1 27-08-2012 Whamcloud, Inc.
3 * Server support for kernels:
4 2.6.18-308.20.1.el5 (RHEL5)
5 2.6.32-279.14.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-308.20.1.el5 (RHEL5)
9 2.6.32-279.14.1.el6 (RHEL6)
10 2.6.29.4-167.fc11 (FC11)
11 2.6.16.60-0.69.1 (SLES10)
13 * Recommended e2fsprogs version: 1.41.90.wc3
14 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
15 Lustre filesystem with 4K stack may cause a stack overflow. For more
16 information, please refer to bugzilla 17630.
17 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
18 mounted with "32bitapi" mount option
20 Severity : enhancement
22 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
24 Severity : enhancement
26 Description: Kernel update [RHEL5.8 2.6.18-308.11.1.el5]
28 Severity : enhancement
30 Description: Kernel update [RHEL6.2 2.6.32-220.17.1]
32 Severity : enhancement
34 Description: Kernel update [RHEL5.8 2.6.18-308.4.1.el5]
36 Severity : enhancement
38 Description: Kernel update [RHEL6.2 2.6.32-220.13.1]
40 Severity : enhancement
42 Description: Kernel update [RHEL6.2 2.6.32-220.4.2]
44 Severity : enhancement
46 Description: Kernel update [RHEL5.7 2.6.18-274.18.1.el5]
48 Severity : enhancement
50 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
52 Severity : enhancement
54 Description: Kernel update [RHEL5.7 2.6.18-274.12.1.el5]
56 Severity : enhancement
57 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
59 Severity : enhancement
60 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
62 Severity : enhancement
63 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
66 Frequency : only with SLES10
68 Description: Use OFED "KMP" provided by Novell
69 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
70 Lustre is now built against this package. That means you need to
71 install the ofed-kmp package from Novell for the patchless client.
72 Note that the ofed-kmp that Novell ships may not exactly match the
73 kernel version but should still be compatible.
75 Severity : enhancement
77 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
78 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
80 Severity : enhancement
82 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
84 --------------------------------------------------------------------------------
86 2010-07-15 Oracle, Inc.
88 * Server support for kernels:
89 2.6.18-164.11.1.el5 (RHEL 5)
90 2.6.18-164.11.1.0.1.el5 (OEL 5)
91 * Client support for unpatched kernels:
92 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
93 2.6.18-164.11.1.el5 (RHEL 5),
94 2.6.18-164.11.1.0.1.el5 (OEL 5)
95 2.6.16.60-0.42.8 (SLES 10),
97 2.6.29.4-167.fc11 (FC11)
98 * Recommended e2fsprogs version: 1.41.10-sun2
100 Severity : enhancement
102 Description: Distribute lfs_migrate script with lustre
103 Details : The lfs_migrate mentioned in the manual is now included in the
104 lustre tree in lustre/scripts and will distributed in lustre RPMs.
106 Severity : enhancement
108 Description: Failover nid list added to import proc file.
110 Severity : enhancement
112 Description: "som_preview" mount option required to use SOM
113 Details : The clients will need to use som_preview mount option
117 Description: Lustre does not do 1MB IOs to HW RAID
118 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
120 Severity : enhancement
122 Description: lfs quota output is not very convenient for awk/sed-parsing
123 Details : Some positions in lfs quota output table could be empty or
124 non-empty which made it hard to parse it with scripts, now
125 a dash is put instead of space where there is not supposed
130 Description: Update to OFED 1.5.1
131 Details : OFED 1.5.1 will be used in environments that don't use
132 vendor supplied OFED.
136 Description: Remove lnet. prefix from lctl params display
137 Details : Also renamed lnet.memused to lnet_memused to
142 Description: Use conf_param -d to remove permanant settings.
146 Description: Add list_param -R
147 Details : lctl list_param now supports recursive mode.
150 Bugzilla : 21829, 22312
151 Description: Fix llobdstat.
152 Details : 1. Now updating the write_bytes counter
153 2. Port script fixes from 1.8.
157 Description: decrease the usage of memory on clients.
158 Details : 1. On clients, recycle dentries and inodes unused.
159 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
164 Description: llobdstat fix and enhancement
165 Details : add a counter to set a limit to how many samples will be returned
166 fix a wildcard in the path to limit to obdfilter stats only
170 Description: lfs check only allowed for root
171 Details : Code cleanups and added a regresion test of non-root use of
176 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
177 Details : Fix hr->hr_index overload.
181 Description: Fix FIEMAP functionality.
185 Description: FC11 patchless client support added
189 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
190 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
194 Description: New acc-sm script and various test-framework enhancments.
198 Description: Previously landed patch was reverted.
202 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
206 Description: Add ldiskfs maxdirsize mount option.
210 Description: Fix mkfs.lustre support for 16TB LUNs.
214 Description: Print mmp_check_interval
215 Details : print mmp_check_interval and make it possible to abort
216 mount operation in case it takes too long.
220 Description: Add cascading_rw to lustre mpi tests.
224 Description: Store SOM data in EA.
225 Details : Also several SOM related cleanups.
229 Description: Add refcount to md_open_data
233 Description: Use refcount on lov_request_set to prevent memory corruption.
237 Description: Statahead: use parent dir's i_mutex for synchronization.
241 Description: loadgen improvements.
245 Description: Enable peer health detection in ksocklnd and ko2iblnd.
249 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
253 Description: Fix race on obd_devcie:md_stats
257 Description: Fix SLES 11 client builds.
258 Details : Resolved uninitalized pointer gcc warnings.
262 Description: Allow non-root access for "lfs check".
263 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
267 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
269 Severity : enhancement
271 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
272 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
276 Description: Sanity test_27p fails.
277 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
278 Sanity tests 27 were cleaned up and improved.
282 Description: Unify req format on client/servers
283 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
284 buffers and swabbers approach..
289 Description: drop join file stuff.
291 Severity : enhancement
293 Description: Add 2.6.27 and SLES11 (patchless client) support
297 Description: Fix in ptlrpc_expire_one_request() to print the signed time
300 Severity : enhancement
302 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
305 Severity : enhancement
307 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
309 Severity : enhancement
311 Description: correctly handle big reply message.
312 Details : send LNet event if reply is bigger then buffer and adjust this buffer
317 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
319 Severity : enhancement
321 Description: Add support for OEL5.
325 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
327 Severity : enhancement
329 Description: "weak-modules" support
330 Details : Implement "weak-modules" support which enables kernel modules
331 to be used with any kernel that implements the same kABI. In
332 order to achieve this modules are now installed in
333 /lib/modules/$(uname -r)/updates/kernel on all distributions.
337 Description: client could not reconnect to OST because of an active request.
338 Details : The client now retries to reconnect to the same server, if a
339 connect request failed with EBUSY or -EAGAIN. Test case 35b
340 conf-sanity.sh has been created.
344 Description: Don't allow make backward step on assignin osc next id.
345 Details : race between allocation next id and ll_sync thread can be cause
346 of set wrong osc next id and can be kill valid ost objects.
350 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
354 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
358 Description: Error handling in osc_statfs_interpret() has been improved.
359 Details : Check in osc_statfs_interpret() for EBADR.
363 Description: Do not send statfs() requests to OSTs disabled by administrator.
364 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
366 Severity : enhancement
368 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
370 Severity : enhancement
372 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
374 Severity : enhancement
376 Description: Enhancement for lfs(1) command to use numeric uid/gid.
380 Description: Improved cleanup procedure for conf-sanity test_46a.
384 Description: Do not start bulk transfer, if client tried to reconnect.
385 Details : Do not start bulk transfer, if client tried to reconnect,
386 but failed due to active rpcs. Check for exp_abort_active_req
387 has been implemented.
391 Description: Prevent inconsistences between linux and lustre mount structures.
392 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
393 Make the sleep interruptible.
395 Severity : enhancement
397 Description: provide server to client comms path
398 Details : server to client communications path via new LDLM_SET_INFO rpc
402 Description: truncate starts GFP_FS allocation under transaction causing deadlock
403 Details : ldiskfs_truncate calls grab_cache_page which may start page
404 allocation under an open transaction. This may lead to
405 calling prune_icache with consequent lustre reentrance.
409 Description: send correctly lsm on open replay
410 Details : MDS is trust to LSM size on replay open, but client can set wrong size
415 Description: enable client interface failover
416 Details : When a child reconnects from another NID, properly update export
417 nid hash position and ldlm reverse import.
419 Severity : enhancement
421 Description: Communicate OST degraded/readonly state via statfs to MDS
422 Details : Flags in the statfs returned from OSTs indicate whether the
423 OST is in a degraded RAID state, or if the filesystem has
424 turned read-only after a filesystem error is detected.
429 Description: don't panic if EPROTO was hit when reading symlink
430 Details : correctly handling request reference in error cases.
432 Severity : enhancement
434 Description: Add LustreNetLink, a kernel-userspace communcation path.
436 Severity : enhancement
438 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
443 Description: access to llog context before init.
444 Details : move handling CATALOGS file at osc layer and forbid access to llog
448 Frequency : rare, only when client is writing to OST on local node
450 Description: Avoid deadlock for local client writes
451 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
452 memory freeing context. This allows OST threads to set the
453 PF_MEMALLOC flag on task structures in order to allocate memory
454 from reserved pools and complete IO.
455 Use GFP_HIGHUSER for OST allocations for non-local client writes,
456 so that the OST threads generate memory pressure and allow
457 inactive pages to be reclaimed.
459 Severity : enhancement
461 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
466 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
467 Details : move ldlm namespace creation in setup phase to avoid grab
468 _lprocfs_lock with cli_sem held.
472 Description: Temporarily disable grant shrink.
473 Details : Disable the feature for debugging.
476 Frequency : only when formatting test filesystems
478 Description: Unable to run several mkfs.lustre on loop devices at the same time
479 Details : mkfs.lustre returns error 256 on the concurrent loop devices
480 formatting. The solution is to proper handle the error.
482 Severity : enhancement
484 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
486 Severity : enhancement
488 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
493 Description: resolve race between obd_disconnect and class_disconnect_exports
494 Details : if obd_disconnect will be called to already disconnected export he
495 forget release one reference and osc module can't unloaded.
497 Severity : enhancement
499 Description: Allow tuning service thread via /proc
500 Details : For each service a new
501 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
502 created that can be used to set min/max thread counts, and get the
503 current number of running threads.
508 Descriptoin: don't return error if have partially created objects for file.
509 Details : lov_update_create_set uses set->set_success as index for created
510 objects, so if some requests will be failed, they will have hole
511 at end of array and qos_shrink_lsm for allocate correct lsm.
513 Severity : enhancement
515 Description: Update OFED support to 1.4.1
520 Description: Reduce small size read RPC
521 Details : Set read-ahead limit for every file and only do read-ahead when
522 available read-ahead pages are bigger than 1M to avoid small size
525 Severity : enhancement
527 Description: move AT tunable parameters for more consistent usage
528 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
530 Severity : enhancement
532 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
533 Details : allow skip disconnected ost for send statfs request and hide error
539 Description: fix racy locking of mballoc block bitmaps causing BUG
540 Details : The locking of the mballoc buddy bitmap and the in-memory
541 block bitmap was using two different spin locks in some
542 cases. This made it possible to incorrectly access the
543 mballoc bitmap while another process was modifying it,
544 causing a sanity assertion to fail. While no on-disk corruption
545 was reported, there was some risk of this happening.
548 Frequency : rare, on llog test 6
550 Descriptoin: don't allow connect to already connected import
551 Details : allowing connect to already connected import is hide connecting problem.
556 Descriptoin: allow kill process which wait statahead result
557 Details : for some reasons 'ls' can stick in waiting result from statahead,
558 in this case need way for kill this process.
560 Severity : enhancement
562 Description: Add state history info file, enhance import info file
563 Details : Track import connection state changes in a new osc/mdc proc file;
564 add overview-type data to the osc/mdc import proc file.
566 Severity : enhancement
568 Description: MDS create should not wait for statfs RPC while holding DLM lock.
570 Severity : enhancement
572 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
577 Descriptoin: OST grant too much space to client even there are not enough space.
578 Details : Client will shrink its grant cache to OST if there are no write
579 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
580 this grant cache if there are already not enough avaible space
581 (left_space < total_clients * 32M).
584 Frequency : start MDS on uncleanly shutdowned MDS device
586 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
587 Details : stay in waiting mds<>ost recovery finished produce random bugs
588 due race between two ll_sync thread for one lov target. send
589 ACTIVATE event only if connect realy finished and import have
593 Frequency : rare, connect and disconnect target at same time
595 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
596 Details : don't call obd_disconnect under lov_lock. this long time
597 operation and can block ptlrpcd which answer to connect request.
602 Descriptoin: don't lose wakeup for imp_recovery_waitq
603 Details : recover_import_no_retry or invalidate_import and import_close can
604 both sleep on imp_recovery_waitq, but we was send only one wakeup
608 Frequency : always with long access acl
610 Descriptoin: mds can't pack reply with long acl.
611 Details : mds don't control size of acl but they limited by reint/getattr
614 Severity : enhancement
616 Description: Update to SLES10 kernel-2.6.16.60-0.33.
618 Severity : enhancement
620 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
623 Frequency : start MDS on uncleanly shutdowned MDS device
625 Descriptoin: aborting recovery hang on MDS
626 Details : don't throttle destroy RPCs for the MDT.
629 Frequency : on remount
631 Description: external journal device not working after the remount
632 Details : clear dev_rdonly flag for external journal devices in
638 Description: shutdown vs evict race
639 Details : client_disconnect_export vs connect request race.
640 if client will evicted at this time - we start invalidate
641 thread without referece to import and import can be freed
645 Frequency : rare, need acl's on inode.
647 Description: client can't handle ost additional correctly
648 Details : if ost was added after client connected to mds client can have
649 hit lnet_try_match_md ... to big messages to wide striped files.
650 in this case need teach client to handle config events about add
651 lov target and update client max ea size at that event.
653 Severity : enhancement
655 Description: Changelogs
656 Details : Changelogs are a lightweight mechanism to track filesystem
657 metadata and namespace changes. The changelog is recorded
658 permanently on the MDTs, and is periodically "consumed" / purged
659 when records are no longer needed.
661 Severity : enhancement
663 Description: compact fld format with extents
664 Details : Store range of seq rather than every seq in FLD. Seq
665 controller update FLD rather than clients. In Case of CMD, mdt0
666 has FLD, all other metadata server act as non persistent proxy
667 for FLD queries and cache fld entries in fld cache.
672 Description: don't skip ost target if they assigned to file
673 Details : Drop slow OSCs if we can, but not for requested start idx.
674 This means "if OSC is slow and it is not the requested
675 start OST, then it can be skipped, otherwise skip it only
676 if it is inactive/recovering/out-of-space.
680 Description: more cleanup in mds_lov
681 Details : not send LOV EA under replay, we can't know about they size at this
682 time. Don't allow client connect to mds before any ost connected,
683 for avoid problems with LOV EA size and returning EIO to client.
685 Severity : enhancement
687 Description: Interoperability at server side (Disk interoperability)
689 Severity : enhancement
691 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
693 Severity : enhancement
695 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
697 Severity : enhancement
699 Description: New client IO stack (CLIO).
701 Severity : enhancement
703 Description: Commit on sharing. Eliminate inter-client dependencies between
704 uncommitted transactions by doing transaction commits.
705 Thereby clients may recovery independently.
708 Frequency : Create a symlink file with a very long name
710 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
711 Details : If there is no extra space in the request for early cancels,
712 ldlm_req_handles_avail() returns 0 instead of a negative value.
714 Severity : enhancement
716 Description: Add /proc entry for import status
717 Details : The mdc, osc, and mgc import directories now have
718 an import directory that contains useful import data for debugging
721 Severity : enhancement
723 Description: Re-disable certain /proc logging
724 Details : Enable and disable client's offset_stats, extents_stats and
725 extents_stats_per_process stats logging on the fly.
728 Frequency : Only on FC kernels 2.6.22+
730 Description: oops in statahead
731 Details : Do not drop reference count for the dentry from VFS when lookup,
732 VFS will do that by itself.
734 Severity : enhancement
736 Description: Generic /proc file permissions
737 Details : Set /Proc file permissions in a more generic way to enable non-
738 root users operate on some /proc files.
742 Description: Hitting mdc_commit_close() ASSERTION
743 Details : Properly handle request reference release in
744 ll_release_openhandle().
748 Description: quota recovery deadlock during mds failover
749 Details : This patch includes att18982, att18236, att18237 in bz14840.
751 1. fix osts hang when mds does failover with quotaon
752 2. prevent watchdog storm when osts threads wait for the
757 Frequency : only patchless client
758 Description: add workaround for race between add/remove dentry from hash
760 Severity : enhancement
762 Description: Allow OST glimpses to return PW locks
766 Description: LBUG when llog conf file is full
767 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
771 Description: Prevent import from entering FULL state when server in recovery
775 Description: service mount cannot take device name with ":"
776 Details : Only when device name contains ":/" will mount treat it as
782 Description: replace ptlrpcd with the statahead thread to interpret the async
783 statahead RPC callback
787 Frequency : on recovery
788 Description: I/O failures after umount during fail back
789 Details : if client reconnected to restarted server we need join to recovery
790 instead of find server handler is changed and process self
791 eviction with cancel all locks.
793 Severity : enhancement
795 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
797 Severity : enhancement
799 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
801 Severity : enhancement
803 Description: Upcall on Lustre log has been dumped
804 Details : Allow for a user mode script to be called once a Lustre log has
805 been dumped. It passes the filename of the dumped log to the
806 script, the location of the script can be specified via
807 /proc/sys/lnet/debug_log_upcall.
812 Description: avoid idr_remove called for id which is not allocated.
813 Details : Move assigment s_dev for clustered nfs to end of initialization,
814 to avoid problem with error handling.
819 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
820 Details : When connection is reused this not moved from CONN_UNUSED_HASH
821 into CONN_USED_HASH and this prodice warning when put connection
822 again in unused hash.
825 Severity : enhancement
827 Description: File striping can now be set to use an arbitrary pool of OSTs.
829 Severity : enhancement
831 Description: Export bytes_read/bytes_write count on OSC/OST.
835 Description: Early reply size mismatch, MGC loses connection
836 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
837 the connect flags are properly negotiated.
842 Description: quotas are not honored with O_DIRECT
843 Details : all writes with the flag O_DIRECT will use grants which leads to
844 this problem. Now using OBD_BRW_SYNC to guard this.
848 Description: add quota statistics
849 Details : 1. sort out quota proc entries and proc code.
850 2. add quota statistics
852 Severity : enhancement
854 Description: enable quota support for HEAD.
858 Description: Properly propagate oinfo flags from lov to osc for statfs
859 Details : restore missing copy oi_flags to lov requests.
861 Severity : enhancement
863 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
864 : l_getgroups(8), lst(8), routerstat(8)
865 Details : included man pages for llobdstat(8), llstat(8),
866 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
868 Severity : enhancement
870 Description: configure's --enable-quota should check the
871 : kernel .config for CONFIG_QUOTA
872 Details : configure is terminated if --enable-quota is passed but
873 : no quota support is in kernel
877 Description: Remove portals compatibility
878 Details : Remove portals compatibility, not interoperable with releases
883 Description: Resolve device initialization race
884 Details : Prevent proc handler from accessing devices added to the
885 obd_devs array but yet be intialized.
887 Severity : enhancement
889 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
891 Severity : enhancement
893 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
898 Description: Using wrong pointer in osc_brw_prep_request
899 Details : Access to array[-1] can produce panic if kernel compiled with
900 CONFIG_PAGE_ALLOC enabled
904 Description: Client runs out of low memory
905 Details : Consider only lowmem when counting initial number of llap pages
909 Description: *optional* service tags registration
910 Details : if the "service tags" package is installed on a Lustre node
911 When the filesystem is mounted, a local-node service tag will
912 be created. See http://inventory.sun.com/ for more information
913 about the Service Tags asset management system.
917 Description: Kernel BUG tries to release flock
918 Details : Lustre does not destroy flock lock before last reference goes
919 away. So always drop flock locks when client is evicted and
920 perform unlock regardless of successfulness of speaking to MDS.
