr and it
2072 should not update the ctime.
2076 Bugzilla : 15713/16362
2077 Description: Assertion in iopen_connect_dentry in 1.6.3
2078 Details : looking up an inode via iopen with the wrong generation number can
2079 populate the dcache with a disconneced dentry while the inode
2080 number is in the process of being reallocated. This causes an
2081 assertion failure in iopen since the inode's dentry list contains
2082 both a connected and disconnected dentry.
2086 Description: assertion failure in ldlm_handle2lock()
2087 Details : fix a race between class_handle_unhash() and class_handle2object()
2088 introduced in lustre 1.6.5 by bug 13622.
2093 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2094 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2099 Description: oops in sd_iostats_seq_show()
2100 Details : unloading/reloading the scsi low level driver triggers a kernel
2101 bug when trying to access the sd iostat file.
2106 Description: Kernel panics during QLogic driver reload
2107 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2108 patch, causing memory corruption.
2113 Description: journal_dev option does not work in b1_6
2114 Details : pass mount option during pre-mount.
2116 Severity : enhancement
2118 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2119 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2120 extent information of a file. It can be used to map logical blocks
2121 in a file to physical blocks in the block device.
2125 Description: LDLM soft lockups - improvement
2126 Details : It is be possible to send the lock handle along with each read
2127 or write request because the client is already doing a lock match
2128 itself so there isn't any reason the OST should have to re-do that
2132 Frequency : only on Cray X2
2134 Description: X2 build failures
2135 Details : fix build failures on Cray X2.
2139 Description: xid & resent requests
2140 Details : Initialize RPC XID from clock at startup (randomly if clock is
2143 Severity : enhancement
2145 Description: Add lustre_start utility to start or stop multiple Lustre servers
2150 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2151 Details : In case of memory pressure, list_del() can be called twice on
2152 req->rq_history_list, causing a kernel oops.
2156 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2157 Details : fix stack overflow in the distributed lock manager by defering
2158 export eviction after a failed AST to the elt thread instead of
2159 handling it in the dlm interpret routine.
2163 Description: Convert some comments to new format.
2164 Details : Update documenting comments to match doxygen conventions.
2168 Description: Grammar fixes.
2169 Details : A couple of trivial spelling fixes.
2173 Description: OSD_COUNTERS-mandatory
2174 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2175 to simplify the code.
2179 Description: simplify cmm_device freeing logic.
2180 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2184 Description: Add lockdep support to dt_object_operations locking interface.
2185 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2186 parameter indicating lock ordering. Update mdd code to use new
2191 Description: Introduce failloc constants for lockless IO tests.
2192 Details : Add two new failloc constants to test lockless IO. Only one of
2193 them in implemented---another is checked in yet to be landed
2198 Description: Add lockdep support for inode mutex.
2199 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2200 in the situations where Lustre has to lock more than one inode
2205 Description: Add optional invariants checking support.
2206 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2207 configure switch. This macro is to be used for consistency and
2208 sanity checks that are too expensive to be left in `production'
2213 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2214 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2219 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2220 Details : Use spin_lock_nested() in (the only) situation where more than
2221 one ldlm_lock is locked simultaneously. Also, fix possible
2222 dead-lock in ldlm_lock_change_resource() by enforcing particular
2227 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2228 Details : Instead of specifying each ldlm_lock call-back through separate
2229 parameter, wrap them into struct ldlm_callback_suite.
2233 Description: Kill join_lru obd method and its callers.
2234 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2235 in memory---a requirement implemented through ->o_join_lru() obd
2236 method in HEAD. Remove this method and its users.
2240 Description: Add asynchronous ldlm ENQUEUE completion handler.
2241 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2242 case that stock ldlm_completion_ast() cannot handle as it waits
2243 until lock is granted. Introduce new ldlm_completion_ast_async()
2244 for this. Also comment ldlm_completion_ast().
2248 Description: ldlm_error <-> errno conversion.
2249 Details : Add functions to map (rather arbitrary) between LDLM error codes
2250 and standard errno values. CLIO needs this to prevent LDLM specific
2251 constants from escaping ldlm and osc.
2255 Description: Kill unused ldlm_handle2lock_ns() function.
2256 Details : Kill unused ldlm_handle2lock_ns() function.
2260 Description: Add lu_ref support to ldlm_lock
2261 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2262 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2263 and ldlm_resource. LDLM interface has to be changed, because code that
2264 releases a reference on a lock, has to "know" what reference this is.
2265 In the most frequent case
2267 lock = ldlm_handle2lock(handle);
2269 LDLM_LOCK_PUT(lock);
2271 no changes are required. When any other reference (received _not_ from
2272 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2273 instead of LDLM_LOCK_PUT().
2275 Arguably, changes are pervasive, and interface requires some discipline
2276 for proper use. On the other hand, it was very instrumental in finding
2277 a few leaked lock references.
2281 Description: Add ldlm_lock_addref_try().
2282 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2283 attempts to addref a lock that might be being canceled
2288 Description: Add ldlm_weigh_callback().
2289 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2290 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2291 hard-coded `number of pages' logic.
2295 Description: Add lockdep annotations to llog code.
2296 Details : Use appropriately tagged _nested() locking calls in the places
2297 where llog takes more than one ->lgh_lock lock.
2301 Description: Add loi_kms_set().
2302 Details : Wrap kms updates into a helper function.
2306 Description: Constify instances of struct lsm_operations.
2307 Details : Constify instances of struct lsm_operations.
2311 Description: lu_conf support.
2312 Details : On a server, a file system object is uniquely identified
2313 by a fid, which is sufficient to locate and load all object
2314 state (inode). On a client, on the other hand, more data are
2315 necessary instantiate an object. Change lu_object_find() and
2316 friends to take additional `lu_conf' argument describing object.
2317 Typically this includes layout information.
2321 Description: lu_context fixes.
2322 Details : Introduce new lu_context functions that are needed on the client
2323 side, where some system threads (ptlrpcd) are shared by multiple
2324 modules, and so cannot be stopped during module shutdown.
2328 Description: Add start and stop methods to lu_device_type_operations.
2329 Details : Introduce two new methods in lu_device_type_operations, that are
2330 invoked when first instance of a given type is created and last one
2331 is destroyed respectively. This is need by CLIO.
2335 Description: Add lu_ref support to struct lu_device.
2336 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2337 track leaked references.
2341 Description: Introduce lu_kmem_descr.
2342 Details : lu_kmem_descr and its companion interface allow to create
2343 and destroy a number of kmem caches at once.
2347 Description: Fix lu_object finalization race.
2348 Details : Fix a race between lu_object_find() finding an object and its
2349 concurrent finalization. This race is (most likely) not possible
2350 on the server, but might happen on the client.
2354 Description: Introduce lu_ref interface.
2355 Details : lu_ref is a debugging module allowing to track references to
2356 a given object. It is quite cpu expensive, and has to be
2357 explicitly enabled with --enable-lu_ref. See usage description
2362 Description: Factor lu_site procfs stats into a separate function.
2363 Details : Separate lu_site stats printing code into a separate function
2364 that can be reused on a client.
2368 Description: Constify instances of struct {lu,dt,md}_device_operations.
2369 Details : Constify instances of struct {lu,dt,md}_device_operations.
2373 Description: Introduce struct md_site and move meta-data specific parts of
2374 struct lu_site here.
2375 Details : Move md-specific fields out of struct lu_site into special struct
2376 md_site, so that lu_site can be used on a client.
2380 Description: Kill mdd_lov_destroy().
2381 Details : Remove unused mdd code.
2385 Description: Add st_block checking to multistat.c.
2386 Details : Add st_block checking to multistat.c.
2390 Description: Add lu_ref support to struct obd_device.
2391 Details : Add lu_ref tracking to obd_device.
2395 Description: Kill obd_set_fail_loc().
2396 Details : Remove unused code.
2400 Description: Add special type for ptlrpc_request interpret functions.
2401 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2402 there. Actual usage will be in CLIO.
2406 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2407 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2408 doesn't work with lockdep.
2412 Description: Add rwv.c test program.
2413 Details : New testing program exercising readv(2) and writev(2) (Qian).
2417 Description: Add sendfile.c test program.
2418 Details : New testing program exercising sendfile(2) (Jay).
2422 Description: Ratelimit a message that can be very frequent.
2423 Details : Ratelimit a memory allocation failure message that can
2428 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2429 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2433 Description: Liblustre build fixes.
2434 Details : Liblustre build fixes.
2438 Description: libcfs: add cfs_{need,cond}_resched() interface.
2439 Details : libcfs: add cfs_{need,cond}_resched() definition and
2440 implementations for Linux, NT, and liblustre.
2442 Severity : enhancement
2444 Description: More exported tunables for mballoc
2445 Details : Add support for tunable preallocation window and new tunables for
2446 large/small requests
2450 Description: Detect corruption of block bitmap and checking for preallocations
2451 Details : Checks validity of on-disk block bitmap. Also it does better
2452 checking of number of applied preallocations. When corruption is
2453 found, it turns filesystem readonly to prevent further corruptions.
2457 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2458 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2459 window to 0. If it does hit cache miss, and read-pattern is still
2460 stride-io mode, does not reset the stride window, but also does not
2461 increase the stride window length in this case.
2465 Frequency : only for big-endian servers
2466 Description: Check if system is big-endian while mounting fs with extents feature
2467 Details : Mounting a filesystem with extents feature will fail on big-endian
2468 systems since ext3-based ldiskfs is not supported on big-endian
2469 systems. This can be over-riden with "bigendian_extents" mount option.
2471 Severity : enhancement
2473 Description: The root squash functionality
2474 Details : A security feature, which is to prevent users from being able
2475 to mount lustre on their desktop, run as root, and delete
2476 all of the files in the filesystem. The goal is accomplished by
2477 remapping user id (UID) and group id (GID) of the root user to
2478 a UID and GID specified by the system administartor via Lustre
2479 configuration management server (MGS). The functionality also
2480 allows to specify sets of clients for which the remapping does
2485 Description: Excessive recovery window
2486 Details : With AT enabled, the recovery window can be excessively long (6000+
2487 seconds). To address this problem, we no longer use
2488 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2489 timeout no longer depends on the service time, it is set to
2490 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2491 time via pb_service_time.
2495 Description: Watchdog triggered on MDS failover
2496 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2497 the OSTs know that the MDS "UUID" can be reused for the same export
2498 from a different NID, so we do not need to wait for the export to be
2502 Frequency : rare, only if using MMP with Linux RAID
2504 Description: MMP doesn't work with Linux RAID
2505 Details : While using HA for Lustre servers with Linux RAID, it is possible
2506 that MMP will not detect multiple mounts. To make this work we
2507 need to unplug the device queue in RAID when the MMP block is being
2508 written. Also while reading the MMP block, we should read it from
2509 disk and not the cached one.
2511 Severity : enhancement
2513 Description: open file using fid
2514 Details : A file can be opened using just its fid, like
2515 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2518 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2521 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2522 Details : add proper locking for creation of jbd_4k slab cache
2526 Description: MMP check in ext3_remount() fails without displaying any error
2527 Details : When multiple mount protection fails during remount, proper error
2530 Severity : enhancement
2532 Description: Allow stripe size to be up to 4G-64k
2533 Details : Fix math logic to allow large stripe sizes.
2537 Description: add check for >8TB ldiskfs filesystems
2538 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2539 Don't allow >8TB ldiskfs filesystems to be mounted without
2540 force_over_8tb mount option
2545 Descriptoin: OSS complains "too many blocks" and crashes
2546 Details : In filter_commitrw_write():
2547 1. lquota_chkquota();
2548 2. obdo_from_inode();
2549 3. lquota_pending_commit();
2550 "2." has changed uid and gid in this case, then caused
2551 lquota_chkquota() and lquota_pending_commit() to use
2552 different uid and gid. That is the root of the bug.
2556 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2557 Details : Detect and print out the 'stripe count = -1' case properly.
2561 Description: Out or order replies might be lost on replay
2562 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2563 request with tid smaller then one currently being added, add it
2564 to the start, not end of the list.
2566 --------------------------------------------------------------------------------
2568 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2570 * Support for kernels:
2571 2.6.5-7.283 (SLES 9),
2572 2.6.9-55.EL (RHEL 4),
2573 2.6.16.46-0.14 (SLES 10),
2574 2.6.18.8 vanilla (kernel.org)
2575 * Client support for unpatched kernels:
2576 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2577 2.6.16 - 2.6.22 vanilla (kernel.org)
2578 * Due to recently discovered recovery problems, we do not recommend
2579 using patchless RHEL 4 clients with this or any earlier release.
2580 * Recommended e2fsprogs version: 1.39.cfs8
2581 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2582 * Starting with this release, the ldiskfs backing filesystem required
2583 by Lustre is now in its own package, lustre-ldiskfs. This package
2584 should be installed. It is versioned separately from Lustre and
2585 may be released separately in future.
2589 Description: block reactivating mgc import until all deactivates complete
2590 Details : Fix race when failing back MDT/MGS to itself (testing)
2593 Frequency : at statup only
2595 Description: mds_lov_synchronize race leads to various problems
2596 Details : simultaneous MDT->OST connections at startup can cause the
2597 sync to abort, leaving the OSC in a bad state.
2599 Severity : enhancement
2601 Description: add optional extra BUILD_VERSION info
2602 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2603 to override the lustre version.
2606 Frequency : 2.6.18 servers only
2608 Description: ll_kern_mount() doesn't release the module reference
2609 Details : The ldiskfs module reference count never drops down to 0
2610 because ll_kern_mount() doesn't release the module reference.
2615 Description: server LBUG when using old ost_num_threads parameter
2616 Details : Accept the old ost_num_threads parameter but warn that it
2617 is deprecated, and fix an off-by-one error that caused an LBUG.
2622 Description: Transient SCSI error results in persistent IO issue
2623 Details : iobuf->dr_error is not reinitialized to 0 between two
2627 Frequency : sometimes when underlying device returns I/O errors
2629 Description: OSTs not going read-only during write failures
2630 Details : OSTs are not remounted read-only when the journal commit threads
2631 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2632 instead of the ext3 wrappers.
2636 Description: poor connect scaling with increasing client count
2637 Details : Don't run filter_grant_sanity_check for more than 100 exports
2638 to improve scaling for large numbers of clients.
2641 Frequency : SLES10 only
2643 Description: sanity-quota.sh quotacheck failed: rc = -22
2644 Details : Quotas cannot be enabled on SLES10.
2647 Frequency : liblustre clients only
2649 Description: getdirentries does not give error when run on compute nodes
2650 Details : getdirentries does not fail when the size specified as an argument
2651 is too small to contain at least one entry
2653 Severity : enhancement
2655 Description: Add LNET router traceability for debug purposes
2656 Details : If a checksum failure occurs with a router as part of the
2657 IO path, the NID of the last router that forwarded the bulk data
2658 is printed so it can be identified.
2663 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2664 Details : Due to a race condition, liblustre clients were occasionally
2665 evicted incorrectly.
2667 Severity : enhancement
2669 Description: lfs setstripe use optional parameters instead of postional
2672 Severity : enhancement
2674 Description: Nanosecond timestamp support for ldiskfs
2675 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2676 resolution timestamps. There is not yet support for this at
2677 the Lustre filesystem level.
2680 Frequency : during server recovery
2682 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2683 Details : request with rq_no_resend flag not awake l_wait_event if they get
2687 Frequency : nfs export on patchless client
2689 Description: connectathon hang when test nfs export over patchless client
2690 Details : Disconnected dentry cannot be found with lookup, so we do not need
2691 to unhash it or make it invalid
2694 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2695 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2696 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2701 Description: open req refcounting wrong on reconnect
2702 Details : If reconnect happened between getting open reply from server and
2703 call to mdc_set_replay_data in ll_file_open, we will schedule
2704 replay for unreferenced request that we are about to free.
2705 Subsequent close will crash in variety of ways.
2706 Check that request is still eligible for replay in
2707 mdc_set_replay_data().
2712 Description: disable writes to filesystem when reading health_check file
2713 Details : the default for reading the health_check proc file has changed
2714 to NOT do a journal transaction and write to disk, because this
2715 can cause reads of the /proc file to hang and block HA state
2716 checking on a healthy but otherwise heavily loaded system. It
2717 is possible to return to the previous behaviour during configure
2718 with --enable-health-write.
2720 Severity : enhancement
2722 Description: 64-bit inode version
2723 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2724 to the inode. This will be required for version-based recovery.
2729 Description: MDS fails to start if a duplicate client export is detected
2730 Details : in some rare cases it was possible for a client to connect to
2731 an MDS multiple times. Upon recovery the MDS would detect this
2732 and fail during startup. Handle this more gracefully.