924 Description: add refcount protection for osc callbacks, avoid panic on shutdown
928 Description: sanity test 65a fails if stripecount of -1 is set
929 Details : handle -1 striping on filesystem in ll_dirstripe_verify
934 Description: ASSERTION(CheckWriteback(page,cmd)) failed
935 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
938 Severity : enhancement
940 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
944 Description: do not process already freed flock
945 Details : flock can possibly be freed by another thread before it reaches
946 to ldlm_flock_completion_ast.
950 Description: LBUG during stress test
951 Details : Need properly lock accesses the flock deadlock detection list.
955 Description: oops in page fault handler
956 Details : kernel page fault handler can return two special 'pages' in error
957 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
961 Description: timeout with invalidate import.
962 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
963 should be handled by ptlrpcd. This produce long age waiting and
964 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
966 Severity : enhancement
968 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
972 Description: enable MGS and MDT services start separately
973 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
974 a co-located MGS without starting the MGS, which is a complement
975 to 'nosvc' mount option.
979 Frequency : after recovery
980 Description: precreate to many object's after del orphan.
981 Details : del orphan st in oscc last_id == next_id and this triger growing
982 count of precreated objects. Set flag LOW to skip increase count
983 of precreated objects.
987 Frequency : rare, on clear nid stats
988 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
989 Details : when clean nid stats sometimes try destroy live entry,
990 and this produce panic in free.
994 Description: Stack overflow during MDS log replay
995 ease stack pressure by using a thread dealing llog_process.
999 Description: wait until IO finished before start new when do lock cancel.
1000 Details : VM protocol want old IO finished before start new, in this case
1001 need wait until PG_writeback is cleared until check dirty flag and
1002 call writepages in lock cancel callback.
1004 Severity : enhancement
1006 Description: using special macro for print time and cleanup in includes.
1010 Description: mds_mfd_close() ASSERTION(rc == 0)
1011 Details : In mds_mfd_close(), we need protect inode's writecount change
1012 within its orphan write semaphore to prevent possible races.
1016 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1020 Frequency : rare, on shutdown ost
1021 Description: don't hit live lock with umount ost.
1022 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1023 use shrink_dcache_sb instead.
1027 Description: don't panic with use echo client
1028 Details : echo client pass NULL as client nid pointer and this produce null
1029 pointer dereference.
1033 Description: fix build on ppc32
1034 Details : compile code with -m64 flag produce wrong object file for ppc32.
1038 Description: add message levels for liblustreapi
1042 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1043 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1044 will result in an -EAGAIN and in the exit call path we call
1045 alloc_rr() which will with increasing aggressiveness attempt to
1046 aquire precreated objects on the minimum number of required OSCs.
1050 Description: Use old size assignment to avoid deadlock
1051 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1052 the scheduling while holding a spinlock. We do not need locking
1053 for size in ll_update_inode() because size is only updated from
1054 the MDS for directories or files without objects, so there is no
1055 other place to do the update, and concurrent access to such inodes
1056 are protected by the inode lock.
1060 Description: resolve "_IOWR redefined" build error on SLES10
1064 Description: dump the memory debugging after all modules are unloaded to
1065 suppress false negative in conf_sanity test 39
1067 Severity : enhancement
1069 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1074 Description: fix SLES kernel versioning
1075 Details : the kernel version for our SLES 10 kernel did not include a "-"
1076 before the "smp" at the end. while this was not a problem in
1077 general, it did mean that software trying to use the kernel
1078 version to try to detect a vendor specific kernel would fail.
1079 this was most evident by the OFED build scripts.
1083 Description: Don't update lov_desc members until making sure they are valid
1084 Details : When updating lov_desc members via proc fs, need fix their
1085 validities before doing the real update.
1089 Description: don't put request into delay list while invalidate in flight.
1090 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1091 import in flight. this produce timeout for invalidate and sometimes
1092 can cause stale data.
1096 Frequency : on ppc only
1097 Description: not convert ost objects for directory because it's not exist.
1098 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1102 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1104 Severity : enhancement
1106 Description: Update to support 2.6.22.14 vanilla kernel.
1110 Frequency : rare, on recovery
1111 Description: read procfs can produce deadlock in some situation
1112 Details : Holding lprocfs lock which send rpc can produce block for destroy
1113 obd objects and this also block reconnect with -EALREADY. This
1114 isn't fix all lprocfs bugs - but make it rare.
1116 Severity : enhancement
1118 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1121 Frequency : frequent on X2 node
1123 Description: mdc_set_open_replay_data LBUG
1124 Details : Set replay data for requests that are eligible for replay.
1128 Description: lustre_mgs: operation 101 on unconnected MGS
1129 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1130 MGC, and late on MGC cannot successfully connect to MGS and a lot
1131 of the error messages complaining that MGS is not connected.
1135 Frequency : on network error
1136 Description: panic with double free request if network error
1137 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1138 but it's true only for synchronus enqueue, for async enqueue
1139 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1142 Severity : enhancement
1144 Description: client-side metadata stat-ahead during readdir(directory readahead)
1145 Details : perform client-side metadata stat-ahead when the client detects
1146 readdir and sequential stat of dir entries therein
1149 Frequency : on start mds
1151 Description: Implement get_info(last_id) in obdfilter.
1154 Frequency : occasional
1156 Description: Correctly check stale fid, not start epoch if ost not support SOM
1157 Details : open with flag O_CREATE need set old fid in op_fid3 because
1158 op_fid2 was overwritten with new generated fid, but mds can answer
1159 with one of these two fids and both is not stale. Setattr
1160 incorrectly started an epoch and assume will be called
1161 done_writing, but without SOM done_writing ever being called.
1164 Frequency : rare, depends on device drivers and load
1166 Description: MDS or OSS nodes crash due to stack overflow
1167 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1168 In some cases, in conjunction with device drivers that use a lot
1169 of stack the MDS (or possibly OSS) service threads could overflow
1170 the stack. One change which was identified to consume additional
1171 stack has been reworked to avoid the extra stack usage.
1174 Frequency : occasional
1176 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1177 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1178 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1179 oscc_has_objects and see OSC has no objests and call
1180 oscc_internal_create to resend create request.
1182 Severity : enhancement
1184 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1186 Severity : enhancement
1188 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1191 Frequency : very rare
1193 Description: Fix replay if there is an un-replied request and open
1194 Details : In some cases, older replay request will revert the
1195 mcd->mcd_last_xid on MDS which is used to record the client's
1196 latest sent request.
1198 Severity : enhancement
1200 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1202 Severity : enhancement
1204 Description: Add rhel5 support to HEAD.
1206 Serverity : enhancement
1208 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1213 Description: Don't allow skipping OSTs if index has been specified.
1214 Details : Don't allow skipping OSTs if index has been specified, make
1215 locking in internal create lots better.
1219 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1220 Details : access to bitfield in structure is always rounded to long
1221 and this produce problem with not atomic change any bit.
1225 Description: Lustre make rpms failed.
1226 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1227 builds Lustre rpms from tarball.
1230 Frequency : rare on shutdown ost
1232 Description: If llog cancel was not send before clean_exports phase, this can
1233 produce deadlock in llog code.
1234 Details : If llog thread has last reference to obd and call class_import_put
1235 this produce deadlock because llog_cleanup_commit_master wait when
1236 last llog_commit_thread exited, but this never success because was
1237 called from llog_commit_thread.
1241 Description: allow userland application know is lost one of stripes.
1242 Details : fill lvb_blocks with error code on ost and return it to
1243 application if error flag found.
1247 Description: NULL lov_tgts causing MDS oops
1248 Details : more safe checks for NULL lov_tgts for avoid oops.
1250 Severity : enhancement
1252 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1256 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1257 Details : If an OST is down the MDS will hang indefinitely in
1258 obd_statfs_rqset() waiting for the statfs data. While for
1259 MDS QOS usage of statfs, it should not stuck in waiting.
1261 Severity : enhancement
1263 Description: remote_acl support
1264 Details : Support ACL-based permission check for remote user.
1265 Support setfacl/getfacl for remote user with the utils
1266 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1267 the system "{s,g}etfacl" utils.
1269 Severity : enhancement
1271 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1273 Severity : enhancement
1275 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1279 Description: make mgs_setparam() handle fsname containing dash
1280 Details : fsname containing a dash does not work with lctl conf_param
1282 Severity : enhancement
1284 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1286 Severity : enhancement
1288 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1292 Frequency : rare, at shutdown
1293 Description: access already free / zero obd_namespace.
1294 Details : if client_disconnect_export was called without force flag set,
1295 and exist connect request in flight, this can produce access to
1296 NULL pointer (or already free pointer) when connect_interpret
1297 store ocd flags in obd_namespace.
1301 Frequency : only at startup
1302 Description: not alloc memory with spinlock held.
1303 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1304 if any spinlock held.
1306 Severity : enhancement
1308 Description: make lustre randomly fail allocating memory
1309 Details : Make lustre randomly failed allocating memory for testing purpose.
1311 Severity : enhancement
1313 Description: lost problems with lov objid file
1314 Details : Fixes some scability and access to not inited memory problems
1315 in work with lov objdid file.
1320 Description: lfs find does not continue on file error
1321 Details : Continue other files processing when a file/dir is absent.
1325 Description: Inconsistent usage of lustre_pack_reply()
1326 Details : Standardize the usage of lustre_pack_reply() such that it
1327 always generate a CERROR on failure.
1330 Frequency : occasional
1332 Description: MDS hang or stay in waiting lock
1333 Details : If client receive lock with CBPENDING flag ldlm need send lock
1334 cancel as separate rpc, to avoid situation when cancel request
1335 can't processed due all i/o threads stay in wait lock.
1339 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1343 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1347 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1351 Description: Update to RHEL4 latest kernel.
1353 Severity : enhancement
1355 Description: Build SLES10 patchless client fails
1356 Details : The configure was broken by run ./configure with
1357 --with-linux-obj=.... argument for patchless client. When the
1358 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1359 can't search header adequately. Use absolute path such as
1360 -I($LINUX)/include instead.
1364 Description: interrupt oig_wait produce painc on resend.
1365 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1366 private set, and this produce situation when rq_ptlrpcd_data not
1367 copyed to new allocated request and triggered LBUG on assert
1368 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1369 ptlrpcd set if request is changed and can be safety changed to use
1374 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1375 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1376 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1377 these macros and replace them with their difinitions instead.
1381 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1382 Details : Update which_patch & target file for SLES10 latest kernel.
1385 Description: Remove client patches from SLES 10 kernel.
1386 Details : This causes SLES 10 clients to behave as patchless clients
1387 even on a Lustre-patched (server) kernel.
1389 Severity : enhancement
1391 Description: self-adjustable client's lru lists
1392 Details : use adaptive algorithm for managing client cached locks lru
1393 lists according to current server load, other client's work
1394 pattern, memory activities, etc. Both, server and client
1395 side namespaces provide number of proc tunables for controlling
1398 Severity : enhancement
1400 Description: light-weight GSS support
1401 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1402 and reduce performance overhead.
1404 Severity : enhancement
1406 Description: Linux keyring support
1407 Details : Support using service of Linux keyring for Lustre GSS internal
1408 context refresh/cache mechanism.
1412 Description: Fix errors in lfs documentation
1413 Details : Fixes man pages
1417 Description: don't use GFP_* in generic Lustre code.
1418 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1422 Description: obdclass is limited by single OBD_ALLOC(idarray)
1423 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1427 Description: warnings with build patchless client with vanila 2.6.19 and up
1428 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1429 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1433 Frequency : Only for SLES
1434 Description: sanity_quota fail test_1
1435 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1436 file, which makes TSTID[2] inunique.
1440 Frequency : Only for SLES9
1441 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1442 Details : Update target/ChangeLog/which_patch .
1446 Frequency : Only for test_10 in sanity-quota.sh
1447 Description: a bug in quota test code
1448 Details : checking a test flag in a wrong place
1452 Frequency : Only for test_13 in sanity-quota.sh
1453 Description: a bug in quota test script
1454 Details : shouldn't check overfull allocated quota
1458 Description: jbd statistics
1459 Details : Port older jbd statistics patch for sles10
1461 Severity : enhancement
1463 Description: Add printing inode info into message about error in writepage.
1468 Description: reply_lock_interpret crash due to race with it and lock cancel.
1469 Details : Do not replay locks that are being cancelled. Do not reference
1470 locks by their address during replay, just by their handle.
1474 Frequency : When flocks are used.
1475 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1476 Details : Flock locks might destroy just granted lock if it could be merged
1477 with another existing flock, this is done in completion handler,
1478 so teach ldlm_cli_enquque_fini that this is a valid case for
1484 Description: Oops in read and write path when failing to allocate lock.
1485 Details : Check if lock allocation failed and return error back.
1489 Description: lstripe command fails for valid OST index
1490 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1491 instead of lov->desc.ld_active_tgt_count.
1495 Description: sanity.sh failed test 103
1496 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1501 Description: df doesn't work properly if diskfs blocksize != 4K
1502 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1506 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1507 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1509 Severity : enhancement
1511 Description: Adaptive timeouts
1512 Details : RPC timeouts adapt to changing server load and network
1513 conditions to reduce resend attempts and improve recovery time.
1517 Description: llapi_file_create() does not allow some changes
1518 Details : add llapi_file_open() that allows specifying the mode and
1519 open flags, and also returns an open file handle.
1523 Description: merge and cleanup kernel patches.
1524 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1528 Description: Add journal checksum support.(Kernel part)
1529 Details : The journal checksum feature adds two new flags i.e
1530 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1531 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1532 indicates that the commit block contains the checksum for
1533 the blocks described by the descriptor blocks. Now commit
1534 record can be sent to disk without waiting for descriptor
1535 blocks to be written to disk. This behavior is controlled
1536 using JBD2_FEATURE_ASYNC_COMMIT flag.
1540 Description: OSS needs mutliple precreate threads
1541 Details : Add ability to start more than one create thread per OSS.
1545 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1546 Details : execute lfs setstripe on client
1550 Description: mds_obd_create error creating tmp object
1551 Details : When the user sets quota on root, llog will be affected and can't
1552 create files and write files.
1555 Frequency : Always on ia64 patchless client, and possibly others.
1557 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1558 Details : This allows the patchless client to be loaded on architectures
1559 without this export.
1563 Description: RedHat Update kernel for RHEL5
1564 Details : Add the kernel config file for RHEL5.
1568 Description: RedHat Update kernel for RHEL5
1569 Details : Modify the kernel config file more closer RHEL5.
1573 Description: Build failure against Centos5 (RHEL5)
1574 Details : Define PAGE_SIZE when it isn't present.
1578 Description: test_7 on sanity-quota.sh can't be used on separate machines
1579 Details : add facet to handle it
1583 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1584 Details : with NFS, the anon dentry's parent was set to itself in
1585 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1586 corresponding dentry other than use rec->ur_name.
1588 Severity : enhancement
1590 Description: lfs setstripe enhancement
1591 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1595 Description: enable data checksumming by default
1596 Details : enable checksum by default, allow --disable-checksum
1597 configure option and "-o nochecksum" mount option. Checksums
1598 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1602 Description: lustre support for RHEL5
1603 Details : Add support for RHEL5.
1607 Description: Client eviction due to failover config
1608 Details : after a connection loss, the lustre client should attempt to
1609 reconnect to the last active server first before trying the
1610 other potential connections.
1614 Description: when mds and osts use different quota unit(32bit and 64bit),
1615 quota will be released repeatly.
1616 Details : void sending multiple quota reqs to mds, which will keep the
1617 status between the reqs.
1621 Description: osts not allocated evenly to files
1622 Details : change the condition to increase offset_idx
1625 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1627 Description: Data corruption for OSTs that are formatted larger than 2TB
1629 Details : When generating the bio request for lustre file writes the
1630 sector number would overflow a temporary variable before being
1631 used for the IO. The data reads correctly from Lustre (which
1632 will overflow in a similar manner) but other file data or
1633 filesystem metadata may be corrupted in some cases.
1637 Description: Tune the kernel for good SCSI performance.
1638 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1639 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1644 Description: rewrite ext2-derived code in obdclass/uuid.c
1645 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1646 from scratch preserving functionality.
1650 Description: "lfs find -obd UUID" prints directories
1651 Details : "lfs find -obd UUID" will return all directory names instead
1652 of just file names. It is incorrect because the directories
1653 do not reside on the OSTs.
1657 Description: use i_size_read and i_size_write in 2.6 port
1658 Details : replace inode->i_size access with i_size_read/write()
1661 Frequency : when removing large files
1663 Description: scheduling issue during removal of large Lustre files
1664 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1665 It causes scheduling issues when removing large files (17TB in the
1669 Frequency : only with liblustre clients on XT3
1671 Description: evictions taking too long
1672 Details : allow llrd to evict clients directly on OSTs
1675 Frequency : only on ppc
1677 Description: /proc/fs/lustre/devices broken on ppc
1678 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1679 We should make sure the type of 'index' is loff_t and then cast
1680 explicitly as needed below. Do not assign an explicitly cast
1684 Frequency : only for rhel5
1686 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1687 Details : Modify the target file & which_kernel.
1689 Severity : enhancement
1691 Description: omit set fsid for export NFS
1692 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1693 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1697 Frequency : Always, for kernels after 2.6.16
1698 Description: Fix warning idr_remove called for id=.. which is not allocated.
1699 Details : Last kernels save old s_dev before kill super and not allow
1700 to restore from callback - restore it before call kill_anon_super.
1704 Description: buffer overruns could theoretically occur
1705 Details : llapi_semantic_traverse() modifies the "path" argument by
1706 appending values to the end of the origin string, and a
1707 overrun may occur. Adding buffer overrun check in liblustreapi.
1711 Description: Fix error on 'ls .' at the top of the Lustre mount.
1712 Details : Don't revalidate dentry if it is a root dentry.
1716 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1717 Details : Modify vm-tunables-rhel4.patch.
1721 Description: Kernel config for 2.6.18-vanilla.
1722 Details : Modify targets/2.6-vanilla.target.in.
1723 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1724 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1725 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1726 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1730 Description: improve handling recoverable errors
1731 Details : if request processig with error which can be recoverable on server
1732 request should be resend, otherwise page released from cache and
1737 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1738 Details : Modify target file & which_patch.
1739 A flaw was found in the IA32 system call emulation provided
1740 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1741 value could be stored in the %RAX register, which could trigger an
1742 out-of-bounds system call table access. An untrusted local user
1743 could exploit this flaw to run code in the kernel
1744 (ie a root privilege escalation). (CVE-2007-4573).
1748 Description: O_DIRECT bypasses client statistics.
1749 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1750 read_ahead_stats were updated. copied the stats section from
1751 osc_send_oap_rpc() into async_internal().
1755 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1759 Description: change order of libsysio includes
1760 Details : '#include sysio.h' should always come before '#include xtio.h'
1764 Description: handle "serious error: objid * already exists" more gracefully
1765 Details : If LAST_ID value on disk is smaller than the objects existing in
1766 the O/0/d* directories, it indicates disk corruption and causes an
1767 LBUG(). If the object is 0-length, then we should use the existing
1768 object. This will help to avoid a full fsck in most cases.
1770 Severity : enhancement
1772 Description: adapt the lustre_config script to support the upgrade case
1773 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1774 server targets to 1.6 in parallel.
1778 Description: To avoid grant space > avaible space when the disk is almost
1779 full. Without this patch you might see the error "grant XXXX >
1780 available" or some LBUG about grant, when the disk is almost
1782 Details : In filter_check_grant, for non_grant cache write, we should
1783 check the left space by if (*left > ungranted + bytes), instead
1784 of (*left > ungranted), because only we are sure the left space
1785 is enough for another "bytes", then the ungrant space should be
1786 increase. In client, we should update cl_avail_grant only there
1787 is OBD_MD_FLGRANT in the reply.
1791 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1792 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1793 described in CVE-2007-4573. This problem could allow untrusted
1794 local users to gain root access.
1797 Frequency : when using O_DIRECT and quotas
1799 Description: Incorrect file ownership on O_DIRECT output files
1800 Details : block usage reported by 'lfs quota' does not take into account
1801 files that have been written with O_DIRECT.
1806 Description: touch file failed when fs is not full
1807 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1808 otherwise we can get ENOSP while fs is not full.
1812 Description: parallel lock callbacks
1813 Details : Instead of sending blocking and completion callbacks as separated
1814 requests, adding them to a set and sending in parallel.
1817 Frequency : only for Cray XT3
1818 Bugzilla : 12829/13455
1819 Description: Changing primary group doesn't change the group lustre assigns to
1821 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1822 overridden with the primary group provided by the group upcall,
1823 whereas the supplied fsgid can be trusted if it is in the list of
1824 supplementary groups returned by the group upcall.
1826 Severity : enhancement
1828 Description: Allow masking D_WARNING, D_ERROR messages from console
1829 Details : Console messages can now be disabled via lnet.printk.
1833 Description: User code with malformed file open parameter crashes client node
1834 Details : Before packing join_file req, all the related reference should be
1835 checked carefully in case some malformed flags cause fake
1836 join_file req on client.
1840 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1841 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1842 mode is equal to the granted mode. However, it should also check
1843 that LDLM_FL_CP_REQD is not set.
1847 Description: Heavy nfs access might result in deadlocks
1848 Details : After ELC code landed, it is now improper to enqueue any mds
1849 locks under och_sem, because enqueue might want to decide to
1850 cancel open locks for same inode we are holding och_sem for.
1854 Description: Client eviction while running blogbench
1855 Details : A lot of unlink operations with concurrent I/O can lead to a
1856 deadlock causing evictions. To address the problem, the number of
1857 oustanding OST_DESTROY requests is now throttled to
1858 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1859 callbacks are processed in priority.
1863 Description: enable ACLs on MDS by default
1864 Details : ACLs must be enabled on MDS by default.
1867 Frequency : PPC/PPC64 only
1869 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1870 Details : fix duplicated definitions between asm-ppc64/types.h and
1871 lustre_types.h on PPC.
1874 Frequency : PPC/PPC64 only
1876 Description: asm-ppc/segment.h does not exist
1877 Details : fix compile issue on PPC.
1881 Description: better handle error messages in extents code
1884 Frequency : RHEL4 only
1886 Description: mkfs is very slow on IA64/RHEL4
1887 Details : A performance regression has been discovered in the MPT Fusion
1888 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1889 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1890 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1892 Severity : enhancement
1894 Description: SNMP support enhancement
1895 Details : Adding total number of sampled request for an MDS node in snmp
1898 Severity : enhancement
1900 Description: Optimize ldlm waiting list processing for PR extent locks
1901 Details : When processing waiting list for read extent lock and meeting read
1902 lock that is same or wider to it that is not contended, skip
1903 processing rest of the list and immediatelly return current
1904 status of conflictness, since we are guaranteed there are no
1905 conflicting locks in the rest of the list.
1909 Description: Time out and refuse to reconnect
1910 Details : When the failover node is the primary node, it is possible
1911 to have two identical connections in imp_conn_list. We must
1912 compare not conn's pointers but NIDs, otherwise we can defeat
1913 connection throttling.
1917 Description: port llog fixes from b1_6 into HEAD
1918 Details : Port llog reference couting and some llog cleanups from b1_6
1919 (bug 10800) into HEAD, for protect from panic and access to already
1920 free llog structures.
1924 Description: Detect stride IO mode in read-ahead
1925 Details : When a client does stride read, read-ahead should detect that and
1926 read-ahead pages according to the detected stride pattern.
1930 Description: data checksumming impacts single node performance
1931 Details : add support for several checksum algorithm. Currently, only CRC32
1932 and Adler-32 are supported. The checksum type can be changed on
1933 the fly via /proc/fs/lustre/osc/*/checksum_type.
1937 Description: use adler32 for page checksums
1938 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1943 Description: build for x2 fails
1944 Details : fix compile issue on Cray systems.
1948 Description: Properly match for duplicate locks
1949 Details : Due to different lock order from skiplists code, we need to
1950 traverse entire list for now
1953 Frequency : only on PPC/SLES10
1955 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1956 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1957 be defined. Add a hack in mkfs_lustre.c to work around this compile
1962 Description: LASSERT on MDS when client holding flock lock dies
1963 Details : ldlm pool logic depends on number of granted locks equal to
1964 number of released locks which is not true for flock locks, so
1965 just exclude such locks from consideration.
1969 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1970 Details : Use fsfilt_sync() for both the whole filesystem sync and
1971 individual file sync to eliminate dangerous inode locking
1972 with I_LOCK that can lead to a deadlock.
1976 Description: performance in 1.6.3
1977 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1978 sure that 1MB requests are not fragmented by the block layer.
1980 Severity : enhancement
1982 Description: organize the server-side client stats on per-nid basis
1983 Details : Change the structure of stats under obdfilter and mds to
1991 The "uuid"s file would list the uuids of _active_ exports.
1992 And the clear entry is to clear all stats and stale nids.
1994 Severity : enhancement
1996 Description: eliminate client locks in face of contention
1997 Details : file contention detection and lockless i/o implementation
1998 for contended files.
2002 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2005 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2007 Description: MDS may lose file striping (and hence file data) in some cases
2008 Details : If there are additional extended attributes stored on the MDS,
2009 in particular ACLs, SELinux, or user attributes (if user_xattr
2010 is specified for the client mount options) then there is a risk
2011 of attribute loss. Additionally, the Lustre file striping
2012 needs to be larger than default (e.g. striped over all OSTs),
2013 and an additional attribute must be stored initially in the
2014 inode and then increase in size enough to be moved to the
2015 external attribute block (e.g. ACL growing in size) for file
2020 Description: skiplist implementation simplification
2021 Details : skiplists are used to group compatible locks on granted list
2022 that was implemented as tracking first and last lock of each
2023 lock group the patch changes that to using doubly linked lists
2027 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2028 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2029 because LASSERT on both the data supplied by a client, and the
2030 data on disk is dangerous and incorrect.
2032 Severity : enhancement
2034 Description: Slow truncate/writes to huge files at high offsets.
2035 Details : Directly associate cached pages to lock that protect those pages,
2036 this allows us to quickly find what pages to write and remove
2037 once lock callback is received.
2041 Description: more ldlm soft lockups
2042 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2043 starve other threads from the resource lock for a long time in
2044 case of long waiting queue, so change the debug level from
2045 D_OTHER to the less frequently used D_INFO.
2047 Severity : enhancement
2049 Description: add -gid, -group, -uid, -user options to lfs find
2053 Description: Hung threads in invalidate_inode_pages2_range
2054 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2055 once one is completed. As a result, some RPCs are stuck in the
2056 queue and are never sent.
2060 Description: filter threads hungs on waiting journal commit
2061 Details : Cleanup filter group llog code, then only filter group llog will
2062 be only created in the MDS/OST syncing process.
2066 Description: Procfs and llog threads access destoryed import sometimes.
2067 Details : Sync the import destoryed process with procfs and llog threads by
2068 the import refcount and semaphore.
2070 Severity : enhancement
2072 Description: openlock cache of b1_6 port to HEAD
2077 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2078 Details : If insertion of an extent fails, then discard the inode
2079 preallocation and free data blocks else it can lead to duplicate
2084 Description: don't always update ctime in ext3_xattr_set_handle()
2085 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2086 In some cases the ctime should not be updated, for example for
2087 2.0->1.8 compatibility it is necessary to delete an xattr and it
2088 should not update the ctime.
2092 Bugzilla : 15713/16362
2093 Description: Assertion in iopen_connect_dentry in 1.6.3
2094 Details : looking up an inode via iopen with the wrong generation number can
2095 populate the dcache with a disconneced dentry while the inode
2096 number is in the process of being reallocated. This causes an
2097 assertion failure in iopen since the inode's dentry list contains
2098 both a connected and disconnected dentry.
2102 Description: assertion failure in ldlm_handle2lock()
2103 Details : fix a race between class_handle_unhash() and class_handle2object()
2104 introduced in lustre 1.6.5 by bug 13622.
2109 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2110 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2115 Description: oops in sd_iostats_seq_show()
2116 Details : unloading/reloading the scsi low level driver triggers a kernel
2117 bug when trying to access the sd iostat file.
2122 Description: Kernel panics during QLogic driver reload
2123 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2124 patch, causing memory corruption.
2129 Description: journal_dev option does not work in b1_6
2130 Details : pass mount option during pre-mount.
2132 Severity : enhancement
2134 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2135 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2136 extent information of a file. It can be used to map logical blocks
2137 in a file to physical blocks in the block device.
2141 Description: LDLM soft lockups - improvement
2142 Details : It is be possible to send the lock handle along with each read
2143 or write request because the client is already doing a lock match
2144 itself so there isn't any reason the OST should have to re-do that
2148 Frequency : only on Cray X2
2150 Description: X2 build failures
2151 Details : fix build failures on Cray X2.
2155 Description: xid & resent requests
2156 Details : Initialize RPC XID from clock at startup (randomly if clock is
2159 Severity : enhancement
2161 Description: Add lustre_start utility to start or stop multiple Lustre servers
2166 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2167 Details : In case of memory pressure, list_del() can be called twice on
2168 req->rq_history_list, causing a kernel oops.
2172 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2173 Details : fix stack overflow in the distributed lock manager by defering
2174 export eviction after a failed AST to the elt thread instead of
2175 handling it in the dlm interpret routine.
2179 Description: Convert some comments to new format.
2180 Details : Update documenting comments to match doxygen conventions.
2184 Description: Grammar fixes.
2185 Details : A couple of trivial spelling fixes.
2189 Description: OSD_COUNTERS-mandatory
2190 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2191 to simplify the code.
2195 Description: simplify cmm_device freeing logic.
2196 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2200 Description: Add lockdep support to dt_object_operations locking interface.
2201 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2202 parameter indicating lock ordering. Update mdd code to use new
2207 Description: Introduce failloc constants for lockless IO tests.
2208 Details : Add two new failloc constants to test lockless IO. Only one of
2209 them in implemented---another is checked in yet to be landed
2214 Description: Add lockdep support for inode mutex.
2215 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2216 in the situations where Lustre has to lock more than one inode
2221 Description: Add optional invariants checking support.
2222 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2223 configure switch. This macro is to be used for consistency and
2224 sanity checks that are too expensive to be left in `production'
2229 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2230 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2235 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2236 Details : Use spin_lock_nested() in (the only) situation where more than
2237 one ldlm_lock is locked simultaneously. Also, fix possible
2238 dead-lock in ldlm_lock_change_resource() by enforcing particular
2243 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2244 Details : Instead of specifying each ldlm_lock call-back through separate
2245 parameter, wrap them into struct ldlm_callback_suite.
2249 Description: Kill join_lru obd method and its callers.
2250 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2251 in memory---a requirement implemented through ->o_join_lru() obd
2252 method in HEAD. Remove this method and its users.
2256 Description: Add asynchronous ldlm ENQUEUE completion handler.
2257 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2258 case that stock ldlm_completion_ast() cannot handle as it waits
2259 until lock is granted. Introduce new ldlm_completion_ast_async()
2260 for this. Also comment ldlm_completion_ast().
2264 Description: ldlm_error <-> errno conversion.
2265 Details : Add functions to map (rather arbitrary) between LDLM error codes
2266 and standard errno values. CLIO needs this to prevent LDLM specific
2267 constants from escaping ldlm and osc.
2271 Description: Kill unused ldlm_handle2lock_ns() function.
2272 Details : Kill unused ldlm_handle2lock_ns() function.
2276 Description: Add lu_ref support to ldlm_lock
2277 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2278 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2279 and ldlm_resource. LDLM interface has to be changed, because code that
2280 releases a reference on a lock, has to "know" what reference this is.
2281 In the most frequent case
2283 lock = ldlm_handle2lock(handle);
2285 LDLM_LOCK_PUT(lock);
2287 no changes are required. When any other reference (received _not_ from
2288 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2289 instead of LDLM_LOCK_PUT().
2291 Arguably, changes are pervasive, and interface requires some discipline
2292 for proper use. On the other hand, it was very instrumental in finding
2293 a few leaked lock references.
2297 Description: Add ldlm_lock_addref_try().
2298 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2299 attempts to addref a lock that might be being canceled
2304 Description: Add ldlm_weigh_callback().
2305 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2306 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2307 hard-coded `number of pages' logic.
2311 Description: Add lockdep annotations to llog code.
2312 Details : Use appropriately tagged _nested() locking calls in the places
2313 where llog takes more than one ->lgh_lock lock.
2317 Description: Add loi_kms_set().
2318 Details : Wrap kms updates into a helper function.
2322 Description: Constify instances of struct lsm_operations.
2323 Details : Constify instances of struct lsm_operations.
2327 Description: lu_conf support.
2328 Details : On a server, a file system object is uniquely identified
2329 by a fid, which is sufficient to locate and load all object
2330 state (inode). On a client, on the other hand, more data are
2331 necessary instantiate an object. Change lu_object_find() and
2332 friends to take additional `lu_conf' argument describing object.
2333 Typically this includes layout information.
2337 Description: lu_context fixes.
2338 Details : Introduce new lu_context functions that are needed on the client
2339 side, where some system threads (ptlrpcd) are shared by multiple
2340 modules, and so cannot be stopped during module shutdown.
2344 Description: Add start and stop methods to lu_device_type_operations.
2345 Details : Introduce two new methods in lu_device_type_operations, that are
2346 invoked when first instance of a given type is created and last one
2347 is destroyed respectively. This is need by CLIO.
2351 Description: Add lu_ref support to struct lu_device.
2352 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2353 track leaked references.
2357 Description: Introduce lu_kmem_descr.
2358 Details : lu_kmem_descr and its companion interface allow to create
2359 and destroy a number of kmem caches at once.
2363 Description: Fix lu_object finalization race.
2364 Details : Fix a race between lu_object_find() finding an object and its
2365 concurrent finalization. This race is (most likely) not possible
2366 on the server, but might happen on the client.
2370 Description: Introduce lu_ref interface.
2371 Details : lu_ref is a debugging module allowing to track references to
2372 a given object. It is quite cpu expensive, and has to be
2373 explicitly enabled with --enable-lu_ref. See usage description
2378 Description: Factor lu_site procfs stats into a separate function.
2379 Details : Separate lu_site stats printing code into a separate function
2380 that can be reused on a client.
2384 Description: Constify instances of struct {lu,dt,md}_device_operations.
2385 Details : Constify instances of struct {lu,dt,md}_device_operations.
2389 Description: Introduce struct md_site and move meta-data specific parts of
2390 struct lu_site here.
2391 Details : Move md-specific fields out of struct lu_site into special struct
2392 md_site, so that lu_site can be used on a client.
2396 Description: Kill mdd_lov_destroy().
2397 Details : Remove unused mdd code.
2401 Description: Add st_block checking to multistat.c.
2402 Details : Add st_block checking to multistat.c.
2406 Description: Add lu_ref support to struct obd_device.
2407 Details : Add lu_ref tracking to obd_device.
2411 Description: Kill obd_set_fail_loc().
2412 Details : Remove unused code.
2416 Description: Add special type for ptlrpc_request interpret functions.
2417 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2418 there. Actual usage will be in CLIO.
2422 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2423 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2424 doesn't work with lockdep.
2428 Description: Add rwv.c test program.
2429 Details : New testing program exercising readv(2) and writev(2) (Qian).
2433 Description: Add sendfile.c test program.
2434 Details : New testing program exercising sendfile(2) (Jay).
2438 Description: Ratelimit a message that can be very frequent.
2439 Details : Ratelimit a memory allocation failure message that can
2444 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2445 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2449 Description: Liblustre build fixes.
2450 Details : Liblustre build fixes.
2454 Description: libcfs: add cfs_{need,cond}_resched() interface.
2455 Details : libcfs: add cfs_{need,cond}_resched() definition and
2456 implementations for Linux, NT, and liblustre.
2458 Severity : enhancement
2460 Description: More exported tunables for mballoc
2461 Details : Add support for tunable preallocation window and new tunables for
2462 large/small requests
2466 Description: Detect corruption of block bitmap and checking for preallocations
2467 Details : Checks validity of on-disk block bitmap. Also it does better
2468 checking of number of applied preallocations. When corruption is
2469 found, it turns filesystem readonly to prevent further corruptions.
2473 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2474 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2475 window to 0. If it does hit cache miss, and read-pattern is still
2476 stride-io mode, does not reset the stride window, but also does not
2477 increase the stride window length in this case.
2481 Frequency : only for big-endian servers
2482 Description: Check if system is big-endian while mounting fs with extents feature
2483 Details : Mounting a filesystem with extents feature will fail on big-endian
2484 systems since ext3-based ldiskfs is not supported on big-endian
2485 systems. This can be over-riden with "bigendian_extents" mount option.
2487 Severity : enhancement
2489 Description: The root squash functionality
2490 Details : A security feature, which is to prevent users from being able
2491 to mount lustre on their desktop, run as root, and delete
2492 all of the files in the filesystem. The goal is accomplished by
2493 remapping user id (UID) and group id (GID) of the root user to
2494 a UID and GID specified by the system administartor via Lustre
2495 configuration management server (MGS). The functionality also
2496 allows to specify sets of clients for which the remapping does
2501 Description: Excessive recovery window
2502 Details : With AT enabled, the recovery window can be excessively long (6000+
2503 seconds). To address this problem, we no longer use
2504 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2505 timeout no longer depends on the service time, it is set to
2506 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2507 time via pb_service_time.
2511 Description: Watchdog triggered on MDS failover
2512 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2513 the OSTs know that the MDS "UUID" can be reused for the same export
2514 from a different NID, so we do not need to wait for the export to be
2518 Frequency : rare, only if using MMP with Linux RAID
2520 Description: MMP doesn't work with Linux RAID
2521 Details : While using HA for Lustre servers with Linux RAID, it is possible
2522 that MMP will not detect multiple mounts. To make this work we
2523 need to unplug the device queue in RAID when the MMP block is being
2524 written. Also while reading the MMP block, we should read it from
2525 disk and not the cached one.
2527 Severity : enhancement
2529 Description: open file using fid
2530 Details : A file can be opened using just its fid, like
2531 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2534 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2537 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2538 Details : add proper locking for creation of jbd_4k slab cache
2542 Description: MMP check in ext3_remount() fails without displaying any error
2543 Details : When multiple mount protection fails during remount, proper error
2546 Severity : enhancement
2548 Description: Allow stripe size to be up to 4G-64k
2549 Details : Fix math logic to allow large stripe sizes.
2553 Description: add check for >8TB ldiskfs filesystems
2554 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2555 Don't allow >8TB ldiskfs filesystems to be mounted without
2556 force_over_8tb mount option
2561 Descriptoin: OSS complains "too many blocks" and crashes
2562 Details : In filter_commitrw_write():
2563 1. lquota_chkquota();
2564 2. obdo_from_inode();
2565 3. lquota_pending_commit();
2566 "2." has changed uid and gid in this case, then caused
2567 lquota_chkquota() and lquota_pending_commit() to use
2568 different uid and gid. That is the root of the bug.
2572 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2573 Details : Detect and print out the 'stripe count = -1' case properly.
2577 Description: Out or order replies might be lost on replay
2578 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2579 request with tid smaller then one currently being added, add it
2580 to the start, not end of the list.
2582 --------------------------------------------------------------------------------
2584 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2586 * Support for kernels:
2587 2.6.5-7.283 (SLES 9),
2588 2.6.9-55.EL (RHEL 4),
2589 2.6.16.46-0.14 (SLES 10),
2590 2.6.18.8 vanilla (kernel.org)
2591 * Client support for unpatched kernels:
2592 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2593 2.6.16 - 2.6.22 vanilla (kernel.org)
2594 * Due to recently discovered recovery problems, we do not recommend
2595 using patchless RHEL 4 clients with this or any earlier release.
2596 * Recommended e2fsprogs version: 1.39.cfs8
2597 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2598 * Starting with this release, the ldiskfs backing filesystem required
2599 by Lustre is now in its own package, lustre-ldiskfs. This package
2600 should be installed. It is versioned separately from Lustre and
2601 may be released separately in future.
2605 Description: block reactivating mgc import until all deactivates complete
2606 Details : Fix race when failing back MDT/MGS to itself (testing)
2609 Frequency : at statup only
2611 Description: mds_lov_synchronize race leads to various problems
2612 Details : simultaneous MDT->OST connections at startup can cause the
2613 sync to abort, leaving the OSC in a bad state.