2734 Severity : enhancement
2736 Description: Add -o localflock option to simulate old noflock
2738 Details : This will achieve local-only flock/fcntl locks
2744 Description: log_commit_thread vs filter_destroy race leads to crash
2745 Details : Take import reference before releasing llog record semaphore
2750 Description: Wrong request locking in request set processing
2751 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2752 imp_delayed_list, in this place should be used imp_lock.
2755 Frequency : when reconnection
2757 Description: Grant Leak when osc reconnect to OST
2758 Details : When osc reconnect ost, OST(filter) should check whether it
2759 should grant more space to client by comparing fed_grant and
2760 cl_avail_grant, and return the granted space to client instead
2761 of "new granted" space, because client will call osc_init_grant
2762 to update the client grant space info.
2765 Frequency : when client reconnect to OST
2767 Description: Grant Leak when osc do resend and replay bulk write
2768 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2769 bulk write request, because the grant info while be sync between
2770 OSC and OST when reconnect, and we should ignore the grant info
2771 these of resend/replay write req.
2776 Description: Grant space more than avaiable left space sometimes.
2777 Details : When then OST is about to be full, if two bulk writing from
2778 different clients came to OST. Accord the avaliable space of the
2779 OST, the first req should be permitted, and the second one
2780 should be denied by ENOSPC. But if the seconde arrived before
2781 the first one is commited. The OST might wrongly permit second
2782 writing, which will cause grant space > avaiable space.
2785 Frequency : when client is evicted
2787 Description: Grant might be wrongly erased when osc is evicted by OST
2788 Details : when the import is evicted by server, it will fork another
2789 thread ptlrpc_invalidate_import_thread to invalidate the
2790 import, where the grant will be set to 0. While the original
2791 thread will update the grant it got when connecting. So if
2792 the former happened latter, the grant will be wrongly errased
2793 because of this race.
2798 Description: Checking Stale with correct fid
2799 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2800 to check whether it is stale, because sometimes, we want the
2801 enqueue happened anyway, and op_data.fid2 will not be initialized.
2803 Severity : enhancement
2805 Description: update patchless client
2806 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2809 Frequency : only with 2.4 kernel
2811 Description: random memory corruption
2812 Details : size of struct ll_inode_info is to big for union inode.u and this
2813 can be cause of random memory corruption.
2818 Description: Memory leak in recovery
2819 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2820 It should also check obd_fail before fsfilt_start, otherwise if
2821 fsfilt_start return -EROFS,(failover mds during mds recovery).
2822 then the req will return with repmsg->transno = 0 and rc = EROFS.
2823 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2824 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2825 be freed no matter whether fsfilt_commit success or not.
2828 Frequency : only with huge count clients
2830 Description: Prevents from taking the superblock lock in llap_from_page for
2832 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2833 for llap_from_page call in ll_removepage prevents from taking the
2834 superblock lock for a soon died page.
2839 Description: Not check open intent error before release open handle
2840 Details : in some rare cases, the open intent error is not checked before
2841 release open handle, which may cause
2842 ASSERTION(open_req->rq_transno != 0), because it tries to release
2843 the failed open handle.
2848 Description: Set cat log bitmap only after create log success.
2849 Details : in some rare cases, the cat log bitmap is set too early. and it
2850 should be set only after create log success.
2854 Description: Accessing a block bevice can re-enable I/O when Lustre is
2855 tearing down a device.
2856 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2861 Description: service threads may hog cpus when there are a lot of requests
2862 Details : Insert cond_resched to give other threads a chance to use some CPU
2867 Description: the cat log was not initialized in recovery
2868 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2869 unlink log on mds will not be initialized until mds post
2870 recovery. And also in mds post recovery, the unlink log will
2871 initialization will be done asynchronausly, so there will be race
2872 between add unlink log and unlink log initialization.
2876 Description: brw_stats were being printed incorrectly
2877 Details : brw_stats were being printed as log2 but all of them were not
2878 recorded as log2. Also remove some code duplication arising from
2879 filter_tally_{read,write}.
2883 Frequency : rare, only in recovery.
2884 Description: ASSERTION(req->rq_type != LI_POISON) failed
2885 Details : imp_lock should be held while iterating over imp_sending_list for
2886 prevent destroy request after get timeout in ptlrpc_queue_wait.
2890 Description: replay-single.sh test 52 fails
2891 Details : A lock's skiplist need to be cleanup when it being unlinked
2892 from its resource list.
2896 Description: Short directio read returns full requested size rather than
2898 Details : Direct I/O operations should return actual amount of bytes
2899 transferred rather than requested size.
2901 Severity : enhancement
2903 Description: metadata RPC reduction (e.g. for rm performance)
2904 Details : decrease the amount of synchronous RPC between clients and servers
2905 by canceling conflicing lock before the operation on the client
2906 and packing thier handles into the main operation RPC to server.
2908 Severity : enhancement
2910 Description: Async OSC create to avoid the blocking unnecessarily.
2911 Details : If a OST has no remain object, system will block on the creating
2912 when need to create a new object on this OST. Now, ways use
2913 pre-created objects when available, instead of blocking on an
2914 empty osc while others are not empty. If we must block, we block
2915 for the shortest possible period of time.
2919 Description: improve handling recoverable errors
2920 Details : if request processig with error which can be recoverable on server
2921 request should be resend, otherwise page released from cache and
2924 Severity : enhancement
2926 Description: refine locking for avoid write wrong info into lov_objid file
2927 Details : fix possible races with add new target and write/update data in
2930 --------------------------------------------------------------------------------
2932 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2937 Frequency : on some architectures
2939 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2940 Details : Uninitialized flags sometimes cause configuration commands to
2944 Frequency : patchless clients only
2946 Description: missing __iget() symbol export
2947 Details : The __iget() symbol export is missing. To avoid the need for
2948 this on patchless clients the deathrow inode reaper is turned
2949 off, and we depend on the VM to clean up old inodes. This
2950 dependency was during via the fix for bug 12181.
2952 --------------------------------------------------------------------------------
2954 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2956 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2957 INTEROPERATE with older versions automatically. In many cases a
2958 special upgrade step is needed. Please read the
2959 user documentation before upgrading any part of a live system.
2960 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2961 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2962 * WARNING: Lustre configuration and startup changes are required with
2963 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2967 Severity : enhancement
2969 Description: MountConf
2970 Details : Lustre configuration is now managed via mkfs and mount
2971 commands instead of lmc and lconf. New obd types (MGS, MGC)
2972 are added for dynamic configuration management. See
2973 https://mail.clusterfs.com/wikis/lustre/MountConf for
2976 Severity : enhancement
2978 Description: dynamic OST addition
2979 Details : OSTs can now be added to a live filesystem
2981 Severity : enhancement
2983 Description: startup order invariance
2984 Details : MDTs and OSTs can be started in any order. Clients only
2985 require the MDT to complete startup.
2987 Severity : enhancement
2989 Description: parallel, asynchronous orphan cleanup
2990 Details : orphan cleanup is now performed in separate threads for each
2991 OST, allowing parallel non-blocking operation.
2993 Severity : enhancement
2995 Description: optimized stripe assignment
2996 Details : stripe assignments are now made based on ost space available,
2997 ost previous usage, and OSS previous usage, in order to try
2998 to optimize storage space and networking resources.
3000 Severity : enhancement
3002 Description: Permanently set tunables
3003 Details : All writable /proc/fs/lustre tunables can now be permanently
3004 set on a per-server basis, at mkfs time or on a live system.
3006 Severity : enhancement
3008 Description: Lustre message v2
3009 Details : Add lustre message format v2.
3011 Severity : enhancement
3013 Description: client OST exclusion list
3014 Details : Clients can be started with a list of OSTs that should be
3015 declared "inactive" for known non-responsive OSTs.
3019 Description: SPEC SFS validation failure on NFS v2 over lustre.
3020 Details : Changes the blocksize for regular files to be 2x RPC size,
3021 and not depend on stripe size.
3023 Severity : enhancement
3025 Description: Multiple MD RPCs in flight.
3026 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3027 To avoid overly-overloading MDS, introduce a limit on number of
3028 MDS RPCs in flight for a single client and add /proc controls
3029 to adjust this limit.
3031 Severity : enhancement
3033 Description: client read/write statistics
3034 Details : Add client read/write call usage stats for performance
3035 analysis of user processes.
3036 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3037 file access. extents_stats shows chunk size distribution.
3038 extents_stats_per_process show chunk size distribution per
3041 Severity : enhancement
3043 Description: mds statistics
3044 Details : Add detailed mds operations statistics in
3045 /proc/fs/lustre/mds/*/stats.
3049 Description: Failure of copying files with lustre special EAs.
3050 Details : Client side always return success for setxattr call for lustre
3051 special xattr (currently only "trusted.lov").
3056 Description: Refcount LNET uuids
3057 Details : The global LNET uuid list grew linearly with every startup;
3058 refcount repeated list entries instead of always adding to
3061 Severity : enhancement
3063 Description: Dynamic service threads
3064 Details : Within a small range, start extra service threads
3065 automatically when the request queue builds up.
3068 Frequency : mixed-endian client/server environments
3070 Description: mixed-endian crashes
3071 Details : The new msg_v2 system had some failures in mixed-endian
3074 Severity : enhancement
3076 Description: Easy OST removal
3077 Details : OSTs can be permanently deactivated with e.g. 'lctl
3078 conf_param lustre-OST0001.osc.active=0'
3080 Severity : enhancement
3082 Description: MGS proc entries
3083 Details : Added basic proc entries for the MGS showing what filesystems
3086 Severity : enhancement
3088 Description: provide MGS failover
3089 Details : Added config lock reacquisition after MGS server failover.
3091 Severity : enhancement
3093 Description: add Linux 2.4 support
3094 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3098 Description: a large application tries to do I/O to the same resource and dies
3099 in the middle of it.
3100 Details : Check the req->rq_arrival time after the call to
3101 ost_brw_lock_get(), but before we do anything about
3102 processing it & sending the BULK transfer request. This
3103 should help move old stale pending locks off the queue as
3104 quickly as obd_timeout.
3107 Frequency : when an incorrect nid is specified during startup
3109 Description: ptlrpc connect to non-existant node causes kernel crash
3110 Details : LNET can't be re-entered from an event callback, which
3111 happened when we expire a message after the export has been
3112 cleaned up. Instead, hand the zombie cleanup off to another
3115 Severity : enhancement
3117 Description: plain/inodebits lock performance improvement
3118 Details : Grouping plain/inodebits in granted list by their request modes
3119 and bits policy, thus improving the performance of search through
3123 Frequency : only if OST filesystem is corrupted
3125 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3126 Details : for a short time RPCs with bulk IO are in the replay list,
3127 but replay of bulk IOs is unimplemented. If the OST filesystem
3128 is corrupted due to disk cache incoherency and then replay is
3129 started it is possible to trip an assertion. Avoid putting
3130 committed RPCs into the replay list at all to avoid this issue.
3133 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3135 Description: System hang on startup
3136 Details : This bug allowed the liblustre (e.g. catamount) client to
3137 return to the app before handling all startup RPCs. This
3138 could leave the node unresponsive to lustre network traffic
3139 and manifested as a server ptllnd timeout.
3141 Severity : enhancement
3143 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3144 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3145 boolean to control whether to print peer debug info when a
3146 client's RPC times out.
3149 Frequency : only for kernels with patches from Lustre below 1.4.3
3151 Description: Remove old rdonly API
3152 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3155 Frequency : only for devices with external journals
3157 Description: Set external device read-only also
3158 Details : During a commanded failover stop, we set the disk device
3159 read-only while the server shuts down. We now also set any
3160 external journal device read-only at the same time.
3163 Frequency : when upgrading from 1.4 while trying to change parameters
3165 Description: The wrong (new) MDC name was used when setting parameters for
3166 upgraded MDT's. Also allows changing of OSC (and MDC)
3167 parameters if --writeconf is specified at tunefs upgrade time.
3170 Frequency : when setting specific ost indicies
3172 Description: QOS code breaks on skipped indicies
3173 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3174 created with --index need not be sequential.
3176 Severity : enhancement
3178 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3179 Details : The uninit_groups feature works in conjunction with the kernel
3180 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3181 up the pass1 processing of e2fsck. This is a read-only feature
3182 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3183 mount filesystems that have had this feature enabled.
3185 Severity : enhancement
3187 Description: Improve multi-block allocation algorithm to avoid fragmentation
3188 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3189 allocation locality and avoid filesystem fragmentation.
3191 ------------------------------------------------------------------------------
3193 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3195 * Support for kernels:
3196 2.4.21-47.0.1.EL (RHEL 3)
3197 2.6.5-7.283 (SLES 9)
3198 2.6.9-42.0.10.EL (RHEL 4)
3199 2.6.12.6 vanilla (kernel.org)
3200 2.6.16.27-0.9 (SLES 10)
3201 * Recommended e2fsprogs version: 1.39.cfs5
3203 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3207 Frequency : occasional, depends on client load and configuration
3208 Bugzilla : 12181, 12203
3209 Description: data loss for recently-modified files
3211 Details : In some cases it is possible that recently written or created
3212 files may not be written to disk in a timely manner (this should
3213 normally be within 30s unless client IO load is very high).
3214 The problem appears as zero-length files or files that are a
3215 multiple of 1MB in size after a client crash or client eviction
3216 that are missing data at the end of the file.
3218 This problem is more likely to be hit on clients where files are
3219 repeatedly created and unlinked in the same directory, clients
3220 have a large amount of RAM, have many CPUs, the filesystem has
3221 many OSTs, the clients are rebooted frequently, and/or the files
3222 are not accessed by other nodes after being written.
3224 The presence of the problem can be detected by looking at
3225 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3226 smaller than the second (nr_unused) then dirty files will not
3227 be flushed automatically to disk. "sync; sleep 10" should be
3228 run several times on the node before unmounting it to update
3229 Lustre (this is also safe to run on nodes without this problem).
3231 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3232 that can cause this same problem, so customers using that kernel
3233 also need to update the kernel in addition to Lustre. In order
3234 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3236 It is normal that files written just before a client crash (less
3237 than 30s) may not yet have been flushed to disk, even for local
3241 Frequency : frequent on thin XT3 nodes
3243 Description: UUID collision on thin XT3 Linux nodes
3244 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3245 randomly, since we relied on an insufficiently-seeded PRNG.
3250 Description: OSS hangs after "All ost request buffers busy"
3251 Details : A deadlock between quota and journal operations caused OSS
3252 hangs after printing "All ost request buffers busy."
3255 Frequency : always on liblustre builds
3257 Description: Cleanup compiler warnings on liblustre
3260 Frequency : always on liblustre builds on XT3
3262 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3266 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3267 > 32000 subdirectories
3268 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3269 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3270 subdirectory count crosses 32000. This will aid e2fsck to
3271 correctly handle more than 32000 subdirectories.
3274 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3276 Description: System hang on startup
3277 Details : This bug allowed the liblustre (e.g. catamount) client to
3278 return to the app before handling all startup RPCs. This
3279 could leave the node unresponsive to lustre network traffic
3280 and manifested as a server ptllnd timeout.
3282 Severity : enhancement
3284 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3285 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3286 boolean to control whether to print peer debug info when a
3287 client's RPC times out.
3292 Description: make O_SYNC working on 2.6 kernels
3293 Details : 2.6 kernels use different method for mark pages for write,
3294 so need add a code to lustre for O_SYNC work.
3299 Description: Failure to close file and release space on NFS
3300 Details : Put inode details into lock acquired in ll_intent_file_open.
3301 Use mdc_intent_lock in ll_intent_open to properly
3302 detect all kind of errors unhandled by mdc_enqueue.
3307 Description: proc file read during shutdown sometimes raced obd removal,
3309 Details : Add lock to prevent obd access after proc file removal.
3312 Frequency : Only for files larger than 4GB on 32-bit clients.
3314 Description: improperly doing page alignment of locks
3315 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3316 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3321 Description: RPCs being resent when they shouldn't be
3322 Details : Some RPCs that should not be resent are being resent. This
3323 can cause inconsistencies in the RPC state machine. Do not
3324 resend such requests.
3327 Frequency : rare, only with NFS export
3329 Description: Crash on NFS re-export node
3330 Details : under very unusual load conditions an assertion is hit in
3331 ll_intent_file_open()
3334 Frequency : only if OST filesystem is corrupted
3336 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3337 Details : for a short time RPCs with bulk IO are in the replay list,
3338 but replay of bulk IOs is unimplemented. If the OST filesystem
3339 is corrupted due to disk cache incoherency and then replay is
3340 started it is possible to trip an assertion. Avoid putting
3341 committed RPCs into the replay list at all to avoid this issue.
3346 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3347 Details : Large single O_DIRECT read and write calls can fail to allocate
3348 a sufficiently large buffer to process the request. In case of
3349 allocation failure the allocation is retried with a smaller
3350 buffer and broken into smaller requests.