2615 Severity : enhancement
2617 Description: add optional extra BUILD_VERSION info
2618 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2619 to override the lustre version.
2622 Frequency : 2.6.18 servers only
2624 Description: ll_kern_mount() doesn't release the module reference
2625 Details : The ldiskfs module reference count never drops down to 0
2626 because ll_kern_mount() doesn't release the module reference.
2631 Description: server LBUG when using old ost_num_threads parameter
2632 Details : Accept the old ost_num_threads parameter but warn that it
2633 is deprecated, and fix an off-by-one error that caused an LBUG.
2638 Description: Transient SCSI error results in persistent IO issue
2639 Details : iobuf->dr_error is not reinitialized to 0 between two
2643 Frequency : sometimes when underlying device returns I/O errors
2645 Description: OSTs not going read-only during write failures
2646 Details : OSTs are not remounted read-only when the journal commit threads
2647 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2648 instead of the ext3 wrappers.
2652 Description: poor connect scaling with increasing client count
2653 Details : Don't run filter_grant_sanity_check for more than 100 exports
2654 to improve scaling for large numbers of clients.
2657 Frequency : SLES10 only
2659 Description: sanity-quota.sh quotacheck failed: rc = -22
2660 Details : Quotas cannot be enabled on SLES10.
2663 Frequency : liblustre clients only
2665 Description: getdirentries does not give error when run on compute nodes
2666 Details : getdirentries does not fail when the size specified as an argument
2667 is too small to contain at least one entry
2669 Severity : enhancement
2671 Description: Add LNET router traceability for debug purposes
2672 Details : If a checksum failure occurs with a router as part of the
2673 IO path, the NID of the last router that forwarded the bulk data
2674 is printed so it can be identified.
2679 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2680 Details : Due to a race condition, liblustre clients were occasionally
2681 evicted incorrectly.
2683 Severity : enhancement
2685 Description: lfs setstripe use optional parameters instead of postional
2688 Severity : enhancement
2690 Description: Nanosecond timestamp support for ldiskfs
2691 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2692 resolution timestamps. There is not yet support for this at
2693 the Lustre filesystem level.
2696 Frequency : during server recovery
2698 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2699 Details : request with rq_no_resend flag not awake l_wait_event if they get
2703 Frequency : nfs export on patchless client
2705 Description: connectathon hang when test nfs export over patchless client
2706 Details : Disconnected dentry cannot be found with lookup, so we do not need
2707 to unhash it or make it invalid
2710 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2711 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2712 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2717 Description: open req refcounting wrong on reconnect
2718 Details : If reconnect happened between getting open reply from server and
2719 call to mdc_set_replay_data in ll_file_open, we will schedule
2720 replay for unreferenced request that we are about to free.
2721 Subsequent close will crash in variety of ways.
2722 Check that request is still eligible for replay in
2723 mdc_set_replay_data().
2728 Description: disable writes to filesystem when reading health_check file
2729 Details : the default for reading the health_check proc file has changed
2730 to NOT do a journal transaction and write to disk, because this
2731 can cause reads of the /proc file to hang and block HA state
2732 checking on a healthy but otherwise heavily loaded system. It
2733 is possible to return to the previous behaviour during configure
2734 with --enable-health-write.
2736 Severity : enhancement
2738 Description: 64-bit inode version
2739 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2740 to the inode. This will be required for version-based recovery.
2745 Description: MDS fails to start if a duplicate client export is detected
2746 Details : in some rare cases it was possible for a client to connect to
2747 an MDS multiple times. Upon recovery the MDS would detect this
2748 and fail during startup. Handle this more gracefully.
2750 Severity : enhancement
2752 Description: Add -o localflock option to simulate old noflock
2754 Details : This will achieve local-only flock/fcntl locks
2760 Description: log_commit_thread vs filter_destroy race leads to crash
2761 Details : Take import reference before releasing llog record semaphore
2766 Description: Wrong request locking in request set processing
2767 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2768 imp_delayed_list, in this place should be used imp_lock.
2771 Frequency : when reconnection
2773 Description: Grant Leak when osc reconnect to OST
2774 Details : When osc reconnect ost, OST(filter) should check whether it
2775 should grant more space to client by comparing fed_grant and
2776 cl_avail_grant, and return the granted space to client instead
2777 of "new granted" space, because client will call osc_init_grant
2778 to update the client grant space info.
2781 Frequency : when client reconnect to OST
2783 Description: Grant Leak when osc do resend and replay bulk write
2784 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2785 bulk write request, because the grant info while be sync between
2786 OSC and OST when reconnect, and we should ignore the grant info
2787 these of resend/replay write req.
2792 Description: Grant space more than avaiable left space sometimes.
2793 Details : When then OST is about to be full, if two bulk writing from
2794 different clients came to OST. Accord the avaliable space of the
2795 OST, the first req should be permitted, and the second one
2796 should be denied by ENOSPC. But if the seconde arrived before
2797 the first one is commited. The OST might wrongly permit second
2798 writing, which will cause grant space > avaiable space.
2801 Frequency : when client is evicted
2803 Description: Grant might be wrongly erased when osc is evicted by OST
2804 Details : when the import is evicted by server, it will fork another
2805 thread ptlrpc_invalidate_import_thread to invalidate the
2806 import, where the grant will be set to 0. While the original
2807 thread will update the grant it got when connecting. So if
2808 the former happened latter, the grant will be wrongly errased
2809 because of this race.
2814 Description: Checking Stale with correct fid
2815 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2816 to check whether it is stale, because sometimes, we want the
2817 enqueue happened anyway, and op_data.fid2 will not be initialized.
2819 Severity : enhancement
2821 Description: update patchless client
2822 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2825 Frequency : only with 2.4 kernel
2827 Description: random memory corruption
2828 Details : size of struct ll_inode_info is to big for union inode.u and this
2829 can be cause of random memory corruption.
2834 Description: Memory leak in recovery
2835 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2836 It should also check obd_fail before fsfilt_start, otherwise if
2837 fsfilt_start return -EROFS,(failover mds during mds recovery).
2838 then the req will return with repmsg->transno = 0 and rc = EROFS.
2839 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2840 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2841 be freed no matter whether fsfilt_commit success or not.
2844 Frequency : only with huge count clients
2846 Description: Prevents from taking the superblock lock in llap_from_page for
2848 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2849 for llap_from_page call in ll_removepage prevents from taking the
2850 superblock lock for a soon died page.
2855 Description: Not check open intent error before release open handle
2856 Details : in some rare cases, the open intent error is not checked before
2857 release open handle, which may cause
2858 ASSERTION(open_req->rq_transno != 0), because it tries to release
2859 the failed open handle.
2864 Description: Set cat log bitmap only after create log success.
2865 Details : in some rare cases, the cat log bitmap is set too early. and it
2866 should be set only after create log success.
2870 Description: Accessing a block bevice can re-enable I/O when Lustre is
2871 tearing down a device.
2872 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2877 Description: service threads may hog cpus when there are a lot of requests
2878 Details : Insert cond_resched to give other threads a chance to use some CPU
2883 Description: the cat log was not initialized in recovery
2884 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2885 unlink log on mds will not be initialized until mds post
2886 recovery. And also in mds post recovery, the unlink log will
2887 initialization will be done asynchronausly, so there will be race
2888 between add unlink log and unlink log initialization.
2892 Description: brw_stats were being printed incorrectly
2893 Details : brw_stats were being printed as log2 but all of them were not
2894 recorded as log2. Also remove some code duplication arising from
2895 filter_tally_{read,write}.
2899 Frequency : rare, only in recovery.
2900 Description: ASSERTION(req->rq_type != LI_POISON) failed
2901 Details : imp_lock should be held while iterating over imp_sending_list for
2902 prevent destroy request after get timeout in ptlrpc_queue_wait.
2906 Description: replay-single.sh test 52 fails
2907 Details : A lock's skiplist need to be cleanup when it being unlinked
2908 from its resource list.
2912 Description: Short directio read returns full requested size rather than
2914 Details : Direct I/O operations should return actual amount of bytes
2915 transferred rather than requested size.
2917 Severity : enhancement
2919 Description: metadata RPC reduction (e.g. for rm performance)
2920 Details : decrease the amount of synchronous RPC between clients and servers
2921 by canceling conflicing lock before the operation on the client
2922 and packing thier handles into the main operation RPC to server.
2924 Severity : enhancement
2926 Description: Async OSC create to avoid the blocking unnecessarily.
2927 Details : If a OST has no remain object, system will block on the creating
2928 when need to create a new object on this OST. Now, ways use
2929 pre-created objects when available, instead of blocking on an
2930 empty osc while others are not empty. If we must block, we block
2931 for the shortest possible period of time.
2935 Description: improve handling recoverable errors
2936 Details : if request processig with error which can be recoverable on server
2937 request should be resend, otherwise page released from cache and
2940 Severity : enhancement
2942 Description: refine locking for avoid write wrong info into lov_objid file
2943 Details : fix possible races with add new target and write/update data in
2946 --------------------------------------------------------------------------------
2948 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2953 Frequency : on some architectures
2955 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2956 Details : Uninitialized flags sometimes cause configuration commands to
2960 Frequency : patchless clients only
2962 Description: missing __iget() symbol export
2963 Details : The __iget() symbol export is missing. To avoid the need for
2964 this on patchless clients the deathrow inode reaper is turned
2965 off, and we depend on the VM to clean up old inodes. This
2966 dependency was during via the fix for bug 12181.
2968 --------------------------------------------------------------------------------
2970 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2972 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2973 INTEROPERATE with older versions automatically. In many cases a
2974 special upgrade step is needed. Please read the
2975 user documentation before upgrading any part of a live system.
2976 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2977 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2978 * WARNING: Lustre configuration and startup changes are required with
2979 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2983 Severity : enhancement
2985 Description: MountConf
2986 Details : Lustre configuration is now managed via mkfs and mount
2987 commands instead of lmc and lconf. New obd types (MGS, MGC)
2988 are added for dynamic configuration management. See
2989 https://mail.clusterfs.com/wikis/lustre/MountConf for
2992 Severity : enhancement
2994 Description: dynamic OST addition
2995 Details : OSTs can now be added to a live filesystem
2997 Severity : enhancement
2999 Description: startup order invariance
3000 Details : MDTs and OSTs can be started in any order. Clients only
3001 require the MDT to complete startup.
3003 Severity : enhancement
3005 Description: parallel, asynchronous orphan cleanup
3006 Details : orphan cleanup is now performed in separate threads for each
3007 OST, allowing parallel non-blocking operation.
3009 Severity : enhancement
3011 Description: optimized stripe assignment
3012 Details : stripe assignments are now made based on ost space available,
3013 ost previous usage, and OSS previous usage, in order to try
3014 to optimize storage space and networking resources.
3016 Severity : enhancement
3018 Description: Permanently set tunables
3019 Details : All writable /proc/fs/lustre tunables can now be permanently
3020 set on a per-server basis, at mkfs time or on a live system.
3022 Severity : enhancement
3024 Description: Lustre message v2
3025 Details : Add lustre message format v2.
3027 Severity : enhancement
3029 Description: client OST exclusion list
3030 Details : Clients can be started with a list of OSTs that should be
3031 declared "inactive" for known non-responsive OSTs.
3035 Description: SPEC SFS validation failure on NFS v2 over lustre.
3036 Details : Changes the blocksize for regular files to be 2x RPC size,
3037 and not depend on stripe size.
3039 Severity : enhancement
3041 Description: Multiple MD RPCs in flight.
3042 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3043 To avoid overly-overloading MDS, introduce a limit on number of
3044 MDS RPCs in flight for a single client and add /proc controls
3045 to adjust this limit.
3047 Severity : enhancement
3049 Description: client read/write statistics
3050 Details : Add client read/write call usage stats for performance
3051 analysis of user processes.
3052 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3053 file access. extents_stats shows chunk size distribution.
3054 extents_stats_per_process show chunk size distribution per
3057 Severity : enhancement
3059 Description: mds statistics
3060 Details : Add detailed mds operations statistics in
3061 /proc/fs/lustre/mds/*/stats.
3065 Description: Failure of copying files with lustre special EAs.
3066 Details : Client side always return success for setxattr call for lustre
3067 special xattr (currently only "trusted.lov").
3072 Description: Refcount LNET uuids
3073 Details : The global LNET uuid list grew linearly with every startup;
3074 refcount repeated list entries instead of always adding to
3077 Severity : enhancement
3079 Description: Dynamic service threads
3080 Details : Within a small range, start extra service threads
3081 automatically when the request queue builds up.
3084 Frequency : mixed-endian client/server environments
3086 Description: mixed-endian crashes
3087 Details : The new msg_v2 system had some failures in mixed-endian
3090 Severity : enhancement
3092 Description: Easy OST removal
3093 Details : OSTs can be permanently deactivated with e.g. 'lctl
3094 conf_param lustre-OST0001.osc.active=0'
3096 Severity : enhancement
3098 Description: MGS proc entries
3099 Details : Added basic proc entries for the MGS showing what filesystems
3102 Severity : enhancement
3104 Description: provide MGS failover
3105 Details : Added config lock reacquisition after MGS server failover.
3107 Severity : enhancement
3109 Description: add Linux 2.4 support
3110 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3114 Description: a large application tries to do I/O to the same resource and dies
3115 in the middle of it.
3116 Details : Check the req->rq_arrival time after the call to
3117 ost_brw_lock_get(), but before we do anything about
3118 processing it & sending the BULK transfer request. This
3119 should help move old stale pending locks off the queue as
3120 quickly as obd_timeout.
3123 Frequency : when an incorrect nid is specified during startup
3125 Description: ptlrpc connect to non-existant node causes kernel crash
3126 Details : LNET can't be re-entered from an event callback, which
3127 happened when we expire a message after the export has been
3128 cleaned up. Instead, hand the zombie cleanup off to another
3131 Severity : enhancement
3133 Description: plain/inodebits lock performance improvement
3134 Details : Grouping plain/inodebits in granted list by their request modes
3135 and bits policy, thus improving the performance of search through
3139 Frequency : only if OST filesystem is corrupted
3141 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3142 Details : for a short time RPCs with bulk IO are in the replay list,
3143 but replay of bulk IOs is unimplemented. If the OST filesystem
3144 is corrupted due to disk cache incoherency and then replay is
3145 started it is possible to trip an assertion. Avoid putting
3146 committed RPCs into the replay list at all to avoid this issue.
3149 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3151 Description: System hang on startup
3152 Details : This bug allowed the liblustre (e.g. catamount) client to
3153 return to the app before handling all startup RPCs. This
3154 could leave the node unresponsive to lustre network traffic
3155 and manifested as a server ptllnd timeout.
3157 Severity : enhancement
3159 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3160 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3161 boolean to control whether to print peer debug info when a
3162 client's RPC times out.
3165 Frequency : only for kernels with patches from Lustre below 1.4.3
3167 Description: Remove old rdonly API
3168 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3171 Frequency : only for devices with external journals
3173 Description: Set external device read-only also
3174 Details : During a commanded failover stop, we set the disk device
3175 read-only while the server shuts down. We now also set any
3176 external journal device read-only at the same time.
3179 Frequency : when upgrading from 1.4 while trying to change parameters
3181 Description: The wrong (new) MDC name was used when setting parameters for
3182 upgraded MDT's. Also allows changing of OSC (and MDC)
3183 parameters if --writeconf is specified at tunefs upgrade time.
3186 Frequency : when setting specific ost indicies
3188 Description: QOS code breaks on skipped indicies
3189 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3190 created with --index need not be sequential.
3192 Severity : enhancement
3194 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3195 Details : The uninit_groups feature works in conjunction with the kernel
3196 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3197 up the pass1 processing of e2fsck. This is a read-only feature
3198 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3199 mount filesystems that have had this feature enabled.
3201 Severity : enhancement
3203 Description: Improve multi-block allocation algorithm to avoid fragmentation
3204 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3205 allocation locality and avoid filesystem fragmentation.
3207 ------------------------------------------------------------------------------
3209 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3211 * Support for kernels:
3212 2.4.21-47.0.1.EL (RHEL 3)
3213 2.6.5-7.283 (SLES 9)
3214 2.6.9-42.0.10.EL (RHEL 4)
3215 2.6.12.6 vanilla (kernel.org)
3216 2.6.16.27-0.9 (SLES 10)
3217 * Recommended e2fsprogs version: 1.39.cfs5
3219 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3223 Frequency : occasional, depends on client load and configuration
3224 Bugzilla : 12181, 12203
3225 Description: data loss for recently-modified files
3227 Details : In some cases it is possible that recently written or created
3228 files may not be written to disk in a timely manner (this should
3229 normally be within 30s unless client IO load is very high).
3230 The problem appears as zero-length files or files that are a
3231 multiple of 1MB in size after a client crash or client eviction
3232 that are missing data at the end of the file.
3234 This problem is more likely to be hit on clients where files are
3235 repeatedly created and unlinked in the same directory, clients
3236 have a large amount of RAM, have many CPUs, the filesystem has
3237 many OSTs, the clients are rebooted frequently, and/or the files
3238 are not accessed by other nodes after being written.
3240 The presence of the problem can be detected by looking at
3241 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3242 smaller than the second (nr_unused) then dirty files will not
3243 be flushed automatically to disk. "sync; sleep 10" should be
3244 run several times on the node before unmounting it to update
3245 Lustre (this is also safe to run on nodes without this problem).
3247 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3248 that can cause this same problem, so customers using that kernel
3249 also need to update the kernel in addition to Lustre. In order
3250 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3252 It is normal that files written just before a client crash (less
3253 than 30s) may not yet have been flushed to disk, even for local
3257 Frequency : frequent on thin XT3 nodes
3259 Description: UUID collision on thin XT3 Linux nodes
3260 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3261 randomly, since we relied on an insufficiently-seeded PRNG.
3266 Description: OSS hangs after "All ost request buffers busy"
3267 Details : A deadlock between quota and journal operations caused OSS
3268 hangs after printing "All ost request buffers busy."
3271 Frequency : always on liblustre builds
3273 Description: Cleanup compiler warnings on liblustre
3276 Frequency : always on liblustre builds on XT3
3278 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3282 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3283 > 32000 subdirectories
3284 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3285 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3286 subdirectory count crosses 32000. This will aid e2fsck to
3287 correctly handle more than 32000 subdirectories.
3290 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3292 Description: System hang on startup
3293 Details : This bug allowed the liblustre (e.g. catamount) client to
3294 return to the app before handling all startup RPCs. This
3295 could leave the node unresponsive to lustre network traffic
3296 and manifested as a server ptllnd timeout.
3298 Severity : enhancement
3300 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3301 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3302 boolean to control whether to print peer debug info when a
3303 client's RPC times out.
3308 Description: make O_SYNC working on 2.6 kernels
3309 Details : 2.6 kernels use different method for mark pages for write,
3310 so need add a code to lustre for O_SYNC work.
3315 Description: Failure to close file and release space on NFS
3316 Details : Put inode details into lock acquired in ll_intent_file_open.
3317 Use mdc_intent_lock in ll_intent_open to properly
3318 detect all kind of errors unhandled by mdc_enqueue.
3323 Description: proc file read during shutdown sometimes raced obd removal,
3325 Details : Add lock to prevent obd access after proc file removal.
3328 Frequency : Only for files larger than 4GB on 32-bit clients.
3330 Description: improperly doing page alignment of locks
3331 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3332 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3337 Description: RPCs being resent when they shouldn't be
3338 Details : Some RPCs that should not be resent are being resent. This
3339 can cause inconsistencies in the RPC state machine. Do not
3340 resend such requests.
3343 Frequency : rare, only with NFS export
3345 Description: Crash on NFS re-export node
3346 Details : under very unusual load conditions an assertion is hit in
3347 ll_intent_file_open()
3350 Frequency : only if OST filesystem is corrupted
3352 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3353 Details : for a short time RPCs with bulk IO are in the replay list,
3354 but replay of bulk IOs is unimplemented. If the OST filesystem
3355 is corrupted due to disk cache incoherency and then replay is
3356 started it is possible to trip an assertion. Avoid putting
3357 committed RPCs into the replay list at all to avoid this issue.
3362 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3363 Details : Large single O_DIRECT read and write calls can fail to allocate
3364 a sufficiently large buffer to process the request. In case of
3365 allocation failure the allocation is retried with a smaller
3366 buffer and broken into smaller requests.
3368 Severity : enhancement
3370 Description: Add -o localflock option to simulate old noflock behaviour.
3371 Details : This will achieve local-only flock/fcntl locks coherentness.
3376 Description: versioning check is incomplete
3377 Details : Checking the version difference of client vs. server, report
3378 error if the gap is too big.
3383 Description: add support PG_writeback bit
3384 Details : add support for PG_writeback bit for Lustre, for more carefull
3385 work with page cache in 2.6 kernel. This also fix some deadlocks
3386 and remove hack for work O_SYNC with 2.6 kernel.
3388 Severity : enhancement
3390 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3391 Details : The uninit_groups feature works in conjunction with the kernel
3392 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3393 up the pass1 processing of e2fsck. This is a read-only feature
3394 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3395 mount filesystems that have had this feature enabled.
3397 Severity : enhancement
3399 Description: Improve multi-block allocation algorithm to avoid fragmentation
3400 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3401 allocation locality and avoid filesystem fragmentation.
3403 ------------------------------------------------------------------------------
3405 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3407 * Support for kernels:
3408 2.6.9-42.0.3.EL (RHEL 4)
3409 2.6.5-7.276 (SLES 9)
3410 2.4.21-47.0.1.EL (RHEL 3)
3411 2.6.12.6 vanilla (kernel.org)
3412 2.6.16.21-0.8 (SLES10)
3413 * Recommended e2fsprogs version: 1.39.cfs2-0
3415 * The backwards-compatible /proc/sys/portals symlink has been removed
3416 in this release. Before upgrading, please ensure that you change
3417 any configuration scripts or /etc/sysctl.conf files that access
3418 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3419 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3420 in advance of the upgrade on any system running Lustre 1.4.6 or
3421 newer, since /proc/sys/lnet was added in that version.
3422 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3426 Frequency : only when quota is used
3428 Description: avoid scanning export list for quota master
3429 Details : Change the algorithms to avoid scanning export list in order
3430 to improve the efficiency.
3433 Frequency : MDS failover only, very rarely
3435 Description: "went back in time" messages on mds failover
3436 Details : The greatest transno may be lost when the current operation
3437 finishes with an error (transno==0) and the client's last_rcvd
3438 record is over-written. Save the greatest transno in the
3439 mds_last_transno for this case.
3442 Frequency : always for specific kernels and striping counts
3444 Description: client may get "Matching packet too big" without ACL support
3445 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3446 when trying to access files in certain configurations. The
3447 clients should in fact be denied when mounting because they do
3448 not understand ACLs.
3451 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3453 Description: many clients connecting with IO in progress causes connect timeouts
3454 Details : Avoid synchronous journal commits to avoid delays caused by many
3455 clients connecting/disconnecting when bulk IO is in progress.
3456 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3457 OST_IO_PORTAL to avoid delays behind potentially many pending
3461 Frequency : occasionally with multiple writers to a single file
3463 Description: shared writes to file may result in wrong size reported by stat()
3464 Details : Allow growing of kms when extent lock is cancelled
3467 Frequency : always with random mmap IO to multi-striped file
3469 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3470 Details : Only if the hole is at the end of OST object so that kms is too
3471 small. Fix is to increase kms accordingly in ll_nopage.
3474 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3476 Description: writes to a missing object would leak memory on the OST
3477 Details : If there is an inconsistency between the MDS and OST filesystems,
3478 such that the MDS references an object that doesn't exist, writes
3479 to that object will leak memory due to incorrect cleanup in the
3480 error handling path, eventually running out of memory on the OST.
3485 Description: Creating too long symlink causes lustre errors
3486 Details : Check symlink and name lengths before sending requests to MDS.
3489 Frequency : only if flock is enabled (not on by default)
3491 Description: posix locks not released on fd closure on 2.6.9+
3492 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3493 kernels, this caused such locks not to be released on fd close and
3494 then assertions on fs unmount about still used locks.
3497 Frequency : MDS failover only, very rarely
3499 Description: clients may get ASSERTION(granted_lock != NULL)
3500 Details : When request was taking a long time, and a client was resending
3501 a getattr by name lock request. The were multiple lock requests
3502 with the same client lock handle and
3503 mds_getattr_name->fixup_handle_for_resent_request found one of the
3504 lock handles but later failed with ASSERTION(granted_lock != NULL).
3509 Description: handle->h_buffer_credits > 0, assertion failure
3510 Details : h_buffer_credits is zero after truncate, causing assertion
3511 failure. This patch extends the transaction or creates a new
3515 Frequency : NFS re-export or patchless client
3516 Bugzilla : 11179, 10796
3517 Description: Crash on NFS re-export node (__d_move)
3518 Details : We do not want to hash the dentry if we don't have a lock.
3519 But if this dentry is later used in d_move, we'd hit uninitialised
3520 list head d_hash, so we just do this to init d_hash field but
3521 leave dentry unhashed.
3524 Frequency : NFS re-export or patchless client
3526 Description: NFS exports has problem with symbolic link
3527 Details : lustre client didn't properly install dentry when re-exported
3528 to NFS or running patchless client.
3531 Frequency : NFS re-export or patchless client
3533 Description: Various nfs/patchless fixes.
3534 Details : fixes reuse disconected alias for lookup process - this fixes
3535 warning "find_exported_dentry: npd != pd",
3536 fix permission error with open files at nfs.
3537 fix apply umask when do revalidate.
3540 Frequency : occasional
3542 Description: Crash on NFS re-export node
3543 Details : calling clear_page() on the wrong pointer triggered oops in
3544 generic_mapping_read().
3547 Frequency : rarely, using O_DIRECT IO
3549 Description: unaligned directio crashes client with LASSERT
3550 Details : check for unaligned buffers before trying any requests.
3553 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3555 Description: stale data returned from RAID cache
3556 Details : If only a small amount of IO is done to the RAID device before
3557 reading it again it is possible to get stale data from the RAID
3558 cache instead of reading it from disk.
3561 Frequency : always for sles10 kernel
3563 Description: sles10 support
3564 Details : ll_follow_link: compile fixes and using of nd_set_link
3565 under newer kernels.
3568 Frequency : depends on arch, kernel and compiler version, always on sles10
3571 Description: recursive or deep enough symlinks cause stack overflow
3572 Details : getting rid of large stack-allocated variable in
3576 Frequency : depends on hardware
3578 Description: lustre write performance loss in the SLES10 kernel
3579 Details : the performance loss is caused by using of write barriers in the
3580 ext3 code. The SLES10 kernel turns barrier support on by
3581 default. The fix is to undo that change for ldiskfs.
3583 ------------------------------------------------------------------------------
3585 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3587 * Support for kernels:
3588 2.6.9-42.0.3EL (RHEL 4)
3589 2.6.5-7.276 (SLES 9)
3590 2.4.21-47.0.1.EL (RHEL 3)
3591 2.6.12.6 vanilla (kernel.org)
3595 Frequency : quota enabled and large files being deleted
3597 Description: releasing more than 4GB of quota at once hangs OST
3598 Details : If a user deletes more than 4GB of files on a single OST it
3599 will cause the OST to spin in an infinite loop. Release
3600 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3605 Description: statfs data retrieved from /proc may be stale or zero
3606 Details : When reading per-device statfs data from /proc, in the
3607 {kbytes,files}_{total,free,avail} files, it may appear
3608 as zero or be out of date.
3611 Frequency : systems with MD RAID1 external journal devices
3613 Description: lconf's call to blkid is confused by RAID1 journal devices
3614 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3615 of returning all block devices that match the journal UUID.
3618 Frequency : always, for aggregate stripe size over 4GB
3620 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3621 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3622 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3625 Frequency : always if "lfs find" used on a local file/directory
3627 Description: "lfs find" segfaults if used on a local file/directory
3628 Details : The case where a directory component was not specified wasn't
3629 handled correctly. Handle this properly.
3632 Frequency : always on ppc64
3634 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3635 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3636 perform endianness conversion.
3639 Frequency : rarely (truncate to non-zero file size after write under load)
3640 Bugzilla : 10730, 10687
3641 Description: Files padded with zeros to next 4K multiple
3642 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3643 it is possible that files that are truncated to a non-zero size
3644 immediately after being written are filled with zero bytes beyond
3645 the truncated size. No file data is lost.
3647 Severity : enhancement
3649 Description: Allow recovery/failover for liblustre clients.
3650 Details : liblustre clients were unaware of failover configurations until
3653 Severity : enhancement
3655 Description: user file locks should fail when not mounting with flock option
3656 Details : Set up an error-returning stub in ll_file_operations.lock field
3657 to prevent incorrect behaviour when client is mounted without
3658 flock option. Also, set up properly f_op->flock field for
3662 Frequency : always on ia64
3664 Description: "lfs df" loops on printing out MDS statfs information
3665 Details : The obd_ioctl_data was not initialized and in some systems
3666 this caused a failure during the ioctl that did not return
3667 an error. Initialize the struct and return an error on failure.
3670 Frequency : SLES 9 only
3672 Description: Error of copying files with lustre special EAs as root
3673 Details : Client side always return success for setxattr call for lustre
3674 special xattr (currently only "trusted.lov").
3677 Frequency : rarely on clusters with both ia64+i386 clients
3679 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3680 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3681 clients are doing concurrent writes to the same file it is
3682 possible that the ia64 clients may LASSERT because the OST
3683 extent locks are not PAGE_SIZE aligned. Ensure that grown
3684 locks are always aligned on the request boundary.
3687 Frequency : specific use, occasional
3689 Description: Overwriting in use executable truncates on-disk binary image
3690 Details : If one node attempts to overwrite an executable in use by
3691 another node, we now correctly return ETXTBSY instead of
3692 truncating the file.
3694 Severity : enhancement
3696 Description: Async OSC create to avoid the blocking unnecessarily.
3697 Details : If a OST has no remain object, system will block on the creating
3698 when need to create a new object on this OST. Now, ways use
3699 pre-created objects when available, instead of blocking on an
3700 empty osc while others are not empty. If we must block, we block
3701 for the shortest possible period of time.
3706 Description: chmod on Lustre root is propagated to other clients
3707 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3708 invalid by the follow_mount time.
3711 Frequency : liblustre clients only
3713 Description: Race in 'instant cancel' lock handling could lead to such locks
3714 never to be granted in case of SMP MDS
3715 Details : Do not destroy not yet granted but cbpending locks in
3719 Frequency : replay/resend of open
3721 Description: non null lock assetion failure in mds_intent_policy
3722 Details : Trying to replay/resend lockless open requests resulted in
3723 mds_open() returning 0 with no lock. Now it sets a flag if
3724 a lock is going to be returned.
3726 Severity : enhancement
3728 Description: Checksum enhancements
3729 Details : New checksum enhancements allow for resending RPCs that failed
3732 Severity : enhancement
3734 Description: Tunables on number of dirty pages in cacche
3735 Details : Allow to set limit on number of dirty pages cached.
3740 Description: client crash on unmount - lock still has references
3741 Details : In some error handling cases it was possible to leak a lock
3742 reference on a client while accessing a file. This was not
3743 harmful to the client during operation, but would cause the
3744 client to crash when the filesystem is unmounted.
3747 Frequency : specific case, rare
3749 Description: ETXTBSY on mds though file not in use by client
3750 Details : ETXTBSY is no longer incorrectly returned when attempting to
3751 chmod or chown a directory that the user previously tried to
3752 execute or a currently-executing binary.
3755 Frequency : extremely rare except on liblustre-based clients
3757 Description: Lustre space not freed when files are deleted
3758 Details : Clean up open-unlinked files after client eviction. Previously
3759 the unlink was skipped and the files remained as orphans.
3764 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3765 Details : In some cases it was possible to send a blocking callback to
3766 a client doing a glimpse, even though that client didn't get
3767 a lock granted. When the glimpse lock is cancelled on the OST
3768 the freed lock is left on the waiting list and corrupted the list.
3771 Frequency : all core dumps
3773 Description: Broke core dumps to lustre
3774 Details : Negative dentry may be unhashed if parent does not have UPDATE
3775 lock, but some callers, e.g. do_coredump, expect dentry to be
3776 hashed after successful create, hash it in ll_create_it.
3778 ------------------------------------------------------------------------------
3780 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3782 * Support for kernels:
3783 2.6.9-42.0.2.EL (RHEL 4)
3784 2.6.5-7.276 (SLES 9)
3785 2.4.21-40.EL (RHEL 3)
3786 2.6.12.6 vanilla (kernel.org)
3790 Frequency : always on RHEL 3
3792 Description: Number of open files grows over time
3793 Details : The number of open files grows over time, whether or not
3794 Lustre is started. This was due to a filp leak introduced
3795 by one of our kernel patches.
3797 ------------------------------------------------------------------------------
3799 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3801 * Support for kernels:
3802 2.6.9-42.EL (RHEL 4)
3803 2.6.5-7.276 (SLES 9)
3804 2.4.21-40.EL (RHEL 3)
3805 2.6.12.6 vanilla (kernel.org)
3810 Bugzilla : 5719, 9635, 9792, 9684
3811 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3812 Details : If a server is under heavy load and cannot reply to new
3813 connection requests before the client resends the (re)connect,
3814 the connection handling code can behave badly if two service
3815 threads are concurrently handing separate (re)connections from
3816 the same client. Add better locking to the connection handling
3817 code, and ensure that only a single connection will be processed
3818 for a given client UUID, even if the lock is dropped.
3820 Severity : enhancement
3822 Description: add TCP zero-copy support to kernel
3823 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3824 sends if the hardware supports scatter-gather and checksumming.
3825 This allows socklnd to do client-write and server-read more
3826 efficiently and reduce CPU utilization from skbuf copying.
3829 Frequency : only if NFS exporting from client
3831 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3832 Details : If setting attributes on a file created under NFS that had
3833 never been opened it would be possible to oops the client
3834 if the file had no objects.
3837 Frequency : always for liblustre
3839 Description: liblustre client does MDS+OSTs setattr RPC for each write
3840 Details : When doing a write from a liblustre client, the client
3841 incorrectly issued an RPC to the MDS and each OST the file was
3842 striped over in order to update the timestamps. When writing
3843 with small chunks and many clients this could overwhelm the MDS
3844 with RPCs. In all cases it would slow down the write because
3845 these RPCs are unnecessary.
3847 Severity : enhancement
3849 Description: allow number of MDS service threads to be changed at module load
3850 Details : It is now possible to change the number of MDS service threads
3851 running. Adding "options mds mds_num_threads={N}" to the MDS's
3852 /etc/modprobe.conf will set the number of threads for the next
3853 time Lustre is restarted (assuming the "mds" module is also
3854 reloaded at that time). The default number of threads will
3855 stay the same, 32 for most systems.
3860 Description: OST crash if filesystem is unformatted or corrupt
3861 Details : If an OST is started on a device that has never been formatted
3862 or if the filesystem is corrupt and cannot even mount then the
3863 error handling cleanup routines would dereference a NULL pointer.
3868 Description: NULL pointer deref in llap_from_page.
3869 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3870 incorrect) mapping if the page was truncated/reclaimed while it was
3871 searched for. Check for this condition and skip such pages when
3872 doing readahead. Introduce extra check to llap_from_page() to
3873 verify page->mapping->host is non-NULL (so page is not anonymous).
3876 Frequency : Sometimes when using sys_sendfile
3878 Description: "page not covered by a lock" warnings from ll_readpage
3879 Details : sendfile called ll_readpage without right page locks present.
3880 Now we introduced ll_file_sendfile that does necessary locking
3881 around call to generic_file_sendfile() much like we do in
3885 Frequency : with certain MDS communication failures at client mount time
3887 Description: NULL pointer deref after failed client mount
3888 Details : a client connection request may delayed by the network layer
3889 and not be sent until after the PTLRPC layer has timed out the
3890 request. If the client fails the mount immediately it will try
3891 to clean up before the network times out the request. Add a
3892 reference from the request import to the obd device and delay
3893 the cleanup until the network drops the request.
3896 Frequency : occasionally during client (re)connect
3898 Description: assertion failure during client (re)connect
3899 Details : processing a client connection request may be delayed by the
3900 client or server longer than the client connect timeout. This
3901 causes the client to resend the connection request. If the
3902 original connection request is replied in this interval, the
3903 client may trip an assertion failure in ptlrpc_connect_interpret()
3904 which thought it would be the only running connect process.
3907 Frequency : only with obd_echo servers and clients that are rebooted
3909 Description: kernel BUG accessing uninitialized data structure
3910 Details : When running an obd_echo server it did not start the ping_evictor
3911 thread, and when a client was evicted an uninitialized data
3912 structure was accessed. Start the ping_evictor in the RPC
3913 service startup instead of the OBD startup.
3915 Severity : enhancement
3916 Bugzilla : 10193 (patchless)
3917 Description: Remove dependency on various unexported kernel interfaces.
3918 Details : No longer need reparent_to_init, exit_mm, exit_files,
3919 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3922 Frequency : rare (only users of deprecated and unsupported LDAP config)
3924 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3925 Details : LDAP apparently contains 'lustreName' attributes instead of
3926 'name'. A simple remapping of the name is sufficient.
3929 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3931 Description: waiting_locks_callback trips kernel BUG if client is evicted
3932 Details : Running with the dump_on_timeout debug flag turned on makes
3933 it possible that the waiting_locks_callback() can try to dump
3934 the Lustre kernel debug logs from an interrupt handler. Defer
3935 this log dumping to the expired_lock_main() thread.
3937 Severity : enhancement
3939 Description: Support NFS exporting on 2.6 kernels.
3940 Details : Implement non-rawops metadata methods for NFS server to use without
3941 changing NFS server code.
3944 Frequency : very rare (synthetic metadata workload only)
3946 Description: two racing renames might cause an MDS thread to deadlock
3947 Details : Running the "racer" program may cause one MDS thread to rename
3948 a file from being the source of a rename to being the target of
3949 a rename at exactly the same time that another thread is doing
3950 so, and the second thread has already enqueued these locks after
3951 doing a lookup of the target and is trying to relock them in
3952 order. Ensure that we don't try to re-lock the same resource.
3955 Frequency : only very large systems with liblustre clients
3957 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3958 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3959 all OSTs in parallel. This allows the eviction of stale liblustre
3960 clients to proceed much faster than if they were done in series,
3961 and also offers similar improvements for other set_info RPCs.
3966 Description: excessive CPU usage during initial read phase on client
3967 Details : During the initial read phase on a client, it would agressively
3968 retry readahead on the file, consuming too much CPU and impacting
3969 performance (since 1.4.5.8). Improve the readahead algorithm
3970 to avoid this, and also improve some other common cases (read
3971 of small files in particular, where "small" is files smaller than
3972 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3977 Description: MDS crash when receiving packet with unknown intent.
3978 Details : Do not LBUG in unknown intent case, just return -EFAULT
3980 Severity : enhancement
3981 Bugzilla : 9293, 9385
3982 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3983 Details : Do not serialize getattr (non-intent version) and statfs.
3986 Frequency : occasional, when OST network is overloaded/intermittent
3988 Description: client evicted by OST after bulk IO timeout
3989 Details : If a client sends a bulk IO request (read or write) the OST
3990 may evict the client if it is unresposive to its data GET/PUT
3991 request. This is incorrect if the network is overloaded (takes
3992 too long to transfer the RPC data) or dropped the OST GET/PUT
3993 request. There is no need to evict the client at all, since
3994 the pinger and/or lock callbacks will handle this, and the
3995 client can restart the bulk request.
3998 Frequency : Always when mmapping file with no objects
4000 Description: client crashes when mmapping file with no objects
4001 Details : Check that we actually have objects in a file before doing any
4002 operations on objects in ll_vm_open, ll_vm_close and
4008 Description: Request leak when working with deleted CWD
4009 Details : Introduce advanced request refcount tracking for requests
4010 referenced from lustre intent.
4012 Severity : Enhancement
4014 Description: Cache open file handles on client.
4015 Details : MDS now will return special lock along with openhandle, if
4016 requested and client is allowed to hold openhandle, even if unused,
4017 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4018 closing files for every read/write openration.
4020 Severity : Enhancement
4022 Description: Cache open negative dentries on client when possible.
4023 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4024 negative dentries on lock revocation.