3352 Severity : enhancement
3354 Description: Add -o localflock option to simulate old noflock behaviour.
3355 Details : This will achieve local-only flock/fcntl locks coherentness.
3360 Description: versioning check is incomplete
3361 Details : Checking the version difference of client vs. server, report
3362 error if the gap is too big.
3367 Description: add support PG_writeback bit
3368 Details : add support for PG_writeback bit for Lustre, for more carefull
3369 work with page cache in 2.6 kernel. This also fix some deadlocks
3370 and remove hack for work O_SYNC with 2.6 kernel.
3372 Severity : enhancement
3374 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3375 Details : The uninit_groups feature works in conjunction with the kernel
3376 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3377 up the pass1 processing of e2fsck. This is a read-only feature
3378 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3379 mount filesystems that have had this feature enabled.
3381 Severity : enhancement
3383 Description: Improve multi-block allocation algorithm to avoid fragmentation
3384 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3385 allocation locality and avoid filesystem fragmentation.
3387 ------------------------------------------------------------------------------
3389 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3391 * Support for kernels:
3392 2.6.9-42.0.3.EL (RHEL 4)
3393 2.6.5-7.276 (SLES 9)
3394 2.4.21-47.0.1.EL (RHEL 3)
3395 2.6.12.6 vanilla (kernel.org)
3396 2.6.16.21-0.8 (SLES10)
3397 * Recommended e2fsprogs version: 1.39.cfs2-0
3399 * The backwards-compatible /proc/sys/portals symlink has been removed
3400 in this release. Before upgrading, please ensure that you change
3401 any configuration scripts or /etc/sysctl.conf files that access
3402 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3403 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3404 in advance of the upgrade on any system running Lustre 1.4.6 or
3405 newer, since /proc/sys/lnet was added in that version.
3406 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3410 Frequency : only when quota is used
3412 Description: avoid scanning export list for quota master
3413 Details : Change the algorithms to avoid scanning export list in order
3414 to improve the efficiency.
3417 Frequency : MDS failover only, very rarely
3419 Description: "went back in time" messages on mds failover
3420 Details : The greatest transno may be lost when the current operation
3421 finishes with an error (transno==0) and the client's last_rcvd
3422 record is over-written. Save the greatest transno in the
3423 mds_last_transno for this case.
3426 Frequency : always for specific kernels and striping counts
3428 Description: client may get "Matching packet too big" without ACL support
3429 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3430 when trying to access files in certain configurations. The
3431 clients should in fact be denied when mounting because they do
3432 not understand ACLs.
3435 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3437 Description: many clients connecting with IO in progress causes connect timeouts
3438 Details : Avoid synchronous journal commits to avoid delays caused by many
3439 clients connecting/disconnecting when bulk IO is in progress.
3440 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3441 OST_IO_PORTAL to avoid delays behind potentially many pending
3445 Frequency : occasionally with multiple writers to a single file
3447 Description: shared writes to file may result in wrong size reported by stat()
3448 Details : Allow growing of kms when extent lock is cancelled
3451 Frequency : always with random mmap IO to multi-striped file
3453 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3454 Details : Only if the hole is at the end of OST object so that kms is too
3455 small. Fix is to increase kms accordingly in ll_nopage.
3458 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3460 Description: writes to a missing object would leak memory on the OST
3461 Details : If there is an inconsistency between the MDS and OST filesystems,
3462 such that the MDS references an object that doesn't exist, writes
3463 to that object will leak memory due to incorrect cleanup in the
3464 error handling path, eventually running out of memory on the OST.
3469 Description: Creating too long symlink causes lustre errors
3470 Details : Check symlink and name lengths before sending requests to MDS.
3473 Frequency : only if flock is enabled (not on by default)
3475 Description: posix locks not released on fd closure on 2.6.9+
3476 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3477 kernels, this caused such locks not to be released on fd close and
3478 then assertions on fs unmount about still used locks.
3481 Frequency : MDS failover only, very rarely
3483 Description: clients may get ASSERTION(granted_lock != NULL)
3484 Details : When request was taking a long time, and a client was resending
3485 a getattr by name lock request. The were multiple lock requests
3486 with the same client lock handle and
3487 mds_getattr_name->fixup_handle_for_resent_request found one of the
3488 lock handles but later failed with ASSERTION(granted_lock != NULL).
3493 Description: handle->h_buffer_credits > 0, assertion failure
3494 Details : h_buffer_credits is zero after truncate, causing assertion
3495 failure. This patch extends the transaction or creates a new
3499 Frequency : NFS re-export or patchless client
3500 Bugzilla : 11179, 10796
3501 Description: Crash on NFS re-export node (__d_move)
3502 Details : We do not want to hash the dentry if we don't have a lock.
3503 But if this dentry is later used in d_move, we'd hit uninitialised
3504 list head d_hash, so we just do this to init d_hash field but
3505 leave dentry unhashed.
3508 Frequency : NFS re-export or patchless client
3510 Description: NFS exports has problem with symbolic link
3511 Details : lustre client didn't properly install dentry when re-exported
3512 to NFS or running patchless client.
3515 Frequency : NFS re-export or patchless client
3517 Description: Various nfs/patchless fixes.
3518 Details : fixes reuse disconected alias for lookup process - this fixes
3519 warning "find_exported_dentry: npd != pd",
3520 fix permission error with open files at nfs.
3521 fix apply umask when do revalidate.
3524 Frequency : occasional
3526 Description: Crash on NFS re-export node
3527 Details : calling clear_page() on the wrong pointer triggered oops in
3528 generic_mapping_read().
3531 Frequency : rarely, using O_DIRECT IO
3533 Description: unaligned directio crashes client with LASSERT
3534 Details : check for unaligned buffers before trying any requests.
3537 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3539 Description: stale data returned from RAID cache
3540 Details : If only a small amount of IO is done to the RAID device before
3541 reading it again it is possible to get stale data from the RAID
3542 cache instead of reading it from disk.
3545 Frequency : always for sles10 kernel
3547 Description: sles10 support
3548 Details : ll_follow_link: compile fixes and using of nd_set_link
3549 under newer kernels.
3552 Frequency : depends on arch, kernel and compiler version, always on sles10
3555 Description: recursive or deep enough symlinks cause stack overflow
3556 Details : getting rid of large stack-allocated variable in
3560 Frequency : depends on hardware
3562 Description: lustre write performance loss in the SLES10 kernel
3563 Details : the performance loss is caused by using of write barriers in the
3564 ext3 code. The SLES10 kernel turns barrier support on by
3565 default. The fix is to undo that change for ldiskfs.
3567 ------------------------------------------------------------------------------
3569 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3571 * Support for kernels:
3572 2.6.9-42.0.3EL (RHEL 4)
3573 2.6.5-7.276 (SLES 9)
3574 2.4.21-47.0.1.EL (RHEL 3)
3575 2.6.12.6 vanilla (kernel.org)
3579 Frequency : quota enabled and large files being deleted
3581 Description: releasing more than 4GB of quota at once hangs OST
3582 Details : If a user deletes more than 4GB of files on a single OST it
3583 will cause the OST to spin in an infinite loop. Release
3584 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3589 Description: statfs data retrieved from /proc may be stale or zero
3590 Details : When reading per-device statfs data from /proc, in the
3591 {kbytes,files}_{total,free,avail} files, it may appear
3592 as zero or be out of date.
3595 Frequency : systems with MD RAID1 external journal devices
3597 Description: lconf's call to blkid is confused by RAID1 journal devices
3598 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3599 of returning all block devices that match the journal UUID.
3602 Frequency : always, for aggregate stripe size over 4GB
3604 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3605 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3606 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3609 Frequency : always if "lfs find" used on a local file/directory
3611 Description: "lfs find" segfaults if used on a local file/directory
3612 Details : The case where a directory component was not specified wasn't
3613 handled correctly. Handle this properly.
3616 Frequency : always on ppc64
3618 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3619 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3620 perform endianness conversion.
3623 Frequency : rarely (truncate to non-zero file size after write under load)
3624 Bugzilla : 10730, 10687
3625 Description: Files padded with zeros to next 4K multiple
3626 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3627 it is possible that files that are truncated to a non-zero size
3628 immediately after being written are filled with zero bytes beyond
3629 the truncated size. No file data is lost.
3631 Severity : enhancement
3633 Description: Allow recovery/failover for liblustre clients.
3634 Details : liblustre clients were unaware of failover configurations until
3637 Severity : enhancement
3639 Description: user file locks should fail when not mounting with flock option
3640 Details : Set up an error-returning stub in ll_file_operations.lock field
3641 to prevent incorrect behaviour when client is mounted without
3642 flock option. Also, set up properly f_op->flock field for
3646 Frequency : always on ia64
3648 Description: "lfs df" loops on printing out MDS statfs information
3649 Details : The obd_ioctl_data was not initialized and in some systems
3650 this caused a failure during the ioctl that did not return
3651 an error. Initialize the struct and return an error on failure.
3654 Frequency : SLES 9 only
3656 Description: Error of copying files with lustre special EAs as root
3657 Details : Client side always return success for setxattr call for lustre
3658 special xattr (currently only "trusted.lov").
3661 Frequency : rarely on clusters with both ia64+i386 clients
3663 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3664 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3665 clients are doing concurrent writes to the same file it is
3666 possible that the ia64 clients may LASSERT because the OST
3667 extent locks are not PAGE_SIZE aligned. Ensure that grown
3668 locks are always aligned on the request boundary.
3671 Frequency : specific use, occasional
3673 Description: Overwriting in use executable truncates on-disk binary image
3674 Details : If one node attempts to overwrite an executable in use by
3675 another node, we now correctly return ETXTBSY instead of
3676 truncating the file.
3678 Severity : enhancement
3680 Description: Async OSC create to avoid the blocking unnecessarily.
3681 Details : If a OST has no remain object, system will block on the creating
3682 when need to create a new object on this OST. Now, ways use
3683 pre-created objects when available, instead of blocking on an
3684 empty osc while others are not empty. If we must block, we block
3685 for the shortest possible period of time.
3690 Description: chmod on Lustre root is propagated to other clients
3691 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3692 invalid by the follow_mount time.
3695 Frequency : liblustre clients only
3697 Description: Race in 'instant cancel' lock handling could lead to such locks
3698 never to be granted in case of SMP MDS
3699 Details : Do not destroy not yet granted but cbpending locks in
3703 Frequency : replay/resend of open
3705 Description: non null lock assetion failure in mds_intent_policy
3706 Details : Trying to replay/resend lockless open requests resulted in
3707 mds_open() returning 0 with no lock. Now it sets a flag if
3708 a lock is going to be returned.
3710 Severity : enhancement
3712 Description: Checksum enhancements
3713 Details : New checksum enhancements allow for resending RPCs that failed
3716 Severity : enhancement
3718 Description: Tunables on number of dirty pages in cacche
3719 Details : Allow to set limit on number of dirty pages cached.
3724 Description: client crash on unmount - lock still has references
3725 Details : In some error handling cases it was possible to leak a lock
3726 reference on a client while accessing a file. This was not
3727 harmful to the client during operation, but would cause the
3728 client to crash when the filesystem is unmounted.
3731 Frequency : specific case, rare
3733 Description: ETXTBSY on mds though file not in use by client
3734 Details : ETXTBSY is no longer incorrectly returned when attempting to
3735 chmod or chown a directory that the user previously tried to
3736 execute or a currently-executing binary.
3739 Frequency : extremely rare except on liblustre-based clients
3741 Description: Lustre space not freed when files are deleted
3742 Details : Clean up open-unlinked files after client eviction. Previously
3743 the unlink was skipped and the files remained as orphans.
3748 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3749 Details : In some cases it was possible to send a blocking callback to
3750 a client doing a glimpse, even though that client didn't get
3751 a lock granted. When the glimpse lock is cancelled on the OST
3752 the freed lock is left on the waiting list and corrupted the list.
3755 Frequency : all core dumps
3757 Description: Broke core dumps to lustre
3758 Details : Negative dentry may be unhashed if parent does not have UPDATE
3759 lock, but some callers, e.g. do_coredump, expect dentry to be
3760 hashed after successful create, hash it in ll_create_it.
3762 ------------------------------------------------------------------------------
3764 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3766 * Support for kernels:
3767 2.6.9-42.0.2.EL (RHEL 4)
3768 2.6.5-7.276 (SLES 9)
3769 2.4.21-40.EL (RHEL 3)
3770 2.6.12.6 vanilla (kernel.org)
3774 Frequency : always on RHEL 3
3776 Description: Number of open files grows over time
3777 Details : The number of open files grows over time, whether or not
3778 Lustre is started. This was due to a filp leak introduced
3779 by one of our kernel patches.
3781 ------------------------------------------------------------------------------
3783 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3785 * Support for kernels:
3786 2.6.9-42.EL (RHEL 4)
3787 2.6.5-7.276 (SLES 9)
3788 2.4.21-40.EL (RHEL 3)
3789 2.6.12.6 vanilla (kernel.org)
3794 Bugzilla : 5719, 9635, 9792, 9684
3795 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3796 Details : If a server is under heavy load and cannot reply to new
3797 connection requests before the client resends the (re)connect,
3798 the connection handling code can behave badly if two service
3799 threads are concurrently handing separate (re)connections from
3800 the same client. Add better locking to the connection handling
3801 code, and ensure that only a single connection will be processed
3802 for a given client UUID, even if the lock is dropped.
3804 Severity : enhancement
3806 Description: add TCP zero-copy support to kernel
3807 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3808 sends if the hardware supports scatter-gather and checksumming.
3809 This allows socklnd to do client-write and server-read more
3810 efficiently and reduce CPU utilization from skbuf copying.
3813 Frequency : only if NFS exporting from client
3815 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3816 Details : If setting attributes on a file created under NFS that had
3817 never been opened it would be possible to oops the client
3818 if the file had no objects.
3821 Frequency : always for liblustre
3823 Description: liblustre client does MDS+OSTs setattr RPC for each write
3824 Details : When doing a write from a liblustre client, the client
3825 incorrectly issued an RPC to the MDS and each OST the file was
3826 striped over in order to update the timestamps. When writing
3827 with small chunks and many clients this could overwhelm the MDS
3828 with RPCs. In all cases it would slow down the write because
3829 these RPCs are unnecessary.
3831 Severity : enhancement
3833 Description: allow number of MDS service threads to be changed at module load
3834 Details : It is now possible to change the number of MDS service threads
3835 running. Adding "options mds mds_num_threads={N}" to the MDS's
3836 /etc/modprobe.conf will set the number of threads for the next
3837 time Lustre is restarted (assuming the "mds" module is also
3838 reloaded at that time). The default number of threads will
3839 stay the same, 32 for most systems.
3844 Description: OST crash if filesystem is unformatted or corrupt
3845 Details : If an OST is started on a device that has never been formatted
3846 or if the filesystem is corrupt and cannot even mount then the
3847 error handling cleanup routines would dereference a NULL pointer.
3852 Description: NULL pointer deref in llap_from_page.
3853 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3854 incorrect) mapping if the page was truncated/reclaimed while it was
3855 searched for. Check for this condition and skip such pages when
3856 doing readahead. Introduce extra check to llap_from_page() to
3857 verify page->mapping->host is non-NULL (so page is not anonymous).
3860 Frequency : Sometimes when using sys_sendfile
3862 Description: "page not covered by a lock" warnings from ll_readpage
3863 Details : sendfile called ll_readpage without right page locks present.
3864 Now we introduced ll_file_sendfile that does necessary locking
3865 around call to generic_file_sendfile() much like we do in
3869 Frequency : with certain MDS communication failures at client mount time
3871 Description: NULL pointer deref after failed client mount
3872 Details : a client connection request may delayed by the network layer
3873 and not be sent until after the PTLRPC layer has timed out the
3874 request. If the client fails the mount immediately it will try
3875 to clean up before the network times out the request. Add a
3876 reference from the request import to the obd device and delay
3877 the cleanup until the network drops the request.
3880 Frequency : occasionally during client (re)connect
3882 Description: assertion failure during client (re)connect
3883 Details : processing a client connection request may be delayed by the
3884 client or server longer than the client connect timeout. This
3885 causes the client to resend the connection request. If the
3886 original connection request is replied in this interval, the
3887 client may trip an assertion failure in ptlrpc_connect_interpret()
3888 which thought it would be the only running connect process.
3891 Frequency : only with obd_echo servers and clients that are rebooted
3893 Description: kernel BUG accessing uninitialized data structure
3894 Details : When running an obd_echo server it did not start the ping_evictor
3895 thread, and when a client was evicted an uninitialized data
3896 structure was accessed. Start the ping_evictor in the RPC
3897 service startup instead of the OBD startup.