4029 Description: Remounting a client read-only wasn't possible with a zconf mount
4030 Details : It wasn't possible to remount a client read-only with llmount.
4032 Severity : enhancement
4033 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4034 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4035 MPICH is included to add Lustre-specific ADIO interfaces.
4036 This is based closely on the UFS ADIO layer and only differs
4037 in file creation, in order to allow the OST striping to be set.
4038 This is user-contributed code and not supported by CFS.
4043 Description: extended inode attributes (immutable, append-only) work improperly
4044 when 2.4 and 2.6 kernels are used on client/server or vice versa
4045 Details : Introduce kernel-independent values for these flags.
4047 Severity : enhancement
4050 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4051 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4052 granularity. Now, Fractional MB granularity is be supported,
4053 this is very useful for low memory system.
4055 Severity : enhancement
4057 Description: Getattr by fid
4058 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4059 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4060 have locks and avoids getattr rpc requests for it on every stat.
4063 Frequency : Always, for filesystems larger than 2TB
4065 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4066 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4067 filesystems larger than 2TB to be mounted with mballoc enabled.
4070 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4072 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4073 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4074 filesystems larger than 2TB is mounted then the kernel will
4075 silently corrupt the start of the filesystem. CONFIG_LBD is
4076 enabled for all CFS-supported kernels, but the possibility of
4077 this happening with a modified kernel config exists.
4079 Severity : enhancement
4081 Description: add client O_DIRECT support for 2.6 kernels
4082 Details : It is now possible to do O_DIRECT reads and writes to files
4083 in the Lustre client mountpoint on 2.6 kernel clients.
4085 Severity : enhancement
4087 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4088 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4089 parallel, not waiting for response from every OST before sending
4090 a rpc to the next OST.
4095 Description: setattr vs write race when updating file timestamps
4096 Details : Client processes that update a file timestamp into the past
4097 right after writing to the file (e.g. tar) it is possible that
4098 the updated file modification time can be reset to the current
4099 time due to a race between processing the setattr and write RPC.
4101 Severity : enhancement
4103 Description: Bring 'lfs find' closer in line with regular Linux find.
4104 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4105 -print0 options and obtains all the needed info through the lustre
4108 Severity : enhancement
4110 Description: support up to 1024 configured devices on one node
4111 Details : change obd_dev array from statically allocated to dynamically
4112 allocated structs as they are first used to reduce memory usage
4117 Description: Flush dirty partially truncated pages during truncate
4118 Details : Immediatelly flush partially truncated pages in filter_setattr,
4119 this way we completely avoid having any pages in page cache on OST
4120 and can retire ugly workarounds during writes to flush such pages.
4125 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4126 Details : protect inode from truncation within vfs_unlink() context
4127 just take a reference before calling vfs_unlink() and release it
4128 when parent's i_sem is free.
4133 Description: last_id value checked outside lock on OST caused LASSERT failure
4134 Details : If there were multiple MDS->OST object precreate requests in
4135 flight, it was possible that the OST's last object id was checked
4136 outside a lock and incorrectly tripped an assertion. Move checks
4137 inside locks, and discard old precreate requests.
4140 Frequency : always, if extents are used on OSTs
4142 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4143 Details : OSTs using the extents format would not zero the high 16 bits of
4144 the index physical block number. This is not a problem for any
4145 OST filesystems smaller than 16TB, and no kernels support ext3
4146 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4147 new/modified files) and can be fixed for existing filesystems
4148 with e2fsprogs-1.39-cfs1.
4153 Description: import connection selection may be incorrect if timer wraps
4154 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4155 import connections to be ignored if the 32-bit jiffies counter
4156 wraps. Use a 64-bit jiffies counter.
4159 Frequency : very large clusters immediately after boot
4161 Description: LNET request buffers exhausted under heavy short-term load
4162 Details : If a large number of client requests are generated on a service
4163 that has previously never seen so many requests it is possible
4164 that the request buffer growth cannot keep up with the spike in
4165 demand. Instead of dropping incoming requests, they are held in
4166 the LND until the RPC service can accept more requests.
4169 Frequency : Sometimes during replay
4171 Description: Assertion failure in ll_local_open after replay.
4172 Details : If replay happened on an open request reply before we were able
4173 to set replay handler, reply will become not swabbed tripping the
4174 assertion in ll_local_open. Now we set the handler right after
4175 recognising of open request
4178 Frequency : very rare
4180 Description: kernel reports "badness in vsnprintf"
4181 Details : Reading from the "recovery_status" /proc file in small chunks
4182 may cause a negative length in lprocfs_obd_rd_recovery_status()
4183 call to vsnprintf() (which is otherwise harmless). Exit early
4184 if there is no more space in the output buffer.
4186 Severity : enhancement
4188 Description: clear OBD RPC statistics by writing to them
4189 Details : It is now possible to clear the OBD RPC statistics by writing
4190 to the "stats" file.
4195 Description: Client mtime is not the same on different clients after utimes
4196 Details : In some cases, the client was using the utimes() syscall on
4197 a file cached on another node. The clients now validate the
4198 ctime from the MDS + OSTs to determine which one is right.
4203 Description: Inability to activate failout mode
4204 Details : lconf script incorrectly assumed that in pythong string's numeric
4205 value is used in comparisons.
4208 Frequency : always with multiple stripes per file
4210 Description: Inefficient object allocation for mutli-stripe files
4211 Details : When selecting which OSTs to stripe files over, for files with
4212 a stripe count that divides evenly into the number of OSTs,
4213 the MDS is always picking the same starting OST for each file.
4214 Return the OST selection heuristic to the original design.
4219 Description: mount failures may take full timeout to return an error
4220 Details : Under some heavy load conditions it is possible that a
4221 failed mount can wait for the full obd_timeout interval,
4222 possibly several minutes, before reporting an error.
4223 Instead return an error as soon as the status is known.
4225 Frequency : quota enabled and large files being deleted
4227 Description: releasing more than 4GB of quota at once hangs OST
4228 Details : If a user deletes more than 4GB of files on a single OST it
4229 will cause the OST to spin in an infinite loop. Release
4230 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4235 Description: statfs data retrieved from /proc may be stale or zero
4236 Details : When reading per-device statfs data from /proc, in the
4237 {kbytes,files}_{total,free,avail} files, it may appear
4238 as zero or be out of date.
4241 Frequency : systems with MD RAID1 external journal devices
4243 Description: lconf's call to blkid is confused by RAID1 journal devices
4244 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4245 of returning all block devices that match the journal UUID.
4248 Frequency : always, for aggregate stripe size over 4GB
4250 Description: assertion fails when trying to use 4GB stripe size
4251 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4252 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4255 Frequency : always on ppc64
4257 Description: the first write on an ext3 filesystem with mballoc got stuck
4258 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4259 perform endianness conversion.
4261 ------------------------------------------------------------------------------
4263 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4265 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4266 INTEROPERATE with older versions automatically. Please read the
4267 user documentation before upgrading any part of a live system.
4268 * WARNING: Lustre networking configuration changes are required with
4269 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4272 * Support for kernels:
4273 2.6.9-22.0.2.EL (RHEL 4)
4274 2.6.5-7.244 (SLES 9)
4275 2.6.12.6 vanilla (kernel.org)
4278 Severity : enhancement
4279 Bugzilla : 7981/8208
4280 Description: Introduced Lustre Networking (LNET)
4281 Details : LNET is new networking infrastructure for Lustre, it includes
4282 a reorganized network configuration mode (see the user
4283 documentation for full details) as well as support for routing
4284 between different network fabrics. Lustre Networking Devices
4285 (LNDs) for the supported network fabrics have also been
4286 created for this new infrastructure.
4288 Severity : enhancement
4289 Description: Introduced Access control lists
4290 Details : clients can set ACLs on files and directories in order to have
4291 more fine-grained permissions than the standard Unix UGO+RWX.
4292 The MDS must be started with the "-o acl" mount option.
4294 Severity : enhancement
4295 Description: Introduced filesystem quotas
4296 Details : Administrators may now establish per-user quotas on the
4299 Severity : enhancement
4301 Description: Configuration change for the XT3
4302 The PTLLND is now used to run Lustre over Portals on the XT3
4303 The configure option(s) --with-cray-portals are no longer used.
4304 Rather --with-portals=<path-to-portals-includes> is used to
4305 enable building on the XT3. In addition to enable XT3 specific
4306 features the option --enable-cray-xt3 must be used.
4311 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4312 Details : A race between allocating a new llcd and re-getting the llcd_lock
4313 allowed another thread to grab newly-allocated llcd.
4315 Severity : enhancement
4317 Description: 2.6 OST async journal commit and locking fix to improve performance
4318 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4319 2.6 kernels are now async as they already were in 2.4 kernels so
4320 that they can commit concurrently with the network bulk transfer.
4321 For block-allocated files the filter allocation semaphore is held
4322 to avoid filesystem fragmentation during allocation. BKL lock
4323 removed for 2.6 xattr operations where it is no longer needed.
4328 Description: lconf incorrectly determined whether two IP networks could talk
4329 Details : In some more complicated routing and multiple-network
4330 configurations, lconf will avoid trying to make a network
4331 connection to a disjoint part of the IP space. It was doing the
4332 math incorrectly for one set of cases.
4337 Description: Fix for potential infinite loop processing records in an llog.
4338 Details : If an llog record is corrupted/zeroed, it is possible to loop
4339 forever in llog_process(). Validate the llog record length
4340 and skip the remainder of the block on error.
4343 Frequency : occasional (liblustre only)
4345 Description: liblustre could not open files whose last component is a symlink
4346 Details : sysio_path_walk() would incorrectly pass the open intent to
4347 intermediate path components.
4350 Frequency : rare (liblustre only with non-standard tuning)
4351 Bugzilla : 7201 (7350)
4352 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4353 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4354 after it has been released, only do this for referenced locks
4356 Severity : enhancement
4358 Description: specifying an (invalid) directory default stripe_size of -1
4359 would reset the directory default striping
4360 Details : stripe_size -1 was used internally to signal directory stripe
4361 removal, now use "all default" to signal dir stripe removal
4362 as a directory striping of "all default" is not useful
4365 Frequency : common for large clusters running liblustre clients
4367 Description: doing an ls when liblustre clients are running is slow
4368 Details : sending a glimpse AST to a liblustre client waits for every AST
4369 to time out, as liblustre clients will not respond. Since they
4370 cannot cache data we refresh the OST lock LVB from disk instead.
4372 Severity : enhancement
4374 Description: doing an ls at the same time as file IO can be slow
4375 Details : enqueue and other "small" requests can be blocked behind many
4376 large IO requests. Create a new OST IO portal for non-IO
4377 requests so they can be processed faster.
4380 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4382 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4383 created file with mode 000
4384 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4385 allow an ftruncate() as a non-root user to a file with mode 000.
4386 Lustre now respects this flag to disable mode checks when
4387 truncating a file owned by the user
4390 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4393 Description: Revoking locks from clients that went dead or catatonic might take
4395 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4396 cancellation of such locks instant on servers without waiting for
4397 any reply from clients. Clients drops these locks when cancel
4398 notification from server is received without replying.
4401 Frequency : liblustre-only, when liblustre client dies or becomes busy
4403 Description: Doing ls on Linux clients can take a long time with active
4405 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4406 granting such locks to it in the first place if possible. Locks
4407 are taken by proxy on the OST during the read or write and
4408 dropped immediately afterward. Add connect flags handling, do
4409 not grant locks to liblustre clients for glimpse ASTs.
4411 Severity : enhancement
4413 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4414 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4415 sequential reads and behaves badly with applications doing random
4416 reads. Improve it by reading ahead at least read region, and
4417 avoiding excessive large RPC for small reads.
4419 Severity : enhancement
4421 Description: Creating more than 1000 files for a single job may cause a load
4422 imbalance on the OSTs if there are also a large number of OSTs.
4423 Details : qos_prep_create() uses an OST index reseed value that is an
4424 even multiple of the number of available OSTs so that if the
4425 reseed happens in the middle of the object allocation it will
4426 still utilize the OSTs as uniformly as possible.
4431 Description: OST or MDS may oops in ping_evictor_main()
4432 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4433 but doesn't restart at beginning of obd_exports_timed list
4436 Severity : enhancement
4438 Description: improve by-nid export eviction on the MDS and OST
4439 Details : allow multiple exports with the same NID to be evicted at one
4440 time without re-searching the exports list.
4443 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4445 Description: MDS may oops in groups_free()
4446 Details : in rare race conditions a newly allocated group_info struct is
4447 freed again, and this can be NULL. The 2.4 compatibility code
4448 for groups_free() checked for a NULL pointer, but 2.6 did not.
4451 Frequency : common for liblustre clients doing little filesystem IO
4452 Bugzilla : 9352, 7313
4453 Description: server may evict liblustre clients accessing contended locks
4454 Details : if a client is granted a lock or receives a completion AST
4455 with a blocking AST already set it would not reply to the AST
4456 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4459 Frequency : lfs setstripe, only systems with more than 160 OSTs
4461 Description: unable to set striping with a starting offset beyond OST 160
4462 Details : llapi_create_file() incorrectly limited the starting stripe
4463 index to the maximum single-file stripe count.
4466 Frequency : LDAP users only
4468 Description: lconf did not handle in-kernel recovery with LDAP properly
4469 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4471 Severity : enhancement
4473 Description: bind OST threads to NUMA nodes to improve performance
4474 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4475 node and do their allocations there to localize memory access
4477 Severity : enhancement
4479 Description: llmount can determine client NID directly from Myrinet (GM)
4480 Details : the client NID code from gmnalnid was moved directly into
4481 llmount, removing the need to use this or specifying the
4482 client NID explicitly when mounting GM clients with zeroconf
4485 Frequency : if client is started with down MDS
4487 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4488 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4489 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4490 ptlrpc_import_delayed_req() because we didn't check if the
4491 request was interrupted, and we also didn't break out of the
4492 event loop if there was no timeout
4497 Description: data loss during non-page-aligned writes to a single file from
4498 both multiple nodes and multiple threads on one node at same time
4499 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4500 inconsistency led to false short-reads, that were cached and later
4501 used by ->prepare_write() to fill in partially written page,
4502 leading to data loss.
4505 Frequency : always, if lconf --abort_recovery used
4507 Description: lconf --abort_recovery fails with 'Operation not supported'
4508 Details : lconf was attempting to abort recovery on the MDT device and not
4511 Severity : enhancement
4513 Description: remove cleanup logs
4514 Details : replace lconf-generated cleanup logs with lustre internal
4515 cleanup routines. Eliminates the need for client-cleanup and
4518 Severity : enhancement
4520 Description: add support for EAs (user and system) on lustre filesystems
4521 Details : it is now possible to store extended attributes in the Lustre
4522 client filesystem, and with the user_xattr mount option it
4523 is possible to allow users to store EAs on their files also
4525 Severity : enhancement
4527 Description: Add possibility (config option) to show minimal available OST free
4529 Details : When compiled with --enable-mindf configure option, statfs(2)
4530 (and so, df) will return least minimal free space available from
4531 all OSTs as amount of free space on FS, instead of summary of
4532 free spaces of all OSTs.
4534 Severity : enhancement
4536 Description: do not expand extent locks acquired on OST-side
4537 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4538 by server: they are not cached anyway.
4541 Frequency : when mmap is used/binaries executed from Lustre
4543 Description: Unmmap pages before throwing them away from read cache.
4544 Details : llap_shrink cache now attempts to unmap pages before discarding
4545 them (if unmapping failed - do not discard). SLES9 kernel has
4546 extra checks that trigger if this unmapping is not done first.
4551 Description: lconf didn't resolve symlinks before checking to see whether a
4552 given mountpoint was already in use
4555 Frequency : when migrating failover services
4556 Bugzilla : 6395, 9514
4557 Description: When migrating a subset of services from a node (e.g. failback
4558 from a failover service node) the remaining services would
4559 time out and evict clients.
4560 Details : lconf --force (implied by --failover) sets the global obd_timeout
4561 to 5 seconds in order to quickly disconnect, but this caused
4562 other RPCs to time out too quickly. Do not change the global
4563 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4565 Severity : enhancement
4566 Frequency : if MDS is started with down OST
4567 Bugzilla : 9439,5706
4568 Description: Allow startup/shutdown of an MDS without depending on the
4569 availability of the OSTs.
4570 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4571 Add appropriate locking and lov-osc refcounts for safe
4572 cleaning. Add osc abort_inflight calls in case the
4573 synchronize never started.
4576 Frequency : occasional (Cray XT3 only)
4578 Description: root not authorized to access files in CRAY_PORTALS environment
4579 Details : The client process capabilities were not honoured on the MDS in
4580 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4581 been accessed by an authorized user then root was able to access
4582 the file on the local client also. The root user capabilities
4583 are now allowed on the MDS, as this environment has secure UID.
4586 Frequency : occasional
4588 Description: ldiskfs "too long searching" message happens too often
4589 Details : A debugging message (otherwise harmless) prints too often on
4590 the OST console. This has been reduced to only happen when
4591 there are fragmentation problems on the filesystem.
4596 Description: Division by zero in statfs when all OSCs are inactive
4597 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4598 lov_statfs divides by value returned by lov_get_stripecnt().
4602 Bugzilla : 9489, 3273
4603 Description: First write from each client to each OST was only 4kB in size,
4604 to initialize client writeback cache, which caused sub-optimal
4605 RPCs and poor layout on disk for the first writen file.
4606 Details : Clients now request an initial cache grant at (re)connect time
4607 and so that they can start streaming writes to the cache right
4608 away and always do full-sized RPCs if there is enough data.
4609 If the OST is rebooted the client also re-establishes its grant
4610 so that client cached writes will be honoured under the grant.
4615 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4616 Details : Now I/O RPCs go to different portal number and (presumably) fast
4617 lock requests (and glimses) and other RPCs get their own service
4618 threads pool that should be able to service those RPCs
4621 Severity : enhancement
4623 Description: Ability to exchange lustre version between client and servers and
4624 issue warnings at client side if client is too old. Also for
4625 liblustre clients there is ability to refuse connection of too old
4627 Details : New 'version' field is added to connect data structure that is
4628 filled with version info. That info is later checked by server and
4632 Frequency : rare, liblustre only.
4633 Bugzilla : 9296, 9581
4634 Description: Two simultaneous writes from liblustre at offset within same page
4635 might proceed at the same time overwriting eachother with stale
4637 Details : I/O lock withing llu_file_prwv was released too early, before data
4638 actually was hitting the wire. Extended lock-holding time until
4639 server acknowledges receiving data.
4642 Frequency : extremely rare. Never observed in practice.
4644 Description: avoid generating lustre_handle cookie of 0.
4645 Details : class_handle_hash() generates handle cookies by incrementing
4646 global counter, and can hit 0 occasionaly (this is unlikely, but
4647 not impossible, because initial value of cookie counter is
4648 selected randonly). Value of 0 is used as a sentinel meaning
4649 "unassigned handle" --- avoid it. Also coalesce two critical
4650 sections in this function into one.
4652 Severity : enhancement
4654 Description: allow liblustre clients to delegate truncate locking to OST
4655 Details : To avoid overhead of locking, liblustre client instructs OST to
4656 take extent lock in ost_punch() on client's behalf. New connection
4657 flag is added to handle backward compatibility.
4659 Severity : enhancement
4660 Bugzilla : 4928, 7341, 9758
4661 Description: allow number of OST service threads to be specified
4662 Details : a module parameter allows the number of OST service threads
4663 to be specified via "options ost ost_num_threads={N}" in the
4664 OSS's /etc/modules.conf or /etc/modprobe.conf.
4668 Bugzilla : 6146, 9635, 9895
4669 Description: servers crash with bad pointer in target_handle_connect()
4670 Details : In rare cases when a client is reconnecting it was possible that
4671 the connection request was the last reference for that export.
4672 We would temporarily drop the export reference and get a new
4673 one, but this may have been the last reference and the export
4674 was just destroyed. Get new reference before dropping old one.
4676 Severity : enhancement
4677 Frequency : if client is started with failover MDS
4679 Description: Allow multiple MDS hostnames in the mount command
4680 Details : Try to read the configuration from all specified MDS
4681 hostnames during a client mount in case the "primary"
4684 Severity : enhancement
4686 Description: Stop sending data to evicted clients as soon as possible.
4687 Details : Check if the client we are about to send or are sending data to
4688 was evicted already. (Check is done every second of waiting,
4689 for which l_wait_event interface was extended to allow checking
4690 of exit condition at specified intervals).