3899 Severity : enhancement
3900 Bugzilla : 10193 (patchless)
3901 Description: Remove dependency on various unexported kernel interfaces.
3902 Details : No longer need reparent_to_init, exit_mm, exit_files,
3903 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3906 Frequency : rare (only users of deprecated and unsupported LDAP config)
3908 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3909 Details : LDAP apparently contains 'lustreName' attributes instead of
3910 'name'. A simple remapping of the name is sufficient.
3913 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3915 Description: waiting_locks_callback trips kernel BUG if client is evicted
3916 Details : Running with the dump_on_timeout debug flag turned on makes
3917 it possible that the waiting_locks_callback() can try to dump
3918 the Lustre kernel debug logs from an interrupt handler. Defer
3919 this log dumping to the expired_lock_main() thread.
3921 Severity : enhancement
3923 Description: Support NFS exporting on 2.6 kernels.
3924 Details : Implement non-rawops metadata methods for NFS server to use without
3925 changing NFS server code.
3928 Frequency : very rare (synthetic metadata workload only)
3930 Description: two racing renames might cause an MDS thread to deadlock
3931 Details : Running the "racer" program may cause one MDS thread to rename
3932 a file from being the source of a rename to being the target of
3933 a rename at exactly the same time that another thread is doing
3934 so, and the second thread has already enqueued these locks after
3935 doing a lookup of the target and is trying to relock them in
3936 order. Ensure that we don't try to re-lock the same resource.
3939 Frequency : only very large systems with liblustre clients
3941 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3942 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3943 all OSTs in parallel. This allows the eviction of stale liblustre
3944 clients to proceed much faster than if they were done in series,
3945 and also offers similar improvements for other set_info RPCs.
3950 Description: excessive CPU usage during initial read phase on client
3951 Details : During the initial read phase on a client, it would agressively
3952 retry readahead on the file, consuming too much CPU and impacting
3953 performance (since 1.4.5.8). Improve the readahead algorithm
3954 to avoid this, and also improve some other common cases (read
3955 of small files in particular, where "small" is files smaller than
3956 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3961 Description: MDS crash when receiving packet with unknown intent.
3962 Details : Do not LBUG in unknown intent case, just return -EFAULT
3964 Severity : enhancement
3965 Bugzilla : 9293, 9385
3966 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3967 Details : Do not serialize getattr (non-intent version) and statfs.
3970 Frequency : occasional, when OST network is overloaded/intermittent
3972 Description: client evicted by OST after bulk IO timeout
3973 Details : If a client sends a bulk IO request (read or write) the OST
3974 may evict the client if it is unresposive to its data GET/PUT
3975 request. This is incorrect if the network is overloaded (takes
3976 too long to transfer the RPC data) or dropped the OST GET/PUT
3977 request. There is no need to evict the client at all, since
3978 the pinger and/or lock callbacks will handle this, and the
3979 client can restart the bulk request.
3982 Frequency : Always when mmapping file with no objects
3984 Description: client crashes when mmapping file with no objects
3985 Details : Check that we actually have objects in a file before doing any
3986 operations on objects in ll_vm_open, ll_vm_close and
3992 Description: Request leak when working with deleted CWD
3993 Details : Introduce advanced request refcount tracking for requests
3994 referenced from lustre intent.
3996 Severity : Enhancement
3998 Description: Cache open file handles on client.
3999 Details : MDS now will return special lock along with openhandle, if
4000 requested and client is allowed to hold openhandle, even if unused,
4001 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4002 closing files for every read/write openration.
4004 Severity : Enhancement
4006 Description: Cache open negative dentries on client when possible.
4007 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4008 negative dentries on lock revocation.
4013 Description: Remounting a client read-only wasn't possible with a zconf mount
4014 Details : It wasn't possible to remount a client read-only with llmount.
4016 Severity : enhancement
4017 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4018 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4019 MPICH is included to add Lustre-specific ADIO interfaces.
4020 This is based closely on the UFS ADIO layer and only differs
4021 in file creation, in order to allow the OST striping to be set.
4022 This is user-contributed code and not supported by CFS.
4027 Description: extended inode attributes (immutable, append-only) work improperly
4028 when 2.4 and 2.6 kernels are used on client/server or vice versa
4029 Details : Introduce kernel-independent values for these flags.
4031 Severity : enhancement
4034 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4035 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4036 granularity. Now, Fractional MB granularity is be supported,
4037 this is very useful for low memory system.
4039 Severity : enhancement
4041 Description: Getattr by fid
4042 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4043 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4044 have locks and avoids getattr rpc requests for it on every stat.
4047 Frequency : Always, for filesystems larger than 2TB
4049 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4050 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4051 filesystems larger than 2TB to be mounted with mballoc enabled.
4054 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4056 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4057 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4058 filesystems larger than 2TB is mounted then the kernel will
4059 silently corrupt the start of the filesystem. CONFIG_LBD is
4060 enabled for all CFS-supported kernels, but the possibility of
4061 this happening with a modified kernel config exists.
4063 Severity : enhancement
4065 Description: add client O_DIRECT support for 2.6 kernels
4066 Details : It is now possible to do O_DIRECT reads and writes to files
4067 in the Lustre client mountpoint on 2.6 kernel clients.
4069 Severity : enhancement
4071 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4072 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4073 parallel, not waiting for response from every OST before sending
4074 a rpc to the next OST.
4079 Description: setattr vs write race when updating file timestamps
4080 Details : Client processes that update a file timestamp into the past
4081 right after writing to the file (e.g. tar) it is possible that
4082 the updated file modification time can be reset to the current
4083 time due to a race between processing the setattr and write RPC.
4085 Severity : enhancement
4087 Description: Bring 'lfs find' closer in line with regular Linux find.
4088 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4089 -print0 options and obtains all the needed info through the lustre
4092 Severity : enhancement
4094 Description: support up to 1024 configured devices on one node
4095 Details : change obd_dev array from statically allocated to dynamically
4096 allocated structs as they are first used to reduce memory usage
4101 Description: Flush dirty partially truncated pages during truncate
4102 Details : Immediatelly flush partially truncated pages in filter_setattr,
4103 this way we completely avoid having any pages in page cache on OST
4104 and can retire ugly workarounds during writes to flush such pages.
4109 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4110 Details : protect inode from truncation within vfs_unlink() context
4111 just take a reference before calling vfs_unlink() and release it
4112 when parent's i_sem is free.
4117 Description: last_id value checked outside lock on OST caused LASSERT failure
4118 Details : If there were multiple MDS->OST object precreate requests in
4119 flight, it was possible that the OST's last object id was checked
4120 outside a lock and incorrectly tripped an assertion. Move checks
4121 inside locks, and discard old precreate requests.
4124 Frequency : always, if extents are used on OSTs
4126 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4127 Details : OSTs using the extents format would not zero the high 16 bits of
4128 the index physical block number. This is not a problem for any
4129 OST filesystems smaller than 16TB, and no kernels support ext3
4130 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4131 new/modified files) and can be fixed for existing filesystems
4132 with e2fsprogs-1.39-cfs1.
4137 Description: import connection selection may be incorrect if timer wraps
4138 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4139 import connections to be ignored if the 32-bit jiffies counter
4140 wraps. Use a 64-bit jiffies counter.
4143 Frequency : very large clusters immediately after boot
4145 Description: LNET request buffers exhausted under heavy short-term load
4146 Details : If a large number of client requests are generated on a service
4147 that has previously never seen so many requests it is possible
4148 that the request buffer growth cannot keep up with the spike in
4149 demand. Instead of dropping incoming requests, they are held in
4150 the LND until the RPC service can accept more requests.
4153 Frequency : Sometimes during replay
4155 Description: Assertion failure in ll_local_open after replay.
4156 Details : If replay happened on an open request reply before we were able
4157 to set replay handler, reply will become not swabbed tripping the
4158 assertion in ll_local_open. Now we set the handler right after
4159 recognising of open request
4162 Frequency : very rare
4164 Description: kernel reports "badness in vsnprintf"
4165 Details : Reading from the "recovery_status" /proc file in small chunks
4166 may cause a negative length in lprocfs_obd_rd_recovery_status()
4167 call to vsnprintf() (which is otherwise harmless). Exit early
4168 if there is no more space in the output buffer.
4170 Severity : enhancement
4172 Description: clear OBD RPC statistics by writing to them
4173 Details : It is now possible to clear the OBD RPC statistics by writing
4174 to the "stats" file.
4179 Description: Client mtime is not the same on different clients after utimes
4180 Details : In some cases, the client was using the utimes() syscall on
4181 a file cached on another node. The clients now validate the
4182 ctime from the MDS + OSTs to determine which one is right.
4187 Description: Inability to activate failout mode
4188 Details : lconf script incorrectly assumed that in pythong string's numeric
4189 value is used in comparisons.
4192 Frequency : always with multiple stripes per file
4194 Description: Inefficient object allocation for mutli-stripe files
4195 Details : When selecting which OSTs to stripe files over, for files with
4196 a stripe count that divides evenly into the number of OSTs,
4197 the MDS is always picking the same starting OST for each file.
4198 Return the OST selection heuristic to the original design.
4203 Description: mount failures may take full timeout to return an error
4204 Details : Under some heavy load conditions it is possible that a
4205 failed mount can wait for the full obd_timeout interval,
4206 possibly several minutes, before reporting an error.
4207 Instead return an error as soon as the status is known.
4209 Frequency : quota enabled and large files being deleted
4211 Description: releasing more than 4GB of quota at once hangs OST
4212 Details : If a user deletes more than 4GB of files on a single OST it
4213 will cause the OST to spin in an infinite loop. Release
4214 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4219 Description: statfs data retrieved from /proc may be stale or zero
4220 Details : When reading per-device statfs data from /proc, in the
4221 {kbytes,files}_{total,free,avail} files, it may appear
4222 as zero or be out of date.
4225 Frequency : systems with MD RAID1 external journal devices
4227 Description: lconf's call to blkid is confused by RAID1 journal devices
4228 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4229 of returning all block devices that match the journal UUID.
4232 Frequency : always, for aggregate stripe size over 4GB
4234 Description: assertion fails when trying to use 4GB stripe size
4235 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4236 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4239 Frequency : always on ppc64
4241 Description: the first write on an ext3 filesystem with mballoc got stuck
4242 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4243 perform endianness conversion.
4245 ------------------------------------------------------------------------------
4247 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4249 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4250 INTEROPERATE with older versions automatically. Please read the
4251 user documentation before upgrading any part of a live system.
4252 * WARNING: Lustre networking configuration changes are required with
4253 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4256 * Support for kernels:
4257 2.6.9-22.0.2.EL (RHEL 4)
4258 2.6.5-7.244 (SLES 9)
4259 2.6.12.6 vanilla (kernel.org)
4262 Severity : enhancement
4263 Bugzilla : 7981/8208
4264 Description: Introduced Lustre Networking (LNET)
4265 Details : LNET is new networking infrastructure for Lustre, it includes
4266 a reorganized network configuration mode (see the user
4267 documentation for full details) as well as support for routing
4268 between different network fabrics. Lustre Networking Devices
4269 (LNDs) for the supported network fabrics have also been
4270 created for this new infrastructure.
4272 Severity : enhancement
4273 Description: Introduced Access control lists
4274 Details : clients can set ACLs on files and directories in order to have
4275 more fine-grained permissions than the standard Unix UGO+RWX.
4276 The MDS must be started with the "-o acl" mount option.
4278 Severity : enhancement
4279 Description: Introduced filesystem quotas
4280 Details : Administrators may now establish per-user quotas on the
4283 Severity : enhancement
4285 Description: Configuration change for the XT3
4286 The PTLLND is now used to run Lustre over Portals on the XT3
4287 The configure option(s) --with-cray-portals are no longer used.
4288 Rather --with-portals=<path-to-portals-includes> is used to
4289 enable building on the XT3. In addition to enable XT3 specific
4290 features the option --enable-cray-xt3 must be used.
4295 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4296 Details : A race between allocating a new llcd and re-getting the llcd_lock
4297 allowed another thread to grab newly-allocated llcd.
4299 Severity : enhancement
4301 Description: 2.6 OST async journal commit and locking fix to improve performance
4302 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4303 2.6 kernels are now async as they already were in 2.4 kernels so
4304 that they can commit concurrently with the network bulk transfer.
4305 For block-allocated files the filter allocation semaphore is held
4306 to avoid filesystem fragmentation during allocation. BKL lock
4307 removed for 2.6 xattr operations where it is no longer needed.
4312 Description: lconf incorrectly determined whether two IP networks could talk
4313 Details : In some more complicated routing and multiple-network
4314 configurations, lconf will avoid trying to make a network
4315 connection to a disjoint part of the IP space. It was doing the
4316 math incorrectly for one set of cases.
4321 Description: Fix for potential infinite loop processing records in an llog.
4322 Details : If an llog record is corrupted/zeroed, it is possible to loop
4323 forever in llog_process(). Validate the llog record length
4324 and skip the remainder of the block on error.
4327 Frequency : occasional (liblustre only)
4329 Description: liblustre could not open files whose last component is a symlink
4330 Details : sysio_path_walk() would incorrectly pass the open intent to
4331 intermediate path components.
4334 Frequency : rare (liblustre only with non-standard tuning)
4335 Bugzilla : 7201 (7350)
4336 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4337 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4338 after it has been released, only do this for referenced locks
4340 Severity : enhancement
4342 Description: specifying an (invalid) directory default stripe_size of -1
4343 would reset the directory default striping
4344 Details : stripe_size -1 was used internally to signal directory stripe
4345 removal, now use "all default" to signal dir stripe removal
4346 as a directory striping of "all default" is not useful
4349 Frequency : common for large clusters running liblustre clients
4351 Description: doing an ls when liblustre clients are running is slow
4352 Details : sending a glimpse AST to a liblustre client waits for every AST
4353 to time out, as liblustre clients will not respond. Since they
4354 cannot cache data we refresh the OST lock LVB from disk instead.
4356 Severity : enhancement
4358 Description: doing an ls at the same time as file IO can be slow
4359 Details : enqueue and other "small" requests can be blocked behind many
4360 large IO requests. Create a new OST IO portal for non-IO
4361 requests so they can be processed faster.
4364 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4366 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4367 created file with mode 000
4368 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4369 allow an ftruncate() as a non-root user to a file with mode 000.
4370 Lustre now respects this flag to disable mode checks when
4371 truncating a file owned by the user
4374 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4377 Description: Revoking locks from clients that went dead or catatonic might take
4379 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4380 cancellation of such locks instant on servers without waiting for
4381 any reply from clients. Clients drops these locks when cancel
4382 notification from server is received without replying.
4385 Frequency : liblustre-only, when liblustre client dies or becomes busy
4387 Description: Doing ls on Linux clients can take a long time with active
4389 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4390 granting such locks to it in the first place if possible. Locks
4391 are taken by proxy on the OST during the read or write and
4392 dropped immediately afterward. Add connect flags handling, do
4393 not grant locks to liblustre clients for glimpse ASTs.
4395 Severity : enhancement
4397 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4398 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4399 sequential reads and behaves badly with applications doing random
4400 reads. Improve it by reading ahead at least read region, and
4401 avoiding excessive large RPC for small reads.
4403 Severity : enhancement
4405 Description: Creating more than 1000 files for a single job may cause a load
4406 imbalance on the OSTs if there are also a large number of OSTs.
4407 Details : qos_prep_create() uses an OST index reseed value that is an
4408 even multiple of the number of available OSTs so that if the
4409 reseed happens in the middle of the object allocation it will
4410 still utilize the OSTs as uniformly as possible.
4415 Description: OST or MDS may oops in ping_evictor_main()
4416 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4417 but doesn't restart at beginning of obd_exports_timed list
4420 Severity : enhancement
4422 Description: improve by-nid export eviction on the MDS and OST
4423 Details : allow multiple exports with the same NID to be evicted at one
4424 time without re-searching the exports list.
4427 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4429 Description: MDS may oops in groups_free()
4430 Details : in rare race conditions a newly allocated group_info struct is
4431 freed again, and this can be NULL. The 2.4 compatibility code
4432 for groups_free() checked for a NULL pointer, but 2.6 did not.
4435 Frequency : common for liblustre clients doing little filesystem IO
4436 Bugzilla : 9352, 7313
4437 Description: server may evict liblustre clients accessing contended locks
4438 Details : if a client is granted a lock or receives a completion AST
4439 with a blocking AST already set it would not reply to the AST
4440 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4443 Frequency : lfs setstripe, only systems with more than 160 OSTs
4445 Description: unable to set striping with a starting offset beyond OST 160
4446 Details : llapi_create_file() incorrectly limited the starting stripe
4447 index to the maximum single-file stripe count.
4450 Frequency : LDAP users only
4452 Description: lconf did not handle in-kernel recovery with LDAP properly
4453 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4455 Severity : enhancement
4457 Description: bind OST threads to NUMA nodes to improve performance
4458 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4459 node and do their allocations there to localize memory access
4461 Severity : enhancement
4463 Description: llmount can determine client NID directly from Myrinet (GM)
4464 Details : the client NID code from gmnalnid was moved directly into
4465 llmount, removing the need to use this or specifying the
4466 client NID explicitly when mounting GM clients with zeroconf
4469 Frequency : if client is started with down MDS
4471 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4472 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4473 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4474 ptlrpc_import_delayed_req() because we didn't check if the
4475 request was interrupted, and we also didn't break out of the
4476 event loop if there was no timeout
4481 Description: data loss during non-page-aligned writes to a single file from
4482 both multiple nodes and multiple threads on one node at same time
4483 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4484 inconsistency led to false short-reads, that were cached and later
4485 used by ->prepare_write() to fill in partially written page,
4486 leading to data loss.
4489 Frequency : always, if lconf --abort_recovery used
4491 Description: lconf --abort_recovery fails with 'Operation not supported'
4492 Details : lconf was attempting to abort recovery on the MDT device and not
4495 Severity : enhancement
4497 Description: remove cleanup logs
4498 Details : replace lconf-generated cleanup logs with lustre internal
4499 cleanup routines. Eliminates the need for client-cleanup and
4502 Severity : enhancement
4504 Description: add support for EAs (user and system) on lustre filesystems
4505 Details : it is now possible to store extended attributes in the Lustre
4506 client filesystem, and with the user_xattr mount option it
4507 is possible to allow users to store EAs on their files also
4509 Severity : enhancement
4511 Description: Add possibility (config option) to show minimal available OST free
4513 Details : When compiled with --enable-mindf configure option, statfs(2)
4514 (and so, df) will return least minimal free space available from
4515 all OSTs as amount of free space on FS, instead of summary of
4516 free spaces of all OSTs.
4518 Severity : enhancement
4520 Description: do not expand extent locks acquired on OST-side
4521 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4522 by server: they are not cached anyway.
4525 Frequency : when mmap is used/binaries executed from Lustre
4527 Description: Unmmap pages before throwing them away from read cache.
4528 Details : llap_shrink cache now attempts to unmap pages before discarding
4529 them (if unmapping failed - do not discard). SLES9 kernel has
4530 extra checks that trigger if this unmapping is not done first.
4535 Description: lconf didn't resolve symlinks before checking to see whether a
4536 given mountpoint was already in use
4539 Frequency : when migrating failover services
4540 Bugzilla : 6395, 9514
4541 Description: When migrating a subset of services from a node (e.g. failback
4542 from a failover service node) the remaining services would
4543 time out and evict clients.
4544 Details : lconf --force (implied by --failover) sets the global obd_timeout
4545 to 5 seconds in order to quickly disconnect, but this caused
4546 other RPCs to time out too quickly. Do not change the global
4547 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4549 Severity : enhancement
4550 Frequency : if MDS is started with down OST
4551 Bugzilla : 9439,5706
4552 Description: Allow startup/shutdown of an MDS without depending on the
4553 availability of the OSTs.
4554 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4555 Add appropriate locking and lov-osc refcounts for safe
4556 cleaning. Add osc abort_inflight calls in case the
4557 synchronize never started.
4560 Frequency : occasional (Cray XT3 only)
4562 Description: root not authorized to access files in CRAY_PORTALS environment
4563 Details : The client process capabilities were not honoured on the MDS in
4564 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4565 been accessed by an authorized user then root was able to access
4566 the file on the local client also. The root user capabilities
4567 are now allowed on the MDS, as this environment has secure UID.
4570 Frequency : occasional
4572 Description: ldiskfs "too long searching" message happens too often
4573 Details : A debugging message (otherwise harmless) prints too often on
4574 the OST console. This has been reduced to only happen when
4575 there are fragmentation problems on the filesystem.
4580 Description: Division by zero in statfs when all OSCs are inactive
4581 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4582 lov_statfs divides by value returned by lov_get_stripecnt().
4586 Bugzilla : 9489, 3273
4587 Description: First write from each client to each OST was only 4kB in size,
4588 to initialize client writeback cache, which caused sub-optimal
4589 RPCs and poor layout on disk for the first writen file.
4590 Details : Clients now request an initial cache grant at (re)connect time
4591 and so that they can start streaming writes to the cache right
4592 away and always do full-sized RPCs if there is enough data.
4593 If the OST is rebooted the client also re-establishes its grant
4594 so that client cached writes will be honoured under the grant.
4599 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4600 Details : Now I/O RPCs go to different portal number and (presumably) fast
4601 lock requests (and glimses) and other RPCs get their own service
4602 threads pool that should be able to service those RPCs
4605 Severity : enhancement
4607 Description: Ability to exchange lustre version between client and servers and
4608 issue warnings at client side if client is too old. Also for
4609 liblustre clients there is ability to refuse connection of too old
4611 Details : New 'version' field is added to connect data structure that is
4612 filled with version info. That info is later checked by server and
4616 Frequency : rare, liblustre only.
4617 Bugzilla : 9296, 9581
4618 Description: Two simultaneous writes from liblustre at offset within same page
4619 might proceed at the same time overwriting eachother with stale
4621 Details : I/O lock withing llu_file_prwv was released too early, before data
4622 actually was hitting the wire. Extended lock-holding time until
4623 server acknowledges receiving data.
4626 Frequency : extremely rare. Never observed in practice.
4628 Description: avoid generating lustre_handle cookie of 0.
4629 Details : class_handle_hash() generates handle cookies by incrementing
4630 global counter, and can hit 0 occasionaly (this is unlikely, but
4631 not impossible, because initial value of cookie counter is
4632 selected randonly). Value of 0 is used as a sentinel meaning
4633 "unassigned handle" --- avoid it. Also coalesce two critical
4634 sections in this function into one.
4636 Severity : enhancement
4638 Description: allow liblustre clients to delegate truncate locking to OST
4639 Details : To avoid overhead of locking, liblustre client instructs OST to
4640 take extent lock in ost_punch() on client's behalf. New connection
4641 flag is added to handle backward compatibility.
4643 Severity : enhancement
4644 Bugzilla : 4928, 7341, 9758
4645 Description: allow number of OST service threads to be specified
4646 Details : a module parameter allows the number of OST service threads
4647 to be specified via "options ost ost_num_threads={N}" in the
4648 OSS's /etc/modules.conf or /etc/modprobe.conf.
4652 Bugzilla : 6146, 9635, 9895
4653 Description: servers crash with bad pointer in target_handle_connect()
4654 Details : In rare cases when a client is reconnecting it was possible that
4655 the connection request was the last reference for that export.
4656 We would temporarily drop the export reference and get a new
4657 one, but this may have been the last reference and the export
4658 was just destroyed. Get new reference before dropping old one.
4660 Severity : enhancement
4661 Frequency : if client is started with failover MDS
4663 Description: Allow multiple MDS hostnames in the mount command
4664 Details : Try to read the configuration from all specified MDS
4665 hostnames during a client mount in case the "primary"
4668 Severity : enhancement
4670 Description: Stop sending data to evicted clients as soon as possible.
4671 Details : Check if the client we are about to send or are sending data to
4672 was evicted already. (Check is done every second of waiting,
4673 for which l_wait_event interface was extended to allow checking
4674 of exit condition at specified intervals).
4677 Frequency : rare, normally only when NFS exporting is done from client
4679 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4681 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4684 Frequency : always, if srand() is called before liblustre initialization
4686 Description: Liblustre uses system PRNG disturbing its usage by user application
4687 Details : Introduce internal to lustre fast and high-quality PRNG for
4688 lustre usage and make liblustre and some other places in generic
4689 lustre code to use it.
4691 Severity : enhancement
4692 Bugzilla : 9477, 9557, 9870
4693 Description: Verify that the MDS configuration logs are updated when xml is
4694 Details : Check if the .xml configuration logs are newer than the config
4695 logs stored on the MDS and report an error if this is the case.
4696 Request --write-conf, or allow starting with --old_conf.
4698 Severity : enhancement
4700 Description: Handle symlinks in the path when checking if Lustre is mounted.
4701 Details : Resolve intermediate symlinks when checking if a client has
4702 mounted a filesystem to avoid duplicate client mounts.
4707 Description: lconf can hit an error exception but still return success.
4708 Details : The lconf command catches the Command error exception at the top
4709 level script context and will exit with the associated exit
4710 status, but doesn't ensure that this exit status is non-zero.
4715 Description: failure of ptlrpc thread startup can cause oops
4716 Details : Starting a ptlrpc service thread can fail if there are a large
4717 number of threads or the server memory is very fragmented.
4718 Handle this without oopsing.
4721 Frequency : always, only if liblustre and non-default acceptor port was used
4723 Description: liblustre cannot connect to servers with non-default acceptor port
4724 Details : tcpnal_set_default_params() was not called and was therefore
4725 ignoring the environment varaible TCPNAL_PORT, as well as other
4726 TCPNAL_ environment variables
4731 Description: two objects could be created on the same OST for a single file
4732 Details : If an OST is down, in some cases it was possible to create two
4733 objects on a single OST for a single file. No problems other
4734 than potential performance impact and spurious error messages.
4738 Bugzilla : 5681, 9562
4739 Description: Client may oops in ll_unhash_aliases
4740 Details : Client dcache may become inconsistent in race condition.
4741 In some cases "getcwd" can fail if the current directory is
4747 Description: Inode refcounting problems in NFS export code
4748 Details : link_raw functions used to call d_instantiate without obtaining
4749 extra inode reference first.
4753 Bugzilla : 9942, 9903
4754 Description: Referencing freed requests leading to crash, memleaks with NFS.
4755 Details : We used to require that call to ll_revalidate_it was always
4756 followed by ll_lookup_it. Also with revalidate_special() it is
4757 possible to call ll_revalidate_it() twice for the same dentry
4758 even if first occurence returned success. This fix changes semantic
4759 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4760 reference on a request referred from the intent.
4761 ll_intent_release() then releases such a request.
4764 Frequency : rare, normally benchmark loads only
4766 Description: unlinked inodes were kept in memory on the client
4767 Details : If a client is repeatedly creating and unlinking files it
4768 can accumulate a lot of stale inodes in the inode slab cache.
4769 If there is no other client load running this can cause the
4770 client node to run out of memory. Instead flush old inodes
4771 from client cache that have the same inode number as a new inode.
4774 Frequency : SLES9 2.6.5 kernel and long filenames only
4775 Bugzilla : 9969, 10379
4776 Description: utime reports stale NFS file handle
4777 Details : SLES9 uses out-of-dentry names in some cases, which confused
4778 the lustre dentry revalidation. Change it to always use the
4782 Frequency : rare, unless heavy write-truncate concurrency is continuous
4783 Bugzilla : 4180, 6984, 7171, 9963, 9331
4784 Description: OST becomes very slow and/or deadlocked during object unlink
4785 Details : filter_destroy() was holding onto the parent directory lock
4786 while truncating+unlinking objects. For very large objects this
4787 may block other threads for a long time and slow overall OST
4788 responsiveness. It may also be possible to get a lock ordering
4789 deadlock in this case, or run out of journal credits because of
4790 the combined truncate+unlink. Solution is to do object truncate
4791 first in one transaction without parent lock, and then do the
4792 final unlink in a new transaction with the parent lock. This
4793 reduces the lock hold time dramatically.
4796 Frequency : rare, 2.4 kernels only
4798 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4799 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4800 before unmounting their filesystem in order to avoid deadlock.
4801 The kernel_locked() function in 2.4 kernels only checks whether
4802 the kernel lock is held, not whether it is this process that is
4803 holding it as 2.6 kernels do.
4808 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4809 Details : The client ptlrpc code may be trying to reconnect to a down
4810 server before a previous connection attempt has timed out.
4811 Increase the reconnect interval to be longer than the connection
4812 timeout interval to avoid sending duplicate connections to
4816 Frequency : echo_client brw_test command
4818 Description: fix echo_client to work with OST preallocated code
4819 Details : OST preallocation code (5137) didn't take echo_client IO path
4820 into account: echo_client calls filter methods outside of any
4821 OST thread and, hence, there is no per-thread preallocated
4822 pages and buffers to use. Solution: hijack pga pages for IO. As
4823 a byproduct, this avoids unnecessary data copying.
4827 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4828 Description: Client can oops in mdc_commit_close() after open replay
4829 Details : It was possible for the MDS to return an open request with no
4830 transaction number in mds_finish_transno() if the client was
4831 evicted, but without actually returning an error. Clients
4832 would later try to replay that open and may trip an assertion
4833 Simplify the client close codepath, and always return an error
4834 from the MDS in case the open is not successful.
4837 Frequency : rare, 2.6 OSTs only
4839 Description: OST may deadlock under high load on fragmented files
4840 Details : If there was a heavy load and highly-fragmented OST filesystems
4841 it was possible to have all the OST threads deadlock waiting on
4842 allocation of biovecs, because the biovecs were not released
4843 until the entire RPC IO was completed. Instead, release biovecs
4844 as soon as they are complete to ensure forward IO progress.
4846 Severity : enhancement
4848 Description: Support for specifying external journal device at mount
4849 Details : If an OST or MDS device is formatted with an external journal
4850 device, this device major/minor is stored in the ext3 superblock
4851 and may not be valid for failover. Allow detecting and
4852 specifying the external journal at mount time.
4857 Description: Mounting an MDS with pending unlinked files may cause oops
4858 Details : target_finish_recovery() calls mds_postrecov() which returned
4859 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4860 considers this an error and immediately begins cleaning up the
4861 lov, just after starting the mds_lov process
4863 Severity : enhancement
4865 Description: Implement 'lfs df' to report actual free space on per-OST basis
4866 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4867 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4868 usage of objects; '-h' to report in human readable format.
4870 ------------------------------------------------------------------------------
4872 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4879 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4880 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4881 at startup during orphan unlinking. Instead, skip buddy update
4882 before setup, as it will be regenerated after recovery is complete.
4887 Description: If an OST is inactive, its locks might reference stale inodes.
4888 Details : lov_change_cbdata() must iterate over all namespaces, even if
4889 they are inactive to clear inode references from the lock.
4891 Severity : enhancement
4892 Frequency : occasional, if non-standard max_dirty_mb used
4894 Description: Client will block write RPCs if not enough grant
4895 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4896 then the client will block writes while waiting for another RPC
4897 to complete instead of consuming its dirty limit. With change
4898 we get improved performance when max_dirty_mb is small.
4900 Severity : enhancement
4901 Bugzilla : 3389, 6253
4902 Description: Add support for supplementary groups on the MDS.
4903 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4904 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4905 lookups for user supplementary groups into a cache.
4909 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4910 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4911 for liblustre O_EXCL is enforced by MDS.
4915 Description: readdir never returns NULL in liblustre.
4916 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4917 dentry in struct dirent.
4921 Frequency : liblustre only, depends on application IO pattern
4922 Description: liblustre clients evicted if not contacting servers
4923 Details : Don't put liblustre clients into the ping_evictor list, so
4924 they will not be evicted by the pinger ever.
4926 Severity : enhancement
4928 Description: Add ability to evict clients by NID from MDS.
4929 Details : By echoing "nid:$NID" string into
4930 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4931 $NID would be instantly evicted from this MDS and from all active
4932 OSTs connected to it.
4936 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4937 Details : lookup_it_finish() used to query file size from OSTs that was not
4942 Description: service threads change working directory to that of init
4943 Details : Starting lustre service threads may pin the working directory
4944 of the parent thread, making that filesystem busy. Threads
4945 now change to the working directory of init to avoid this.
4949 Frequency : during shutdown only
4950 Description: shutdown with a failed MDS or OST can cause unmount to hang
4951 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4956 Frequency : 2.6 only
4957 Description: chmod/chown may include an extra supplementary group
4958 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4959 supplementary group and if none is specified this is used.
4962 Bugzilla : 5479 (6816)
4964 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4965 Details : If lookup is in progress on a file that is unlinked we might try
4966 to revalidate the inode and fail in revalidate after lookup is
4967 complete and ll_file_open() enqueues the open again but
4968 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4971 Frequency : always, if lconf --abort_recovery used
4973 Description: lconf --abort_recovery fails with 'Operation not supported'
4974 Details : lconf was attempting to abort recovery on the MDT device and not
4977 ------------------------------------------------------------------------------
4979 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4984 Frequency : rare (only unsupported configurations with a node running as an
4986 Bugzilla : 6514, 5137
4987 Description: Mounting a Lustre file system on a node running as an OST could
4989 Details : OSTs now preallocates memory needed to write out data at
4990 startup, instead of when needed, to avoid having to
4991 allocate memory in possibly low memory situations.
4992 Specifically, if the file system is mounted on on OST,
4993 memory pressure could force it to try to write out data,
4994 which it needed to allocate memory to do. Due to the low
4995 memory, it would be unable to do so and the node would
4996 become unresponsive.