4693 Frequency : rare, normally only when NFS exporting is done from client
4695 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4697 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4700 Frequency : always, if srand() is called before liblustre initialization
4702 Description: Liblustre uses system PRNG disturbing its usage by user application
4703 Details : Introduce internal to lustre fast and high-quality PRNG for
4704 lustre usage and make liblustre and some other places in generic
4705 lustre code to use it.
4707 Severity : enhancement
4708 Bugzilla : 9477, 9557, 9870
4709 Description: Verify that the MDS configuration logs are updated when xml is
4710 Details : Check if the .xml configuration logs are newer than the config
4711 logs stored on the MDS and report an error if this is the case.
4712 Request --write-conf, or allow starting with --old_conf.
4714 Severity : enhancement
4716 Description: Handle symlinks in the path when checking if Lustre is mounted.
4717 Details : Resolve intermediate symlinks when checking if a client has
4718 mounted a filesystem to avoid duplicate client mounts.
4723 Description: lconf can hit an error exception but still return success.
4724 Details : The lconf command catches the Command error exception at the top
4725 level script context and will exit with the associated exit
4726 status, but doesn't ensure that this exit status is non-zero.
4731 Description: failure of ptlrpc thread startup can cause oops
4732 Details : Starting a ptlrpc service thread can fail if there are a large
4733 number of threads or the server memory is very fragmented.
4734 Handle this without oopsing.
4737 Frequency : always, only if liblustre and non-default acceptor port was used
4739 Description: liblustre cannot connect to servers with non-default acceptor port
4740 Details : tcpnal_set_default_params() was not called and was therefore
4741 ignoring the environment varaible TCPNAL_PORT, as well as other
4742 TCPNAL_ environment variables
4747 Description: two objects could be created on the same OST for a single file
4748 Details : If an OST is down, in some cases it was possible to create two
4749 objects on a single OST for a single file. No problems other
4750 than potential performance impact and spurious error messages.
4754 Bugzilla : 5681, 9562
4755 Description: Client may oops in ll_unhash_aliases
4756 Details : Client dcache may become inconsistent in race condition.
4757 In some cases "getcwd" can fail if the current directory is
4763 Description: Inode refcounting problems in NFS export code
4764 Details : link_raw functions used to call d_instantiate without obtaining
4765 extra inode reference first.
4769 Bugzilla : 9942, 9903
4770 Description: Referencing freed requests leading to crash, memleaks with NFS.
4771 Details : We used to require that call to ll_revalidate_it was always
4772 followed by ll_lookup_it. Also with revalidate_special() it is
4773 possible to call ll_revalidate_it() twice for the same dentry
4774 even if first occurence returned success. This fix changes semantic
4775 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4776 reference on a request referred from the intent.
4777 ll_intent_release() then releases such a request.
4780 Frequency : rare, normally benchmark loads only
4782 Description: unlinked inodes were kept in memory on the client
4783 Details : If a client is repeatedly creating and unlinking files it
4784 can accumulate a lot of stale inodes in the inode slab cache.
4785 If there is no other client load running this can cause the
4786 client node to run out of memory. Instead flush old inodes
4787 from client cache that have the same inode number as a new inode.
4790 Frequency : SLES9 2.6.5 kernel and long filenames only
4791 Bugzilla : 9969, 10379
4792 Description: utime reports stale NFS file handle
4793 Details : SLES9 uses out-of-dentry names in some cases, which confused
4794 the lustre dentry revalidation. Change it to always use the
4798 Frequency : rare, unless heavy write-truncate concurrency is continuous
4799 Bugzilla : 4180, 6984, 7171, 9963, 9331
4800 Description: OST becomes very slow and/or deadlocked during object unlink
4801 Details : filter_destroy() was holding onto the parent directory lock
4802 while truncating+unlinking objects. For very large objects this
4803 may block other threads for a long time and slow overall OST
4804 responsiveness. It may also be possible to get a lock ordering
4805 deadlock in this case, or run out of journal credits because of
4806 the combined truncate+unlink. Solution is to do object truncate
4807 first in one transaction without parent lock, and then do the
4808 final unlink in a new transaction with the parent lock. This
4809 reduces the lock hold time dramatically.
4812 Frequency : rare, 2.4 kernels only
4814 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4815 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4816 before unmounting their filesystem in order to avoid deadlock.
4817 The kernel_locked() function in 2.4 kernels only checks whether
4818 the kernel lock is held, not whether it is this process that is
4819 holding it as 2.6 kernels do.
4824 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4825 Details : The client ptlrpc code may be trying to reconnect to a down
4826 server before a previous connection attempt has timed out.
4827 Increase the reconnect interval to be longer than the connection
4828 timeout interval to avoid sending duplicate connections to
4832 Frequency : echo_client brw_test command
4834 Description: fix echo_client to work with OST preallocated code
4835 Details : OST preallocation code (5137) didn't take echo_client IO path
4836 into account: echo_client calls filter methods outside of any
4837 OST thread and, hence, there is no per-thread preallocated
4838 pages and buffers to use. Solution: hijack pga pages for IO. As
4839 a byproduct, this avoids unnecessary data copying.
4843 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4844 Description: Client can oops in mdc_commit_close() after open replay
4845 Details : It was possible for the MDS to return an open request with no
4846 transaction number in mds_finish_transno() if the client was
4847 evicted, but without actually returning an error. Clients
4848 would later try to replay that open and may trip an assertion
4849 Simplify the client close codepath, and always return an error
4850 from the MDS in case the open is not successful.
4853 Frequency : rare, 2.6 OSTs only
4855 Description: OST may deadlock under high load on fragmented files
4856 Details : If there was a heavy load and highly-fragmented OST filesystems
4857 it was possible to have all the OST threads deadlock waiting on
4858 allocation of biovecs, because the biovecs were not released
4859 until the entire RPC IO was completed. Instead, release biovecs
4860 as soon as they are complete to ensure forward IO progress.
4862 Severity : enhancement
4864 Description: Support for specifying external journal device at mount
4865 Details : If an OST or MDS device is formatted with an external journal
4866 device, this device major/minor is stored in the ext3 superblock
4867 and may not be valid for failover. Allow detecting and
4868 specifying the external journal at mount time.
4873 Description: Mounting an MDS with pending unlinked files may cause oops
4874 Details : target_finish_recovery() calls mds_postrecov() which returned
4875 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4876 considers this an error and immediately begins cleaning up the
4877 lov, just after starting the mds_lov process
4879 Severity : enhancement
4881 Description: Implement 'lfs df' to report actual free space on per-OST basis
4882 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4883 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4884 usage of objects; '-h' to report in human readable format.
4886 ------------------------------------------------------------------------------
4888 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4895 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4896 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4897 at startup during orphan unlinking. Instead, skip buddy update
4898 before setup, as it will be regenerated after recovery is complete.
4903 Description: If an OST is inactive, its locks might reference stale inodes.
4904 Details : lov_change_cbdata() must iterate over all namespaces, even if
4905 they are inactive to clear inode references from the lock.
4907 Severity : enhancement
4908 Frequency : occasional, if non-standard max_dirty_mb used
4910 Description: Client will block write RPCs if not enough grant
4911 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4912 then the client will block writes while waiting for another RPC
4913 to complete instead of consuming its dirty limit. With change
4914 we get improved performance when max_dirty_mb is small.
4916 Severity : enhancement
4917 Bugzilla : 3389, 6253
4918 Description: Add support for supplementary groups on the MDS.
4919 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4920 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4921 lookups for user supplementary groups into a cache.
4925 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4926 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4927 for liblustre O_EXCL is enforced by MDS.
4931 Description: readdir never returns NULL in liblustre.
4932 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4933 dentry in struct dirent.
4937 Frequency : liblustre only, depends on application IO pattern
4938 Description: liblustre clients evicted if not contacting servers
4939 Details : Don't put liblustre clients into the ping_evictor list, so
4940 they will not be evicted by the pinger ever.
4942 Severity : enhancement
4944 Description: Add ability to evict clients by NID from MDS.
4945 Details : By echoing "nid:$NID" string into
4946 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4947 $NID would be instantly evicted from this MDS and from all active
4948 OSTs connected to it.
4952 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4953 Details : lookup_it_finish() used to query file size from OSTs that was not
4958 Description: service threads change working directory to that of init
4959 Details : Starting lustre service threads may pin the working directory
4960 of the parent thread, making that filesystem busy. Threads
4961 now change to the working directory of init to avoid this.
4965 Frequency : during shutdown only
4966 Description: shutdown with a failed MDS or OST can cause unmount to hang
4967 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4972 Frequency : 2.6 only
4973 Description: chmod/chown may include an extra supplementary group
4974 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4975 supplementary group and if none is specified this is used.
4978 Bugzilla : 5479 (6816)
4980 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4981 Details : If lookup is in progress on a file that is unlinked we might try
4982 to revalidate the inode and fail in revalidate after lookup is
4983 complete and ll_file_open() enqueues the open again but
4984 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4987 Frequency : always, if lconf --abort_recovery used
4989 Description: lconf --abort_recovery fails with 'Operation not supported'
4990 Details : lconf was attempting to abort recovery on the MDT device and not
4993 ------------------------------------------------------------------------------
4995 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5000 Frequency : rare (only unsupported configurations with a node running as an
5002 Bugzilla : 6514, 5137
5003 Description: Mounting a Lustre file system on a node running as an OST could
5005 Details : OSTs now preallocates memory needed to write out data at
5006 startup, instead of when needed, to avoid having to
5007 allocate memory in possibly low memory situations.
5008 Specifically, if the file system is mounted on on OST,
5009 memory pressure could force it to try to write out data,
5010 which it needed to allocate memory to do. Due to the low
5011 memory, it would be unable to do so and the node would
5012 become unresponsive.
5014 Severity : enhancement
5016 Description: Addition of lconf --service command line option
5017 Details : lconf now accepts a '--service <arg>' option, which is
5018 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5020 Severity : enhancement
5022 Description: Failover mode is now the default for OSTs.
5023 Details : By default, OSTs will now run in failover mode. To return to
5024 the old behaviour, add '--failout' to the lmc line for OSTs.
5026 Severity : enhancement
5028 Description: Health checks are now provided for MDS and OSTs
5029 Details : Additional detailed health check information on MSD and OSTs
5030 is now provided through the procfs health_check value.
5033 Frequency : occasional, depends on IO load
5035 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5036 after numerous create/delete cycles
5037 Details : The ext3 inode allocation policy would not allocate new inodes
5038 very well on the OSTs because there are no new directories
5039 being created. Instead we look for groups with free space if
5040 the parent directories are nearly full.
5045 Description: Network or server problems during mount may cause partially
5046 mounted clients instead of returning an error.
5047 Details : The config llog parsing code may overwrite the error return
5048 code during mount error handling, returning success instead
5054 Description: MDS can fail to allocate large reply buffers
5055 Details : After long uptimes the MDS can fail to allocate large reply
5056 buffers (e.g. zconf client mount config records) due to memory
5057 fragmentation or consumption by the buffer cache. Preallocate
5058 some large reply buffers so that these replies can be sent even
5059 under memory pressure.
5063 Frequency : rare (liblustre)
5064 Description: fsx running with liblustre complained that using truncate() to
5065 extend the file doesn't work. This patch corrects that issue.
5066 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5067 ATTR_SIZE and lsm use in llu_setattr_raw.
5071 Frequency : rare, only 2.6 kernels
5072 Description: Unusual file access patterns on the MDS may result in inode
5073 data being lost in very rare circumstances.
5074 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5075 code in the 2.6 kernel caused the inode and/or EA data not to
5076 be read from disk, causing single-file corruption.
5080 Frequency : rare, only 2.6 filesystems using extents
5081 Description: Heavy concurrent write and delete load may cause data corruption.
5082 Details : It was possible under high-load situations to have an extent
5083 metadata block in the block device cache from a just-unlinked
5084 file overwrite a newly-allocated data block. We now unmap any
5085 metadata buffers that alias just-allocated data blocks.
5089 Frequency : filesystems with default stripe_count larger than 77
5090 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5091 Details : lconf specifies an inode size of 4096 bytes when the default
5092 stripe_count is larger than 77. This conflicts with the default
5093 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5094 this case to avoid pinning too much memory for large EAs.
5096 ------------------------------------------------------------------------------
5098 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5103 Frequency : rare (extremely heavy IO load with hundreds of clients)
5105 Description: Client is evicted, gets IO error writing to file
5106 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5107 caused clients to be evicted for AST timeouts. The fixes in
5108 bug 5192 mean we no longer need to have such short AST timeouts
5109 so ldlm_timeout has been increased.
5112 Frequency : occasional during --force or --failover shutdown under load
5113 Bugzilla : 5949, 4834
5114 Description: Server oops/LBUG if stopped with --force or --failover under load
5115 Details : a collection of import/export refcount and cleanup ordering
5116 issues fixed for safer force cleanup
5119 Frequency : only filesystems larger than 120 OSTs
5120 Bugzilla : 5990, 6223
5121 Description: lfs getstripe would oops on a very large filesystem
5122 Details : lov_getconfig used kfree on vmalloc'd memory
5125 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5126 Bugzilla : 6242, 6243
5127 Description: reading from files that had been truncated to a non-zero size
5128 but never opened returned no data
5129 Details : ll_file_read() reads zeros from no-object files to EOF
5134 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5135 Details : There's little harm in aborting MDS/OSS recovery and letting it
5136 try again, so I removed the LASSERT and return an error instead.
5138 Severity : enhancement
5140 Description: New debugging infrastructure for tracking down data corruption
5141 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5142 (b) cover more of the client-side code path, and (c) try to narrow
5143 down where problems occurred
5147 Bugzilla : 3819, 4364, 4397, 6313
5148 Description: Racing close and eviction MDS could cause assertion in mds_close
5149 Details : It was possible to get multiple mfd references during close and
5150 client eviction, leading to one thread referencing a freed mfd.
5152 Severity: : enhancement
5153 Bugzilla : 3262, 6359
5154 Description: Attempts to reconnect to servers are now more aggressive.
5155 Details : This builds on the enhanced upcall-less recovery that was added
5156 in 1.4.2. When trying to reconnect to servers, clients will
5157 now try each server in the failover group every 10 seconds. By
5158 default, clients would previously try one server every 25 seconds.
5163 Description: After recovery, certain operations trigger a failed
5164 assertion on a client.
5165 Details : Failing over an mds, using lconf -d --failover, while a
5166 client was doing a readdir() call would cause the client to
5167 LBUG after recovery completed and the readdir() was resent.
5169 Severity : enhancement
5171 Description: Default groups are now added by lconf
5172 Details : You can now run lconf --group <servicename> without having to
5173 manually add groups with lmc.
5176 Frequency : occasional
5178 Description: Nodes with an elan id of 0 trigger a failed assertion
5181 Frequency : always when accessing e.g. tty/console device nodes
5183 Description: tty and some other devices nodes cannot be used on lustre
5184 Details : file's private_data field is used by device data and lustre
5185 values in there got lost. New field was added to struct file to
5186 store fs-specific private data.
5189 Frequency : when exporting Lustre via NFS
5191 Description: NFSD failed occasionally when looking up a path component
5192 Details : NFSD is looking up ".." which was broken in ext3 directories
5193 that had grown large enough to become hashed.
5196 Frequency : Clusters with multiple interfaces not on the same subnet
5198 Description: Nodes will repeatedly try to reconnect to an interface which it
5199 cannot reach and report an error to the log.
5200 Details : Extra peer list entries will be created by lconf with some peers
5201 unreachable. lconf now validates the peer before adding it.
5204 Frequency : Only if a default stripe is set on the filesystem root.
5206 Description: Setting a default stripe on the filesystem root prevented the
5207 filesystem from being remounted.
5208 Details : The client was sending extra request flags in the root getattr
5209 request and did not allocate a reply buffer for the dir EA.
5212 Frequency : occasional, higher if lots of files are accessed by one client
5213 Bugzilla : 6159, 6097
5214 Description: Client trips assertion regarding lsm mismatch/magic
5215 Details : While revalidating inodes the VFS looks up inodes with ifind()
5216 and in rare cases can find an inode that is being freed.
5217 The ll_test_inode() code will free the lsm during ifind()
5218 when it finds an existing inode and then the VFS later attaches
5219 this free lsm to a new inode.
5223 Bugzilla : 6422, 7030
5224 Description: MDS deadlock between mkdir and client eviction
5225 Details : Creating a new file via mkdir or mknod (starting a transaction
5226 and getting the ns lock) can deadlock with client eviction
5227 (gets ns lock and trying to finish a synchronous transaction).
5230 Frequency : occasional
5231 Description: While starting a server, the fsfilt_ext3 module could not be
5233 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5234 kernels. Previously, lconf would still use the ext3 name
5235 when trying to load modules. Now, it will correctly use
5236 ext3 on 2.4 and ldiskfs on 2.6.
5238 Severity : enhancement
5239 Description: The default stripe count has been changed to 1
5240 Details : The interpretation of the default stripe count (0, to lfs
5241 or lmc) has been changed to mean striping across a single
5242 OST, rather than all available. For general usage we have
5243 found a stripe count of 1 or 2 works best.
5245 Severity : enhancement
5246 Description: Add support for compiling against Cray portals.
5247 Details : Conditional compiling for some areas that are different
5251 Frequency : occasional
5252 Bugzilla : 6409, 6834
5253 Description: Creating files with an explicit stripe count may lead to
5254 a failed assertion on the MDS
5255 Details : If some OSTs are full or unavailable, creating files may
5256 trigger a failed assertion on the MDS. Now, Lustre will
5257 try to use other servers or return an error to the
5261 Frequency : occasional
5263 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5264 caused lock timeout during readahead (since 1.4.2).
5265 Details : Processes doing readahead might match a lock that hasn't been
5266 granted yet if there are overlapping and conflicting lock
5267 requests. The readahead process waits on ungranted lock
5268 (original lock is CBPENDING), while OST waits for that process
5269 to cancel CBPENDING read lock and eventually evicts client.
5271 Severity : enhancement
5273 Description: Initial enabling of flock support for clients
5274 Details : Implements fcntl advisory locking and file status functions.
5275 This feature is provided as an optional mount flag (default
5276 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5277 locking are implemented yet, and those that are are not guaranteed
5278 to be completely correct in production environments.
5279 mount -t lustre -o [flock|noflock] ...
5282 Frequency : occasional
5284 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5285 assertion in the ext3 JBD (journaling) layer.
5286 Details : The b_committed_data struct is protected by the big kernel lock
5287 in 2.4 kernels, serializing journal_commit_transaction() and
5288 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5289 access to this struct. In 2.6 kernels there is finer grained
5290 locking to improve SMP performance of the JBD layer.
5294 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5299 Description: merge and cleanup kernel patches.
5302 -----------------------------------------------------------------------------
5304 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5306 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5307 versions of Lustre. Please update all servers and clients to
5308 version 1.4.2 or later at the same time. You must also run
5309 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5310 to update the configuration logs.
5312 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5313 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5314 - swab llog records of type '0' so we get proper header size/idx (5861)
5315 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5316 - fix rename of one directory over another leaking an inode (5953)
5317 - avoid SetPageDirty on 2.6 (5981)
5318 - don't re-add just-being-destroyed locks to the waiting list (5653)
5319 - when creating new directories, inherit the parent's custom
5320 striping settings if present parent (3048)
5321 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5322 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5323 - don't hold client locks on temporary worklist from l_lru (5666)
5324 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5325 - automatically evict dead clients (5921)
5326 - Update file size properly in create+truncate+fstat case (6196)
5327 - Do not unhash mountpoint dentries, do not allow removal of
5329 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5330 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5331 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5332 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5333 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5334 - Correctly handle reads of files with no objects (6243)
5335 - lctl recover will also mark a device active if deactivate used (5933)
5337 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5338 - allow --write-conf on an MDS with different nettype than client (5619)
5339 - don't write config llogs to MDS for mounts not from that MDS (5617)
5340 - lconf should create multiple TCP connections from a client (5201)
5341 - init scripts are now turned off by default; run chkconfig --on
5342 lustre and chkconfig --on lustrefs to use them
5343 - upcalls are no longer needed for clients to recover to failover
5345 - add --abort-recovery option to lconf to abort recovery on device
5347 - add support for an arbitrary number of OSTs (3026)
5348 - Quota support protocol changes.