4998 Severity : enhancement
5000 Description: Addition of lconf --service command line option
5001 Details : lconf now accepts a '--service <arg>' option, which is
5002 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5004 Severity : enhancement
5006 Description: Failover mode is now the default for OSTs.
5007 Details : By default, OSTs will now run in failover mode. To return to
5008 the old behaviour, add '--failout' to the lmc line for OSTs.
5010 Severity : enhancement
5012 Description: Health checks are now provided for MDS and OSTs
5013 Details : Additional detailed health check information on MSD and OSTs
5014 is now provided through the procfs health_check value.
5017 Frequency : occasional, depends on IO load
5019 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5020 after numerous create/delete cycles
5021 Details : The ext3 inode allocation policy would not allocate new inodes
5022 very well on the OSTs because there are no new directories
5023 being created. Instead we look for groups with free space if
5024 the parent directories are nearly full.
5029 Description: Network or server problems during mount may cause partially
5030 mounted clients instead of returning an error.
5031 Details : The config llog parsing code may overwrite the error return
5032 code during mount error handling, returning success instead
5038 Description: MDS can fail to allocate large reply buffers
5039 Details : After long uptimes the MDS can fail to allocate large reply
5040 buffers (e.g. zconf client mount config records) due to memory
5041 fragmentation or consumption by the buffer cache. Preallocate
5042 some large reply buffers so that these replies can be sent even
5043 under memory pressure.
5047 Frequency : rare (liblustre)
5048 Description: fsx running with liblustre complained that using truncate() to
5049 extend the file doesn't work. This patch corrects that issue.
5050 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5051 ATTR_SIZE and lsm use in llu_setattr_raw.
5055 Frequency : rare, only 2.6 kernels
5056 Description: Unusual file access patterns on the MDS may result in inode
5057 data being lost in very rare circumstances.
5058 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5059 code in the 2.6 kernel caused the inode and/or EA data not to
5060 be read from disk, causing single-file corruption.
5064 Frequency : rare, only 2.6 filesystems using extents
5065 Description: Heavy concurrent write and delete load may cause data corruption.
5066 Details : It was possible under high-load situations to have an extent
5067 metadata block in the block device cache from a just-unlinked
5068 file overwrite a newly-allocated data block. We now unmap any
5069 metadata buffers that alias just-allocated data blocks.
5073 Frequency : filesystems with default stripe_count larger than 77
5074 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5075 Details : lconf specifies an inode size of 4096 bytes when the default
5076 stripe_count is larger than 77. This conflicts with the default
5077 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5078 this case to avoid pinning too much memory for large EAs.
5080 ------------------------------------------------------------------------------
5082 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5087 Frequency : rare (extremely heavy IO load with hundreds of clients)
5089 Description: Client is evicted, gets IO error writing to file
5090 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5091 caused clients to be evicted for AST timeouts. The fixes in
5092 bug 5192 mean we no longer need to have such short AST timeouts
5093 so ldlm_timeout has been increased.
5096 Frequency : occasional during --force or --failover shutdown under load
5097 Bugzilla : 5949, 4834
5098 Description: Server oops/LBUG if stopped with --force or --failover under load
5099 Details : a collection of import/export refcount and cleanup ordering
5100 issues fixed for safer force cleanup
5103 Frequency : only filesystems larger than 120 OSTs
5104 Bugzilla : 5990, 6223
5105 Description: lfs getstripe would oops on a very large filesystem
5106 Details : lov_getconfig used kfree on vmalloc'd memory
5109 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5110 Bugzilla : 6242, 6243
5111 Description: reading from files that had been truncated to a non-zero size
5112 but never opened returned no data
5113 Details : ll_file_read() reads zeros from no-object files to EOF
5118 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5119 Details : There's little harm in aborting MDS/OSS recovery and letting it
5120 try again, so I removed the LASSERT and return an error instead.
5122 Severity : enhancement
5124 Description: New debugging infrastructure for tracking down data corruption
5125 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5126 (b) cover more of the client-side code path, and (c) try to narrow
5127 down where problems occurred
5131 Bugzilla : 3819, 4364, 4397, 6313
5132 Description: Racing close and eviction MDS could cause assertion in mds_close
5133 Details : It was possible to get multiple mfd references during close and
5134 client eviction, leading to one thread referencing a freed mfd.
5136 Severity: : enhancement
5137 Bugzilla : 3262, 6359
5138 Description: Attempts to reconnect to servers are now more aggressive.
5139 Details : This builds on the enhanced upcall-less recovery that was added
5140 in 1.4.2. When trying to reconnect to servers, clients will
5141 now try each server in the failover group every 10 seconds. By
5142 default, clients would previously try one server every 25 seconds.
5147 Description: After recovery, certain operations trigger a failed
5148 assertion on a client.
5149 Details : Failing over an mds, using lconf -d --failover, while a
5150 client was doing a readdir() call would cause the client to
5151 LBUG after recovery completed and the readdir() was resent.
5153 Severity : enhancement
5155 Description: Default groups are now added by lconf
5156 Details : You can now run lconf --group <servicename> without having to
5157 manually add groups with lmc.
5160 Frequency : occasional
5162 Description: Nodes with an elan id of 0 trigger a failed assertion
5165 Frequency : always when accessing e.g. tty/console device nodes
5167 Description: tty and some other devices nodes cannot be used on lustre
5168 Details : file's private_data field is used by device data and lustre
5169 values in there got lost. New field was added to struct file to
5170 store fs-specific private data.
5173 Frequency : when exporting Lustre via NFS
5175 Description: NFSD failed occasionally when looking up a path component
5176 Details : NFSD is looking up ".." which was broken in ext3 directories
5177 that had grown large enough to become hashed.
5180 Frequency : Clusters with multiple interfaces not on the same subnet
5182 Description: Nodes will repeatedly try to reconnect to an interface which it
5183 cannot reach and report an error to the log.
5184 Details : Extra peer list entries will be created by lconf with some peers
5185 unreachable. lconf now validates the peer before adding it.
5188 Frequency : Only if a default stripe is set on the filesystem root.
5190 Description: Setting a default stripe on the filesystem root prevented the
5191 filesystem from being remounted.
5192 Details : The client was sending extra request flags in the root getattr
5193 request and did not allocate a reply buffer for the dir EA.
5196 Frequency : occasional, higher if lots of files are accessed by one client
5197 Bugzilla : 6159, 6097
5198 Description: Client trips assertion regarding lsm mismatch/magic
5199 Details : While revalidating inodes the VFS looks up inodes with ifind()
5200 and in rare cases can find an inode that is being freed.
5201 The ll_test_inode() code will free the lsm during ifind()
5202 when it finds an existing inode and then the VFS later attaches
5203 this free lsm to a new inode.
5207 Bugzilla : 6422, 7030
5208 Description: MDS deadlock between mkdir and client eviction
5209 Details : Creating a new file via mkdir or mknod (starting a transaction
5210 and getting the ns lock) can deadlock with client eviction
5211 (gets ns lock and trying to finish a synchronous transaction).
5214 Frequency : occasional
5215 Description: While starting a server, the fsfilt_ext3 module could not be
5217 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5218 kernels. Previously, lconf would still use the ext3 name
5219 when trying to load modules. Now, it will correctly use
5220 ext3 on 2.4 and ldiskfs on 2.6.
5222 Severity : enhancement
5223 Description: The default stripe count has been changed to 1
5224 Details : The interpretation of the default stripe count (0, to lfs
5225 or lmc) has been changed to mean striping across a single
5226 OST, rather than all available. For general usage we have
5227 found a stripe count of 1 or 2 works best.
5229 Severity : enhancement
5230 Description: Add support for compiling against Cray portals.
5231 Details : Conditional compiling for some areas that are different
5235 Frequency : occasional
5236 Bugzilla : 6409, 6834
5237 Description: Creating files with an explicit stripe count may lead to
5238 a failed assertion on the MDS
5239 Details : If some OSTs are full or unavailable, creating files may
5240 trigger a failed assertion on the MDS. Now, Lustre will
5241 try to use other servers or return an error to the
5245 Frequency : occasional
5247 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5248 caused lock timeout during readahead (since 1.4.2).
5249 Details : Processes doing readahead might match a lock that hasn't been
5250 granted yet if there are overlapping and conflicting lock
5251 requests. The readahead process waits on ungranted lock
5252 (original lock is CBPENDING), while OST waits for that process
5253 to cancel CBPENDING read lock and eventually evicts client.
5255 Severity : enhancement
5257 Description: Initial enabling of flock support for clients
5258 Details : Implements fcntl advisory locking and file status functions.
5259 This feature is provided as an optional mount flag (default
5260 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5261 locking are implemented yet, and those that are are not guaranteed
5262 to be completely correct in production environments.
5263 mount -t lustre -o [flock|noflock] ...
5266 Frequency : occasional
5268 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5269 assertion in the ext3 JBD (journaling) layer.
5270 Details : The b_committed_data struct is protected by the big kernel lock
5271 in 2.4 kernels, serializing journal_commit_transaction() and
5272 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5273 access to this struct. In 2.6 kernels there is finer grained
5274 locking to improve SMP performance of the JBD layer.
5278 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5283 Description: merge and cleanup kernel patches.
5286 -----------------------------------------------------------------------------
5288 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5290 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5291 versions of Lustre. Please update all servers and clients to
5292 version 1.4.2 or later at the same time. You must also run
5293 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5294 to update the configuration logs.
5296 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5297 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5298 - swab llog records of type '0' so we get proper header size/idx (5861)
5299 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5300 - fix rename of one directory over another leaking an inode (5953)
5301 - avoid SetPageDirty on 2.6 (5981)
5302 - don't re-add just-being-destroyed locks to the waiting list (5653)
5303 - when creating new directories, inherit the parent's custom
5304 striping settings if present parent (3048)
5305 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5306 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5307 - don't hold client locks on temporary worklist from l_lru (5666)
5308 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5309 - automatically evict dead clients (5921)
5310 - Update file size properly in create+truncate+fstat case (6196)
5311 - Do not unhash mountpoint dentries, do not allow removal of
5313 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5314 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5315 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5316 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5317 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5318 - Correctly handle reads of files with no objects (6243)
5319 - lctl recover will also mark a device active if deactivate used (5933)
5321 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5322 - allow --write-conf on an MDS with different nettype than client (5619)
5323 - don't write config llogs to MDS for mounts not from that MDS (5617)
5324 - lconf should create multiple TCP connections from a client (5201)
5325 - init scripts are now turned off by default; run chkconfig --on
5326 lustre and chkconfig --on lustrefs to use them
5327 - upcalls are no longer needed for clients to recover to failover
5329 - add --abort-recovery option to lconf to abort recovery on device
5331 - add support for an arbitrary number of OSTs (3026)
5332 - Quota support protocol changes.
5333 - forward compatibility changes to wire structs (6007)
5334 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5335 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5336 - improved "lustre status" script
5337 - initialize blocksize for non-regular files (6062)
5338 - added --disable-server and --disable-client configure options (5782)
5339 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5340 - Vanilla 2.4.29 support
5341 - increase maximum number of obd devices to 520 (6242)
5342 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5343 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5344 SLES 9 2.6.5 kernels
5345 - limit stripes per file to 160 (the maximum EA size) (6093)
5347 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5350 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5351 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5352 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5353 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5354 - small ext3 extents cleanups and fixes (5733)
5355 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5356 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5357 - avoid needless client->OST connect, fix handle mismatch (5317)
5358 - fix DLM error path that led to out-of-sync client, long delays (5779)
5359 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5360 - fix several locking issues related to i_size (5492,5624,5654,5672)
5361 - don't move pending lock onto export if it is already evicted (5683)
5362 - fix kernel oops when creating .foo in unlinked directory (5548)
5363 - fix deadlock in obdfilter statistics vs. object create (5811)
5364 - use time_{before,after} to avoid timer jiffies wrap (5882)
5365 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5366 - Do not leak request if server was not able to process it (5154)
5367 - If mds_open unable to find parent dir, make that negative lookup(5154)
5368 - don't create new directories with extent-mapping (5909, 5936)
5370 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5371 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5372 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5373 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5374 - support for RHEL4 kernel on i686 (5773)
5375 - provide error messages when incompatible logs are encountered (5898)
5377 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5378 * version 1.4.0.10 (1.4.1 release candidate 1)
5380 - don't keep a lock reference when lock is not granted (4238)
5381 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5382 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5383 - drop import inflight refcount on signal_completed_replay error (5255)
5384 - unlock page after async write error during send (3677)
5385 - handle missing objects in filter_preprw_read properly (5265)
5386 - no transno return for symlink open, don't save no-trasno open (3440)
5387 - don't try to complete elan receive that already failed (4012)
5388 - free RPC server reply state on error (5406)
5389 - clean up thread from ptlrpc_start_thread() on error (5160)
5390 - readahead could read extra page into cache that wasn't ejected (5388)
5391 - prevent races in class_attach/setup/cleanup/detach (5260)
5392 - don't dereference de->d_inode after l_dput of de (5458)
5393 - use "int" for stripe value returned from lock_to_stripe (5544)
5394 - mballoc allocation and error-checking fixes in 2.6 (5504)
5395 - block device patches to fix I/O request sizes in 2.6 (5482)
5396 - look up hostnames for IB nals (5602)
5397 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5398 - don't start multiple acceptors for the same port (5277)
5399 - fix incorrect LASSERT in mds_getattr_name (5635)
5400 - export a proc file for general "ping" checking (5628)
5401 - fix "lfs check" to not block when the MDS is down (5628)
5403 - service request history (4965)
5404 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5405 - create an "evict_client" /proc entry on OSTs, like the MDS has
5406 - fix mount usage message, return errors per mount(8) (5168)
5407 - change grep [] to grep "[]" in tests so they work in more UMLs
5408 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5409 - remove ancient LOV_MAGIC_V0 EA support (5047)
5410 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5411 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5412 - allow readahead allocations to fail when low on memory (5383)
5413 - mmap locking landed again, after considerable improvement (2828)
5414 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5416 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5419 - send OST transaction number in read/write reply to free req (4966)
5420 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5421 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5422 - the watchdog thread now runs as interruptible (5246)
5423 - flock/lockf fixes (but it's still disabled, pending 5135)
5424 - don't use EXT3 constants in llite code (5094)
5425 - memory shortage at startup could cause assertion (5176)
5427 - reorganization of lov code
5428 - single portals codebase
5430 - add extents/mballoc support (5025)
5431 - direct I/O reads in the obdfilter (4048)
5432 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5434 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5437 - send OST transaction number in read/write reply to free req (4966)
5438 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5439 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5440 - the watchdog thread now runs as interruptible (5246)
5441 - handle missing objects in filter_preprw_read properly (5265)
5442 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5443 - drop import inflight refcount on signal_completed_replay error (5255)
5444 - unlock page after async write error during send (3677)
5445 - return original error code on reconstructed replies (3761)
5446 - no transno return for symlink open, don't save no-trasno open (3440)
5448 - add pid to ldlm debugging output (4922)
5449 - bump the watchdog timeouts -- we can't handle 30sec yet
5450 - extra debugging for orphan dentry/inode bug (5259)
5452 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5455 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5456 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5457 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5458 - fix NULL dereference in /proc/sys/portals/routes (4827)
5459 - allow failed mdc_close() operations to be interrupted (4561)
5460 - stop precreate on OST before MDS would time out on it (4778)
5461 - don't send partial-page writes before EOF from client (4410)
5462 - discard client grant for sub-page writes on large-page clients (4520)
5463 - don't free dentries not owned by NFS code, check generation (4806)
5464 - fix lsm leak if mds_create_objects() fails (4801)
5465 - limit debug_daemon file size, always print CERROR messages (4789)
5466 - use transno after validating reply (3892)
5467 - process timed out requests if import state changes (3754)
5468 - update mtime on OST during writes, return in glimpse (4829)
5469 - add mkfsoptions to LDAP (4679)
5470 - use ->max_readahead method instead of zapping global ra (5039)
5471 - don't interrupt __l_wait_event() during strace
5473 - add software watchdogs to catch hung threads quickly (4941)
5474 - make lustrefs init script start after nfs is mounted
5475 - fix CWARN/ERROR duplication (4930)
5476 - return async write errors to application if possible (2248)
5477 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5478 - print NAL number in %x format (4645)
5479 - update barely-supported suse-2.4.21-171 series (4842)
5480 - support for sles 9 %post scripts
5481 - support for building 2.6 kernel-source packages
5482 - support for sles km_* packages
5484 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5487 - ignore -ENOENT errors in osc_destroy (3639)
5488 - notify osc create thread that OSC is being cleaned up (4600)
5489 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5490 - reconstruct ost_handle() like mds_handle() (4657)
5491 - create a new thread to do import eviction to avoid deadlock (3969)
5492 - let lconf resolve symlinked-to devices (4629)
5493 - don't unlink "objects" from directory with default EA (4554)
5494 - hold socknal file ref over connect in case target is down (4394)
5495 - allow more than 32000 subdirectories in a single directory (3244)
5496 - fix blocks count for O_DIRECT writes (3751)
5497 - OST returns ENOSPC from object create when no space left (4539)
5498 - don't send truncate RPC if file size isn't changing (4410)
5499 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5500 - bind to privileged port in socknal and tcpnal (3689)
5502 - rate limit CERROR/CWARN console message to avoid overload (4519)
5503 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5504 - basic mmap support (3918)
5505 - kernel patch series update from b1_4 (4711)
5507 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5510 - avoid crash during MDS cleanup with OST shut down (2775)
5511 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5512 - don't use bad inodes on the MDS (3744)
5513 - dynamic object preallocation to improve recovery speed (4236)
5514 - don't hold spinlock over lock dumping or change debug flags (4401)
5515 - don't zero obd_dev when it is force cleaned (3651)
5516 - print grants to console if they go negative (4431)
5517 - "lctl deactivate" will stop automatic recovery attempts (3406)
5518 - look for existing locks in ldlm_handle_enqueue() (3764)
5519 - don't resolve lock handle twice in recovery avoiding race (4401)
5520 - revalidate should check working dir is a directory (4134)
5522 - don't always mark "slow" obdfilter messages as errors (4418)
5524 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5527 - don't close LustreDB during write_conf until it is done (3860)
5528 - fix typo in lconf for_each_profile (3821)
5529 - allow dumping logs from multiple threads at one time (3820)
5530 - don't allow multiple threads in OSC recovery (3812)
5531 - fix debug_size parameters (3864)
5532 - fix mds_postrecov to initialize import for llog ctxt (3121)
5533 - replace config semaphore with spinlock (3306)
5534 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5535 - don't allow enqueue to complete on a destroyed export (3822)
5536 - down write_lock before checking llog header bitmap (3825)
5537 - recover from lock replay timeout (3764)
5538 - up llog sem before sending rpc (3652)
5539 - reduce ns lock hold times when setting kms (3267)
5540 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5541 - fix NULL deref and obd_dev leak on setup error (3312)
5542 - replace some LBUG about llog ops with error handling (3841)
5543 - don't match INVALID dentries from d_lookup and spin (3784)
5544 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5545 - fix invalid assertion in ptlrpc_set_wait (3880)
5547 - add libwrap support for the TCP acceptor (3996)
5548 - add /proc/sys/portals/routes for non-root route listing (3994)
5549 - allow setting MDS UUID in .xml (2580)
5550 - print the stack of a process that LBUGs (4228)
5552 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5555 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5556 - make sure to unset replay flag from failed open requests (3440)
5557 - if default stripe count is 0, use OST count for inode size (3636)
5558 - update parent mtime/ctime on client for create/unlink (2611)
5559 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5560 - free recovery state on server during a forced cleanup (3571)
5561 - unregister_reply for resent reqs (3063)
5562 - loop back devices mounting and status check on 2.6 (3563)
5563 - fix resource-creation race that can provoke i_size == 0 (3513)
5564 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5565 - more debugging for page-accounting assertion (3746)
5566 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5567 - avoid deadlock after precreation failure (3758)
5568 - fix race and lock order deadlock in orphan handling (3450, 3750)
5569 - add validity checks when grabbing inodes from l_ast_data (3599)
5571 - add /proc/.../recovery_status to obdfilter (3428)
5572 - lightweight CDEBUG infrastructure, debug daemon (3668)
5573 - change default OSC RPC parameters to be better on small clusters
5574 - turn off OST read cache for files smaller than 32MB
5575 - install man pages and include them in rpms (3100)
5576 - add new init script for (un)mounting lustre filesystems (2593)
5577 - run chkconfig in %post for init scripts (3701)
5578 - drop scimac NAL (unmaintained)
5580 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5583 - clean kiobufs before and after use (3485)
5584 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5585 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5586 - change init script to fail more clearly if not run as root (1528)
5587 - allow clients to reconnect during replay (1742)
5588 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5589 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5590 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5591 - fix cond_resched (3554)
5592 - only evict unfinished clients after recovery (3515)
5593 - allow bulk resend, prevent data loss (3570)
5594 - dynamic ptlrpc request buffer allocation (2102)
5595 - don't allow unlinking open directory if it isn't empty (2904)
5596 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5597 - remove extraneous obd dereference causing LASSERT failure (3334)
5598 - don't use get_cycles() when creating temp. files on the mds (3156)
5599 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5600 - handle EEXIST for set-stripe, set proper directory name (3336)
5602 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5603 - fix ksocknal_fmb_callback() error messages (2918)
5605 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5608 - don't copy lvb into (possibly NULL) reply on error (2983)
5609 - don't deref dentry after dput, don't free lvb on error (2922)
5610 - use the kms to determine writeback rpc length (2947)
5611 - increment oti_logcookies when osc is inactive (2948)
5612 - update client's i_blocks count via lvb messages (2543)
5613 - handle intent open/close of special files properly (1557)
5614 - mount MDS with errors=remount-ro, like obdfilter (2009)
5615 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5616 - don't use cancelling-locks' kms values (2947)
5617 - use highest lock extent for kms, not last one (2925)
5618 - don't dereference ERR_PTR() dentry in error handling path (3107)
5619 - fix thread race in portals_debug_dumplog() (3122)
5620 - create lprocfs device entries at setup instead of at attach (1519)
5621 - common AST error handler, don't evict client on completion race (3145)
5622 - zero nameidata in detach_mnt in 2.6 (3118)
5623 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5624 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5625 - initialize RPC timeout timer earlier for 2.6 (3219)
5626 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5627 - print nal/nid for unknown nid (3258)
5628 - additional checks for oscc recovery before doing precreate (3284)
5629 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5630 - don't crash in mdc_close for bad permissions on open (3285)
5631 - zero i_rdev for non-device files (3147)
5632 - clear page->private before handing to FS, better assertion (3119)
5633 - tune the read pipeline (3236)
5634 - fix incorrect decref of invalidated dentry (2350)
5635 - provide read-ahead stats and refine rpc in flight stats (3328)
5636 - don't hold journal transaction open across create RPC (3313)
5637 - update atime on MDS at close time (3265)
5638 - close LDAP connection when recovering to avoid server load (3315)
5639 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5640 - don't leak open file on MDS after open resend (3325)
5641 - serialize filter_precreate and filter_destroy_precreated (3329)
5642 - loop device shouldn't call sync_dev() for nul device (3092)
5643 - clear page cache after eviction (2766)
5644 - resynchronize MDS->OST in background (2824)
5645 - refuse to mount the same filesystem twice on same mountpoint (3394)
5646 - allow llmount to create routes for mounting behind routers (3320)
5647 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5648 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5649 - fix rare problem with rename on htree directories (3417)
5651 - allow default OST striping configuration per directory (1414)
5652 - fix compilation for qswnal for 2.6 kernels (3125)
5653 - increase maximum number of MDS request buffers for large systems
5654 - change liblustreapi to be useful for external progs like lfsck (3098)
5655 - increase local configuration timeout for slow disks (3353)
5656 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5658 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5661 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5662 - don't overwrite extent policy data in reply if lock was blocked (2901)
5663 - drop filter export grants atomically with removal from device (2663)
5664 - del obd_self_export from work_list in class_disconnect_exports (2908)
5665 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5666 - swab reply message in mdc_close, other PPC fixes (2464)
5667 - fix destroying of named logs (2325)
5668 - overwrite old logs when running lconf --write_conf (2264)
5669 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5670 - fix race in target_handle_connect (2898)
5671 - mds_reint_create() should take same inode create lock (2926)
5672 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5673 - don't close files for self_export to avoid uninitialized obd (2936)
5674 - allow MDS with the same name as client node (2939)
5675 - hold dentry reference for closed log files for unlink (2325)
5676 - reserve space for all logs during transactions (2059)
5677 - don't evict page beyond end of stripe extent (2925)
5678 - don't oops on a deleted current working directory (2399)
5679 - handle hard links to targets without a parent properly (2517)
5680 - don't dereference NULL lock when racing during eviction (2867)
5681 - don't grow lock extents when lots of conflicting locks (2919)
5683 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5686 - account for cache space usage on clients to avoid data loss (974)
5687 - lfsck support in lustre kernel code (2349)
5688 - reduce journal credits needed for BRW writes (2370)
5689 - orphan handling to avoid losing space on client/server crashes
5690 - ptlrpcd can be blocked, stopping ALL progress (2477)
5691 - use lock value blocks to assist in proper KMS, faster stat (1021)
5692 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5693 - recovery for initial connections (2355)
5694 - fixes for mds_cleanup_orphans (1934)
5695 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5696 - block all file creations until orphan recovery completes (1901)
5697 - client remove rq_connection from request struct (2423)
5698 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5699 - recovery timer race (2670)
5700 - mdc_close recovey bug (2532)
5701 - ptlrpc cleanup bug (2710)
5702 - mds timeout on local locks (2588)
5703 - namespace lock held during RPCs (2431)
5704 - handle interrupted sync write properly (2503)
5705 - don't try to handle a message that hasn't been replied to (2699)
5706 - client assert failure during cleanup after abort recovery (2701)
5707 - leak mdc device after failed mount (2712)
5708 - ptlrpc_check_set allows timedout requests to complete (2714)
5709 - wait for inflight reqs when ptlrpcd finishes (2710)
5710 - make sure unregistered services are removed from the srv_list
5711 - reset bulk XID's when resending them (caught by 1138 test)
5712 - unregister_bulk after timeout
5713 - fix lconf error (2694)
5714 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5715 - readahead locks pages, leaves pending causing memory pressure (2673)
5716 - increase OST request buffers to 4096 on large machines (2729)
5717 - fix up permission of existing directories in simple_mkdir (2661)
5718 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5719 - don't assign transno to errored transactions (2742)
5720 - don't delete objects on OST if given a bogus objid from MDS (2751)
5721 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5722 - if rq_no_resend, then timeout request after recovery (2432)
5723 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5724 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5725 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5726 - MDS recovery shouldn't delete objects that it has given out (2730)
5727 - if enqueue arrives after completion, don't clobber LVB (2819)
5728 - don't unlock pages twice when trigger_group_io returns error (2814)
5729 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5730 - don't write pages to disk if there was an error (1450)
5731 - don't ping imports that have recovery disabled (2676)
5732 - take buffered bytes into account when balancing socknal conn (2817)
5733 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5734 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5735 - remove little-endian swabbing of llog records (1987)
5736 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5737 - retry reposting request buffers if they fail (1191)
5738 - grow extent at grant time to avoid granting a revoked lock (2809)
5739 - lock revoke doesn't evict page if covered by a second lock (2765)
5740 - disable VM readahead to avoid reading outside lock extents (2805)
5742 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5743 - updated kernel patches for hp-2.4.20 kernel (2681)
5745 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5748 - fix truncated write corruption (2366)
5749 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5751 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5752 - ensure len(uuid) < 37 in lmc (1171)
5753 - fix ia64 OOPS in llog_test (2255)
5754 - zero end of page at obdfilter for partial page writes (2648)
5755 - don't leave stale dentries around after renames (bug 2428)
5756 - fix timeouts when evicting a client with a single lock held (2642)
5757 - set deadline for the initial HELLO message to drain (2634)
5758 - print out dotted-quad IP addresses in the socknal (2302)
5760 - additional debugging for MDS client eviction problem (2443)
5761 - fix mkfsoptions support for osts (2603, 2604)
5763 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5766 - add series for the vanilla 2.6.0 kernel
5767 - add series for the vanilla 2.4.24 kernel
5768 - add series for a cray x86/64 UL kernel drop
5769 - fix xattr patches for the vanilla 2.4.19 series
5771 - generate true UUIDs in lmc (1171)
5772 - have portals stack dumping break in UML (2466)
5773 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5774 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5775 - wake callback waiting threads on client eviction (2460)
5776 - Add --ptldebug and --subsystem to lmc (1719)
5777 - update assertion to allow safe interrupt allocation
5778 - set rq_no_resend for cancel requests (2432)
5779 - recalculate ptlrpcd timeout after resend (2494)
5780 - call vfs_rmdir when removing pending directories (2368)
5781 - fix renaming a file to itself (2429)
5782 - lmc creates a default one-stripe lov (2454)
5783 - expand procfs space to handle large clusters (2326)
5784 - increase UML stack to avoid overflow
5785 - update lconf's list of debug and subsystem masks
5786 - fix lfs find --obd (2510)
5787 - /proc tunable for disabling filter read caching (2591)
5788 - stop rpm packages from altering slapd.conf (2301)
5789 - disable nagle in the socknal under 0conf (2578)
5790 - choose mds inode size based on stripe count (2572)
5791 - fix kernel-source rpm problems (2516)
5793 - add --disable-doc to avoid pdf generation (2421)
5794 - update documentation, tests, type-os, comments
5795 - avoid format warnings on ia64
5796 - remove the TOE NAL
5797 - tiny code cleanups by removing unused fields
5799 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5802 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5803 - remove the most blatant lies from BUILDING file (2371)
5804 - change default debug level to reasonable production setting
5805 - reduce client side cache size to reduce cache flush time
5806 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5807 - make TCP zerocopy and pinger support enabled by default (2476)
5808 - sync writes completed after process exits caused crashes (2319)
5809 - maintain correct mount count on the MDS (2356)
5810 - backout 1557, because 2316 wasn't really fixed
5811 - better file I/O statistics gathering in /proc
5812 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5813 - another kernel patch to fix zero-copy TCP function export
5814 - don't take duplicate lock when processing re-sent getattr (2420)
5815 - lctl uses obd_self_export instead of creating new conn (2353)
5816 - MDS/OST recovery case which requires object creation asserted (2425)
5817 - move lfs from /usr/sbin to /usr/bin in packages
5818 - fix race between mds_client_add and mds_client_free (2417)
5819 - use kmalloc instead of slabs in portals (2430)
5820 - don't create duplicate records when a failover MDS is present (2442)
5821 - remove unnecessary mount age check (2332)
5822 - don't remove directory inodes from locks prematurely (2451)
5823 - don't break if MDS service name is the same as hostname (2103)
5824 - fix races in client write RPC generation when cache full (2482)
5826 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5829 - remove now-unused request->rq_obd (278)
5830 - if an allocation fails, print out how much memory we've used (1933)
5831 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5832 - add the "configurable stack size" patch to most series files (1256)
5833 - ability to write large log records, for 100+ OST configs (2306)
5834 - fix NULL deref when filter_prep fails (2314)
5835 - fix operator precedence error in filter_sync
5836 - dynamic allocation of socknal TX descriptors (2315)
5837 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5838 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5839 - ensure that CWARN messages in Portals always get to the syslog
5840 - __init/__exit are not for prototype decls (ldlm_init/exit)
5841 - x86-64 compile warning fixes
5842 - fix gateway LMC keyword conflict (2318)
5843 - fix MDS lock inversions in getattr/reint paths (1844)
5844 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5845 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5846 - DLM race condition prevented some lock evictions (2328)
5847 - ENOMEM detection and retry on socknal sends (2230)
5848 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5849 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5850 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5851 - better error messages when a client is rejected during recovery (1505)
5852 - avoid cancelling locks which were never granted, after failure (2330)
5853 - fix i_sem/journal inversion in mds_client_add (2333)
5854 - fix truncate/getattr lock cycle deadlock (2334)
5855 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5856 - fix two rare exit paths which could leak an l_lock() ref (2321)
5857 - fencepost error in MDS/OST orphan recovery (2226)
5858 - make log record alignment 8 bytes (1988)
5859 - lstripe now fails when requested offset > ost_count (2237)
5860 - ensure that all kernel series have a complete list.h (1607)
5861 - fix crashes in special-file operations (2316)
5862 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5864 - allow configurable automake binary, for testing new versions
5865 - small update to the lfs documentation
5867 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5869 * fix negative export reference count in fsfilt_sync (2312)
5871 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5872 * release candidate 0.9.1
5874 - orphans are moved into the PENDING directory for possible recovery
5875 - replayed opens now open by fid for orphan/rename safety (1042)
5876 - last close of an orphan inode generates a transno (683)
5877 - chdir() and mount() now pin the directory entry (1020)
5878 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5879 - discard very old requests without processing them (1502)
5880 - remove some common, well-understood CERRORs (1505)
5881 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5882 - clear "grant" flags in OST replies until OST grant code lands (1644)
5883 - fix read performance by not clobbering i_blksize on client (1598)
5884 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5885 - make LRU size a /proc tunable, clears locks when reduced (707)
5886 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5887 - server locks take references on exports now (1558)
5888 - build fixes for 2.4.20-rh trees (1663)
5889 - return an error from lov_create if all OSCs are inactive (1751)
5890 - fix import levels when a reconnect happens without a timeout (1597)
5891 - exit early from mds_open if we get a lookup error (1749)
5892 - partial page read at EOF wouldn't wait for disk before sending (1642)
5893 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5894 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5895 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5896 - write/truncate lock inversion (1639)
5897 - Don't auto-load obdclass, portals modules during cleanup (1495)
5898 - fix timestamps from jumping to "now" (1763)
5899 - extra journal assertions (1648)
5900 - add an extra multiunlink test (1771)
5901 - fix read_record/write_record API (1776)
5902 - fix leak of offset_extent, possible incorrect i_size later (1772)
5903 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5904 - Debugging for the kqswnal_get_idle_tx problems (1820)
5905 - Allow recovery to be attempted multiple times (1536)
5906 - Write out MDS last_rcvd file after it is first created (1600)
5907 - Fix tx_descriptor leak in failed transmit situations (1827)
5908 - ext3 journaling fixes for assertion failure after IO error (1871)
5909 - class_export_put() on freed export after completion AST error (1896)
5910 - Fix revalidate looping in VFS (1322)
5911 - Don't access a freed export during MDS_REINT timeout (1521)
5912 - Add open-unlink recovery support on the MDS (1673,1764)
5913 - Return an error if no MDS data was read from last_rcvd (1946)
5914 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5915 - Don't setup a disk device that doesn't match exported UUID (317)
5916 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5917 - avoid committing NULL handle in force close
5918 - local.sh is now a one-stripe LOV configuration
5919 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5920 - fix ext3 htree duplicate directory entry corruption (1516)
5921 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5922 - update to new LOV EA format (2097)
5923 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5925 2003-06-15 Phil Schwan <phil@clusterfs.com>
5928 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5929 - per-import/export recovery handling (958, 931, 959)
5930 - multiple last-rcvd slots, for serving multiple FSes (949)
5931 - connections are again shared between multiple imp/exports (963, 964)
5932 - "umount -f" would hang if any requests needed to be sent (393, 978)
5933 - avoid pinning large req buffer by copying for queued messages (989)
5934 - add "uuid" to "lctl device" command to help upcalls (991)
5935 - "open" RPCs with transnos would confuse recovery counters (1037)
5936 - do proper endian conversion of all wire messages (288, 340, 891)
5937 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5938 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5939 - fix ptlrpc_connection leak in target_handle_connect (1174)
5940 - fix import refcounting bug in OST and MDS cleanup (1134)
5941 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5942 - fix very unlikely obd_types race condition (501)
5943 - remove osc_open hack for echo_client (1187)
5944 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5945 - a failure in read_inode2 leads to deadlock (1139)
5946 - cancel ack-locks as soon as transaction is committed (1072)
5947 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5948 - make sure to commitrw after any preprw to avoid deadlock (1162)
5949 - failing to execute a file in a lustre FS would lock inode (1203)
5950 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5951 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5952 - obd_uuid2tgt would walk off the end of the list (1255)
5953 - on IA64 the osc would give portals incorrect bulk size (1258)
5954 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5955 - fix lock inversion caused by new llite matching code (1282)
5956 - limit the number of dirty pages on a client to 10MB (1286)
5957 - timed out locks were not being corrected cancelled (1289)
5958 - fix O_DIRECT above 4GB on IA-32 (1292)
5959 * major user-visible changes
5960 - fail out/fail over policy now controlled by the upcall (993)
5962 - add OBD_PING to check server availability and failure (954)
5963 - lustre messages are now sent in sending host order (288, 340, 891)
5964 - add eadatalen to MDS getattr reply (340)
5965 - OST read replies may contain second buffer, with per-page status (593)
5967 2003-03-11 Phil Schwan <phil@clusterfs.com>
5970 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5971 - failed open()s could cause deadlock; fixed (867, 869)
5972 - stop cancelling OST locks when files are closed (481)
5973 - overlapping XID spaces caused network corruption (851, 853)
5974 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5975 - setattr_raw added, to do single-RPC, server-side setattrs
5976 - lmc/lconf syntax change for OST UUIDs
5977 - fix crashy race condition between ptlrpc_free_req and osc_close
5978 - don't use request in mdc_enqueue if we hit a timeout (889)
5979 - don't set the inode i_size for regular files from the MDS (896)
5980 - handle out of order completion AST (842)
5981 - don't LBUG if a lock request times out after receiving AST (913)
5982 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5983 - if a bad lock AST arrives, send an error instead of dropping entirely
5984 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5985 - fix leak in bulk IO when only partially completed (899, 900, 926)
5986 - fix O_DIRECT for ia64 (55)
5987 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5988 - C-z after timeout could hang a process forever; fixed (977)
5990 - client-side I/O cache (678, 924, 929, 941, 970)
5992 - READPAGE and SETATTRs which don't take server-side locks get
5995 2003-02-11 Phil Schwan <phil@clusterfs.com>
5998 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5999 - Fix fsfilt_extN_readpage() to read a full page of directory
6000 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6001 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6002 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6003 - Most intent operations were removed, in favour of a new RPC mode
6004 that does a single RPC to the server and bypasses most of the VFS
6005 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6006 - Aggressively cancel local locks on DLM servers
6007 - mds_reint_unlink sends EA to the client if it's the last nlink.
6008 client uses that EA to unlink OST objects.
6009 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6010 - recursive symlinks were fixed (439)
6011 - fixed NULL deref in DEBUG_REQ
6012 - filter_update_lastobjid no longer calls sync, which annoyed extN
6013 - fixed multi-client small-writes to a single file problem (445)
6014 - fixed mtime updates during file writes (607)
6015 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6016 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6017 - fixed Linux OST size reporting problem (444, 656)
6018 - OST now updates object mtime with writes or setattr (607, 619)
6019 - client verifies file size before zeroing page past EOF (445)
6020 - OST now writes last allocated objid to disk with allocation (108)
6021 - LOV on echo now works (409)
6023 - mds_reint_unlink sends a new buffer, with the EA included. this
6024 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6025 set if a regular file was being unlinked, and it was the last link
6026 - use PtlGet from the target for bulk writes (315)
6027 - OST now updates object mtime with writes or setattr (607, 619)
6028 - LDLM now has a grant-time callback to revalidate locked items, if
6030 - Many MDS operations were reorganized to combat race conditions
6032 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6033 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6035 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6038 - Fully reactivate OST imports after reconnection (512, others)
6039 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6040 - More graceful error handling for truncating on dead OST (515)
6041 - Don't error out unless we're actually accessing dead stripes (474)
6042 - Fix garbage sizes when stripes are missing (410)
6043 - LRU counters were broken, causing constant lock purge (433, 432)
6044 - garbage on read from stripes with failed OSTs (441)
6045 - mark OSCs as active before reconnecting during recovery (438)
6046 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6047 - lfind did not preserve OST order in output (443)
6048 - symlinks cause hung clients, incorrect data (439)
6049 - stop dereferencing request after dropping refcount (457)
6050 - don't LASSERT(spin_is_locked) on non-SMP (455)
6051 - fixes for many rename() bugs
6052 - fstat didn't correctly synchronize attributes (399)
6053 - server must handle lock cancellation during blocking AST prep (487)
6054 - bulk descriptors were free()d too soon (511)
6055 - fix paths in lconf, which would load incorrect modules (451, 507)
6056 - fix confusing lconf 'host not found' error message (386)
6057 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6058 - fix race condition in mdc_blocking_ast() for inode access (526)
6059 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6060 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6061 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6062 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6063 - fix lstripe in several ways (kernel side) (527)
6064 - fix request leak in ldlm_cli_enqueue (262)
6065 - incorrect OSC was marked inactive after OST failure
6066 - call mds_fs_cleanup before unmounting filesystem (524)
6067 - fix races between taking ns_lock and ldlm_lock_change_resource
6068 - fix races updating LOV export open file list
6069 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6070 - fix recovery NULL deref in ldlm_cli_cancel_unused
6071 - fix some DLM races by using new hash table for lock handles (419)
6072 - permit the client to specify desired inodes, at replay
6073 - duplicate requests when we queue them for replay reintegration
6074 - fix last_rcvd offset calculation
6075 - sync after each recovered transaction, so we always make progress
6076 - never, not always, ERESTART requests without transnos
6077 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6079 - skip replay if the MDS says that the client is already connected
6080 - don't check for a recovery-enabled export to match lctl's UUID
6081 - don't INC_USE_COUNT for phantom exports
6082 - don't crash when cleaning up phantom exports (567)
6083 - don't double-finish or set replay data for errored mdc_open requests
6084 - abort requests when they time out, so we don't get old replies
6085 - send/receive replies for AST messages again
6086 - if the client says that it doesn't have the lock, cancel it on the
6088 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6089 mark it as destroyed, it will all work out when decref is called
6090 - fix module use counts (22, 581)
6092 - ASTs now expect a reply (server cancels lock on error reply)
6094 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6097 - fix many simultaneous client startup (392)
6098 - fix dentry->d_it clobbering
6099 - credentials weren't being shipped for readdir/getattr operations
6100 - remove invalid assertions triggered during some concurrent MD
6102 - proper Lustre versions added (336, 389)
6103 - fix memory leak for create error case (398)
6104 - fix LOV locking bug that would get cli/srv out of sync
6105 - fix echo client over LOV (409)
6106 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6107 - fix double-O_EXCL intent crash (424)
6108 - avoid sending multiple lock CANCELs (352)
6110 - MDS can do multi-client recovery (modulo bugs in new code)
6112 - many updates, edits, cleanups
6114 2002-11-18 Phil Schwan <phil@clusterfs.com>
6117 - fix null d_it dereference (346)
6118 - fix full OST/dbench hang (333)
6119 - fix permission problem with file removal (286)
6120 - fix removal of OSCs from LOV when they fail
6121 - fix NULL deref during bulk timeout (214)
6122 - fix problems related to multiple filesystems on one MDS (241)
6123 - fixed serious subtle metadata locking bugs
6124 - free locks on clients when inodes are removed due to memory
6126 - fix inode pointer in lock data (285)
6127 - partial support for multiple MDS on a single host (241)
6128 - data locks weren't cancelled at clear_inode time (290, 311)
6129 - intent locks could lead to unbounded lock growth (205)
6130 - added a maximum lock count, an LRU list, and a flusher
6131 - fix multiple rename (365)
6132 - properly abstracted the echo client
6133 - OSC locked 1 byte too many; fixed
6134 - rewrote brw callback code:
6135 - fixed recovery bugs related to LOVs (306)
6136 - fixed too-many-pages-in-one-write crash (191)
6137 - fixed (again) crash in sync_io_timeout (214)
6138 - probably fixed callback-related race (385)
6140 - Add capability to MDS protocol
6141 - LDLM cancellations and callbacks on different portals
6143 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6146 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6147 - defer open object destruction to close time (601981, 138)
6148 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6149 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6150 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6151 - don't call obd_setattr in ll_file_release for destroyed objects
6152 * protocol change to lustre_msg: move |version| and add |flags|
6153 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6154 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6155 * added replay of create, unlink, link and rename operations during
6156 MDS failover; recovery should be much more robust now
6157 * remove failed OSCs from LOVs (only lov_create uses this so far)
6158 * the lustre-HOWTO was brought (more) up to date (582544)
6160 2002-10-23 Phil Schwan <phil@clusterfs.com>
6163 - in-use dentries weren't being reused properly (617851)
6164 - prevent multiple LDLM setup (599178)
6165 - fix LOV size calculations for truncate (617853)
6166 - fix client handling of MDS intent errors (POSIX)
6167 - fix permission bug in lovstripe.c test (624321)
6168 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6169 - truncate past end of file could corrupt data
6170 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6171 - a race in recovery could return ETIMEDOUT to apps (623947)
6172 - building outside the source directory was fixed
6173 * the lustre-HOWTO was brought (more) up to date (582544)
6174 * major progress was made on recovery functionality
6176 2002-10-10 Phil Schwan <phil@clusterfs.com>
6179 - recovery deadlock fix
6180 - rm -rf causes LBUG fix (617817)
6181 - file open by multiple tasks fix (618962)
6182 - directory permissions bugs (602707 and 620007)
6183 - journal_stop fixed with locking (611313)
6184 - O_APPEND failures resolved (618273, perhaps 614459)
6185 - lconf PATH fix (619770)
6186 - IA64 build fix (621450)
6187 - RPC buffer sizes scale with amount of memory
6189 2002-10-01 Phil Schwan <phil@clusterfs.com>
6192 - locks would be cancelled without throwing away data pages,
6193 resulting in inconsistent data (605627)
6194 - inode attributes were not always being refreshed (605627, 612449)
6195 - lconf now continues to cleanup after lctl reports an error
6196 - MDS now enforces user permissions (602707)
6197 - lprocfs cleanup fixed, but not yet enabled (614157)
6198 - fixed infinite server hang, should a client not respond to an AST
6199 - avoid going into recovery if user calls readlink() with a buffer
6200 that's too small (613941)
6201 - AST RPCs no longer require replies (614867) -- this may be changed
6202 - don't crash server if client sends an IOV that's too big (611336)
6203 - fixed lock conversion deadlock (611892)
6204 - fixed the following of symlinks (614622)
6205 * recovery: the server can remove locks from a client that dies, other
6206 clients can make progress
6207 * more extN patch fixes
6208 * compile-time configurable ptlrpc buffer allocations
6210 - collaborative read cache document
6211 - Lustre Lite Performance CDR document-in-progress
6213 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6216 - fix typo in patch-2.4.18
6218 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6221 - clear ptlrpc request each time in handle_incoming_request()
6222 - unlink of files now destroys the object on the OST
6224 2002-09-19 Peter Braam <braam@clusterfs.com>
6226 * add hard link support
6227 * change obdfile creation method
6228 * kernel patch changed
6230 2002-09-19 Peter Braam <braam@clusterfs.com>
6233 - stack overflow bug in extN fixed
6235 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6237 * documentation updates
6238 - add man pages for config tools
6239 - update tests/README to describe testing with new config tools
6240 - finish metadata API descriptions
6241 * bug fixes and cleanups
6242 - statfs workaround for 16TB limit
6243 - LOV stripe allocation improved, can stripe on subset of OSTs
6244 - LOV file size/IO offset was wrong for files > 4GB in size
6245 - object EA data was being dropped, caused files to be unreadable
6246 - memory overflow with non-LOV OST caused memory corruption
6247 - fixed regression tests to work with new config tools, obdfilter
6248 - fixed bug when directory size became larger than 1 block
6249 - fixed bug (for single client case) when PWD was deleted
6250 - invalidate local directory pages when doing intent-based ops
6251 - avoid LDLM oops when lock callback contained bad data
6253 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6255 * documentation updates
6256 * bug fixes and cleanups
6257 - configuration tools
6260 - 64-bit compile warnings
6261 - 64-bit internal statfs data
6263 * test_brw on persistent OST devices
6265 * lprocfs (disabled)
6267 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6269 * documentation updates
6270 * bug fixes and cleanups
6271 * configuration tools
6273 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6276 * many small fixes to 0_5_4
6277 * io/network handling
6278 * thinkos in MDS operations
6280 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6283 * crucial basic fixes to 0.5.3
6284 * IOR, Iozone work over Elan
6287 2002-08-07 Phil Schwan <phil@clusterfs.com>
6288 * version 0_5_3, our first alpha
6289 * we use the new Portals iovs
6290 * documentation updates
6291 * bug fixes and cleanups
6292 * small changes in the DLM wire protocol
6294 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6295 * version 0_5_1 with some initial stability,
6296 * locking on MD and file I/O.
6297 * documentation updates
6298 * several bug fixes since 0.5.0
6299 * small changes in wire protocol
6301 2002-07-18 Phil Schwan <phil@clusterfs.com>
6303 * delivered as Lustre Light Alpha
6304 * fixed a crash after handling invalid MDS requests
6305 * fixed directory pages for architectures with non-4k pages sizes
6307 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6308 * release version v0_4_4
6309 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6311 * Fixes a number of interruption problems with OST operations.
6312 * Update documentation for portals header changes
6313 * Move all wire protocol structs/defines to lustre_idl.h
6314 * Fixes symlink length bug.
6315 * Add tcpdump to repository.
6317 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6318 * release version v0_4_3
6319 * Fixes statfs for inodes on extN.
6320 * Fixes bug in runtests which would delete /etc/hosts.
6321 * Use 64-bit object IDs wherever possible (not into VFS though)
6322 Remove ost_get_info, which is unused by lustre, and out of date.
6324 2002-07-03 Peter Braam <braam@clusterfs.com>
6325 * release version v0_4_2 Fixes a lookup error (type not passed)
6326 * move forward to head of Portals
6327 * move forward to latest Lustre kernel
6329 2002-06-25 Peter Braam <braam@clusterfs.com>
6330 * release version v0_4_1. Hopefully stable on single node use.