5349 - forward compatibility changes to wire structs (6007)
5350 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5351 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5352 - improved "lustre status" script
5353 - initialize blocksize for non-regular files (6062)
5354 - added --disable-server and --disable-client configure options (5782)
5355 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5356 - Vanilla 2.4.29 support
5357 - increase maximum number of obd devices to 520 (6242)
5358 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5359 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5360 SLES 9 2.6.5 kernels
5361 - limit stripes per file to 160 (the maximum EA size) (6093)
5363 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5366 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5367 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5368 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5369 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5370 - small ext3 extents cleanups and fixes (5733)
5371 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5372 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5373 - avoid needless client->OST connect, fix handle mismatch (5317)
5374 - fix DLM error path that led to out-of-sync client, long delays (5779)
5375 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5376 - fix several locking issues related to i_size (5492,5624,5654,5672)
5377 - don't move pending lock onto export if it is already evicted (5683)
5378 - fix kernel oops when creating .foo in unlinked directory (5548)
5379 - fix deadlock in obdfilter statistics vs. object create (5811)
5380 - use time_{before,after} to avoid timer jiffies wrap (5882)
5381 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5382 - Do not leak request if server was not able to process it (5154)
5383 - If mds_open unable to find parent dir, make that negative lookup(5154)
5384 - don't create new directories with extent-mapping (5909, 5936)
5386 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5387 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5388 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5389 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5390 - support for RHEL4 kernel on i686 (5773)
5391 - provide error messages when incompatible logs are encountered (5898)
5393 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5394 * version 1.4.0.10 (1.4.1 release candidate 1)
5396 - don't keep a lock reference when lock is not granted (4238)
5397 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5398 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5399 - drop import inflight refcount on signal_completed_replay error (5255)
5400 - unlock page after async write error during send (3677)
5401 - handle missing objects in filter_preprw_read properly (5265)
5402 - no transno return for symlink open, don't save no-trasno open (3440)
5403 - don't try to complete elan receive that already failed (4012)
5404 - free RPC server reply state on error (5406)
5405 - clean up thread from ptlrpc_start_thread() on error (5160)
5406 - readahead could read extra page into cache that wasn't ejected (5388)
5407 - prevent races in class_attach/setup/cleanup/detach (5260)
5408 - don't dereference de->d_inode after l_dput of de (5458)
5409 - use "int" for stripe value returned from lock_to_stripe (5544)
5410 - mballoc allocation and error-checking fixes in 2.6 (5504)
5411 - block device patches to fix I/O request sizes in 2.6 (5482)
5412 - look up hostnames for IB nals (5602)
5413 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5414 - don't start multiple acceptors for the same port (5277)
5415 - fix incorrect LASSERT in mds_getattr_name (5635)
5416 - export a proc file for general "ping" checking (5628)
5417 - fix "lfs check" to not block when the MDS is down (5628)
5419 - service request history (4965)
5420 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5421 - create an "evict_client" /proc entry on OSTs, like the MDS has
5422 - fix mount usage message, return errors per mount(8) (5168)
5423 - change grep [] to grep "[]" in tests so they work in more UMLs
5424 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5425 - remove ancient LOV_MAGIC_V0 EA support (5047)
5426 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5427 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5428 - allow readahead allocations to fail when low on memory (5383)
5429 - mmap locking landed again, after considerable improvement (2828)
5430 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5432 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5435 - send OST transaction number in read/write reply to free req (4966)
5436 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5437 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5438 - the watchdog thread now runs as interruptible (5246)
5439 - flock/lockf fixes (but it's still disabled, pending 5135)
5440 - don't use EXT3 constants in llite code (5094)
5441 - memory shortage at startup could cause assertion (5176)
5443 - reorganization of lov code
5444 - single portals codebase
5446 - add extents/mballoc support (5025)
5447 - direct I/O reads in the obdfilter (4048)
5448 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5450 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5453 - send OST transaction number in read/write reply to free req (4966)
5454 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5455 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5456 - the watchdog thread now runs as interruptible (5246)
5457 - handle missing objects in filter_preprw_read properly (5265)
5458 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5459 - drop import inflight refcount on signal_completed_replay error (5255)
5460 - unlock page after async write error during send (3677)
5461 - return original error code on reconstructed replies (3761)
5462 - no transno return for symlink open, don't save no-trasno open (3440)
5464 - add pid to ldlm debugging output (4922)
5465 - bump the watchdog timeouts -- we can't handle 30sec yet
5466 - extra debugging for orphan dentry/inode bug (5259)
5468 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5471 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5472 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5473 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5474 - fix NULL dereference in /proc/sys/portals/routes (4827)
5475 - allow failed mdc_close() operations to be interrupted (4561)
5476 - stop precreate on OST before MDS would time out on it (4778)
5477 - don't send partial-page writes before EOF from client (4410)
5478 - discard client grant for sub-page writes on large-page clients (4520)
5479 - don't free dentries not owned by NFS code, check generation (4806)
5480 - fix lsm leak if mds_create_objects() fails (4801)
5481 - limit debug_daemon file size, always print CERROR messages (4789)
5482 - use transno after validating reply (3892)
5483 - process timed out requests if import state changes (3754)
5484 - update mtime on OST during writes, return in glimpse (4829)
5485 - add mkfsoptions to LDAP (4679)
5486 - use ->max_readahead method instead of zapping global ra (5039)
5487 - don't interrupt __l_wait_event() during strace
5489 - add software watchdogs to catch hung threads quickly (4941)
5490 - make lustrefs init script start after nfs is mounted
5491 - fix CWARN/ERROR duplication (4930)
5492 - return async write errors to application if possible (2248)
5493 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5494 - print NAL number in %x format (4645)
5495 - update barely-supported suse-2.4.21-171 series (4842)
5496 - support for sles 9 %post scripts
5497 - support for building 2.6 kernel-source packages
5498 - support for sles km_* packages
5500 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5503 - ignore -ENOENT errors in osc_destroy (3639)
5504 - notify osc create thread that OSC is being cleaned up (4600)
5505 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5506 - reconstruct ost_handle() like mds_handle() (4657)
5507 - create a new thread to do import eviction to avoid deadlock (3969)
5508 - let lconf resolve symlinked-to devices (4629)
5509 - don't unlink "objects" from directory with default EA (4554)
5510 - hold socknal file ref over connect in case target is down (4394)
5511 - allow more than 32000 subdirectories in a single directory (3244)
5512 - fix blocks count for O_DIRECT writes (3751)
5513 - OST returns ENOSPC from object create when no space left (4539)
5514 - don't send truncate RPC if file size isn't changing (4410)
5515 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5516 - bind to privileged port in socknal and tcpnal (3689)
5518 - rate limit CERROR/CWARN console message to avoid overload (4519)
5519 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5520 - basic mmap support (3918)
5521 - kernel patch series update from b1_4 (4711)
5523 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5526 - avoid crash during MDS cleanup with OST shut down (2775)
5527 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5528 - don't use bad inodes on the MDS (3744)
5529 - dynamic object preallocation to improve recovery speed (4236)
5530 - don't hold spinlock over lock dumping or change debug flags (4401)
5531 - don't zero obd_dev when it is force cleaned (3651)
5532 - print grants to console if they go negative (4431)
5533 - "lctl deactivate" will stop automatic recovery attempts (3406)
5534 - look for existing locks in ldlm_handle_enqueue() (3764)
5535 - don't resolve lock handle twice in recovery avoiding race (4401)
5536 - revalidate should check working dir is a directory (4134)
5538 - don't always mark "slow" obdfilter messages as errors (4418)
5540 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5543 - don't close LustreDB during write_conf until it is done (3860)
5544 - fix typo in lconf for_each_profile (3821)
5545 - allow dumping logs from multiple threads at one time (3820)
5546 - don't allow multiple threads in OSC recovery (3812)
5547 - fix debug_size parameters (3864)
5548 - fix mds_postrecov to initialize import for llog ctxt (3121)
5549 - replace config semaphore with spinlock (3306)
5550 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5551 - don't allow enqueue to complete on a destroyed export (3822)
5552 - down write_lock before checking llog header bitmap (3825)
5553 - recover from lock replay timeout (3764)
5554 - up llog sem before sending rpc (3652)
5555 - reduce ns lock hold times when setting kms (3267)
5556 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5557 - fix NULL deref and obd_dev leak on setup error (3312)
5558 - replace some LBUG about llog ops with error handling (3841)
5559 - don't match INVALID dentries from d_lookup and spin (3784)
5560 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5561 - fix invalid assertion in ptlrpc_set_wait (3880)
5563 - add libwrap support for the TCP acceptor (3996)
5564 - add /proc/sys/portals/routes for non-root route listing (3994)
5565 - allow setting MDS UUID in .xml (2580)
5566 - print the stack of a process that LBUGs (4228)
5568 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5571 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5572 - make sure to unset replay flag from failed open requests (3440)
5573 - if default stripe count is 0, use OST count for inode size (3636)
5574 - update parent mtime/ctime on client for create/unlink (2611)
5575 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5576 - free recovery state on server during a forced cleanup (3571)
5577 - unregister_reply for resent reqs (3063)
5578 - loop back devices mounting and status check on 2.6 (3563)
5579 - fix resource-creation race that can provoke i_size == 0 (3513)
5580 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5581 - more debugging for page-accounting assertion (3746)
5582 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5583 - avoid deadlock after precreation failure (3758)
5584 - fix race and lock order deadlock in orphan handling (3450, 3750)
5585 - add validity checks when grabbing inodes from l_ast_data (3599)
5587 - add /proc/.../recovery_status to obdfilter (3428)
5588 - lightweight CDEBUG infrastructure, debug daemon (3668)
5589 - change default OSC RPC parameters to be better on small clusters
5590 - turn off OST read cache for files smaller than 32MB
5591 - install man pages and include them in rpms (3100)
5592 - add new init script for (un)mounting lustre filesystems (2593)
5593 - run chkconfig in %post for init scripts (3701)
5594 - drop scimac NAL (unmaintained)
5596 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5599 - clean kiobufs before and after use (3485)
5600 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5601 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5602 - change init script to fail more clearly if not run as root (1528)
5603 - allow clients to reconnect during replay (1742)
5604 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5605 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5606 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5607 - fix cond_resched (3554)
5608 - only evict unfinished clients after recovery (3515)
5609 - allow bulk resend, prevent data loss (3570)
5610 - dynamic ptlrpc request buffer allocation (2102)
5611 - don't allow unlinking open directory if it isn't empty (2904)
5612 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5613 - remove extraneous obd dereference causing LASSERT failure (3334)
5614 - don't use get_cycles() when creating temp. files on the mds (3156)
5615 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5616 - handle EEXIST for set-stripe, set proper directory name (3336)
5618 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5619 - fix ksocknal_fmb_callback() error messages (2918)
5621 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5624 - don't copy lvb into (possibly NULL) reply on error (2983)
5625 - don't deref dentry after dput, don't free lvb on error (2922)
5626 - use the kms to determine writeback rpc length (2947)
5627 - increment oti_logcookies when osc is inactive (2948)
5628 - update client's i_blocks count via lvb messages (2543)
5629 - handle intent open/close of special files properly (1557)
5630 - mount MDS with errors=remount-ro, like obdfilter (2009)
5631 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5632 - don't use cancelling-locks' kms values (2947)
5633 - use highest lock extent for kms, not last one (2925)
5634 - don't dereference ERR_PTR() dentry in error handling path (3107)
5635 - fix thread race in portals_debug_dumplog() (3122)
5636 - create lprocfs device entries at setup instead of at attach (1519)
5637 - common AST error handler, don't evict client on completion race (3145)
5638 - zero nameidata in detach_mnt in 2.6 (3118)
5639 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5640 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5641 - initialize RPC timeout timer earlier for 2.6 (3219)
5642 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5643 - print nal/nid for unknown nid (3258)
5644 - additional checks for oscc recovery before doing precreate (3284)
5645 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5646 - don't crash in mdc_close for bad permissions on open (3285)
5647 - zero i_rdev for non-device files (3147)
5648 - clear page->private before handing to FS, better assertion (3119)
5649 - tune the read pipeline (3236)
5650 - fix incorrect decref of invalidated dentry (2350)
5651 - provide read-ahead stats and refine rpc in flight stats (3328)
5652 - don't hold journal transaction open across create RPC (3313)
5653 - update atime on MDS at close time (3265)
5654 - close LDAP connection when recovering to avoid server load (3315)
5655 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5656 - don't leak open file on MDS after open resend (3325)
5657 - serialize filter_precreate and filter_destroy_precreated (3329)
5658 - loop device shouldn't call sync_dev() for nul device (3092)
5659 - clear page cache after eviction (2766)
5660 - resynchronize MDS->OST in background (2824)
5661 - refuse to mount the same filesystem twice on same mountpoint (3394)
5662 - allow llmount to create routes for mounting behind routers (3320)
5663 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5664 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5665 - fix rare problem with rename on htree directories (3417)
5667 - allow default OST striping configuration per directory (1414)
5668 - fix compilation for qswnal for 2.6 kernels (3125)
5669 - increase maximum number of MDS request buffers for large systems
5670 - change liblustreapi to be useful for external progs like lfsck (3098)
5671 - increase local configuration timeout for slow disks (3353)
5672 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5674 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5677 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5678 - don't overwrite extent policy data in reply if lock was blocked (2901)
5679 - drop filter export grants atomically with removal from device (2663)
5680 - del obd_self_export from work_list in class_disconnect_exports (2908)
5681 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5682 - swab reply message in mdc_close, other PPC fixes (2464)
5683 - fix destroying of named logs (2325)
5684 - overwrite old logs when running lconf --write_conf (2264)
5685 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5686 - fix race in target_handle_connect (2898)
5687 - mds_reint_create() should take same inode create lock (2926)
5688 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5689 - don't close files for self_export to avoid uninitialized obd (2936)
5690 - allow MDS with the same name as client node (2939)
5691 - hold dentry reference for closed log files for unlink (2325)
5692 - reserve space for all logs during transactions (2059)
5693 - don't evict page beyond end of stripe extent (2925)
5694 - don't oops on a deleted current working directory (2399)
5695 - handle hard links to targets without a parent properly (2517)
5696 - don't dereference NULL lock when racing during eviction (2867)
5697 - don't grow lock extents when lots of conflicting locks (2919)
5699 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5702 - account for cache space usage on clients to avoid data loss (974)
5703 - lfsck support in lustre kernel code (2349)
5704 - reduce journal credits needed for BRW writes (2370)
5705 - orphan handling to avoid losing space on client/server crashes
5706 - ptlrpcd can be blocked, stopping ALL progress (2477)
5707 - use lock value blocks to assist in proper KMS, faster stat (1021)
5708 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5709 - recovery for initial connections (2355)
5710 - fixes for mds_cleanup_orphans (1934)
5711 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5712 - block all file creations until orphan recovery completes (1901)
5713 - client remove rq_connection from request struct (2423)
5714 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5715 - recovery timer race (2670)
5716 - mdc_close recovey bug (2532)
5717 - ptlrpc cleanup bug (2710)
5718 - mds timeout on local locks (2588)
5719 - namespace lock held during RPCs (2431)
5720 - handle interrupted sync write properly (2503)
5721 - don't try to handle a message that hasn't been replied to (2699)
5722 - client assert failure during cleanup after abort recovery (2701)
5723 - leak mdc device after failed mount (2712)
5724 - ptlrpc_check_set allows timedout requests to complete (2714)
5725 - wait for inflight reqs when ptlrpcd finishes (2710)
5726 - make sure unregistered services are removed from the srv_list
5727 - reset bulk XID's when resending them (caught by 1138 test)
5728 - unregister_bulk after timeout
5729 - fix lconf error (2694)
5730 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5731 - readahead locks pages, leaves pending causing memory pressure (2673)
5732 - increase OST request buffers to 4096 on large machines (2729)
5733 - fix up permission of existing directories in simple_mkdir (2661)
5734 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5735 - don't assign transno to errored transactions (2742)
5736 - don't delete objects on OST if given a bogus objid from MDS (2751)
5737 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5738 - if rq_no_resend, then timeout request after recovery (2432)
5739 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5740 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5741 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5742 - MDS recovery shouldn't delete objects that it has given out (2730)
5743 - if enqueue arrives after completion, don't clobber LVB (2819)
5744 - don't unlock pages twice when trigger_group_io returns error (2814)
5745 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5746 - don't write pages to disk if there was an error (1450)
5747 - don't ping imports that have recovery disabled (2676)
5748 - take buffered bytes into account when balancing socknal conn (2817)
5749 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5750 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5751 - remove little-endian swabbing of llog records (1987)
5752 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5753 - retry reposting request buffers if they fail (1191)
5754 - grow extent at grant time to avoid granting a revoked lock (2809)
5755 - lock revoke doesn't evict page if covered by a second lock (2765)
5756 - disable VM readahead to avoid reading outside lock extents (2805)
5758 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5759 - updated kernel patches for hp-2.4.20 kernel (2681)
5761 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5764 - fix truncated write corruption (2366)
5765 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5767 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5768 - ensure len(uuid) < 37 in lmc (1171)
5769 - fix ia64 OOPS in llog_test (2255)
5770 - zero end of page at obdfilter for partial page writes (2648)
5771 - don't leave stale dentries around after renames (bug 2428)
5772 - fix timeouts when evicting a client with a single lock held (2642)
5773 - set deadline for the initial HELLO message to drain (2634)
5774 - print out dotted-quad IP addresses in the socknal (2302)
5776 - additional debugging for MDS client eviction problem (2443)
5777 - fix mkfsoptions support for osts (2603, 2604)
5779 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5782 - add series for the vanilla 2.6.0 kernel
5783 - add series for the vanilla 2.4.24 kernel
5784 - add series for a cray x86/64 UL kernel drop
5785 - fix xattr patches for the vanilla 2.4.19 series
5787 - generate true UUIDs in lmc (1171)
5788 - have portals stack dumping break in UML (2466)
5789 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5790 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5791 - wake callback waiting threads on client eviction (2460)
5792 - Add --ptldebug and --subsystem to lmc (1719)
5793 - update assertion to allow safe interrupt allocation
5794 - set rq_no_resend for cancel requests (2432)
5795 - recalculate ptlrpcd timeout after resend (2494)
5796 - call vfs_rmdir when removing pending directories (2368)
5797 - fix renaming a file to itself (2429)
5798 - lmc creates a default one-stripe lov (2454)
5799 - expand procfs space to handle large clusters (2326)
5800 - increase UML stack to avoid overflow
5801 - update lconf's list of debug and subsystem masks
5802 - fix lfs find --obd (2510)
5803 - /proc tunable for disabling filter read caching (2591)
5804 - stop rpm packages from altering slapd.conf (2301)
5805 - disable nagle in the socknal under 0conf (2578)
5806 - choose mds inode size based on stripe count (2572)
5807 - fix kernel-source rpm problems (2516)
5809 - add --disable-doc to avoid pdf generation (2421)
5810 - update documentation, tests, type-os, comments
5811 - avoid format warnings on ia64
5812 - remove the TOE NAL
5813 - tiny code cleanups by removing unused fields
5815 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5818 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5819 - remove the most blatant lies from BUILDING file (2371)
5820 - change default debug level to reasonable production setting
5821 - reduce client side cache size to reduce cache flush time
5822 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5823 - make TCP zerocopy and pinger support enabled by default (2476)
5824 - sync writes completed after process exits caused crashes (2319)
5825 - maintain correct mount count on the MDS (2356)
5826 - backout 1557, because 2316 wasn't really fixed
5827 - better file I/O statistics gathering in /proc
5828 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5829 - another kernel patch to fix zero-copy TCP function export
5830 - don't take duplicate lock when processing re-sent getattr (2420)
5831 - lctl uses obd_self_export instead of creating new conn (2353)
5832 - MDS/OST recovery case which requires object creation asserted (2425)
5833 - move lfs from /usr/sbin to /usr/bin in packages
5834 - fix race between mds_client_add and mds_client_free (2417)
5835 - use kmalloc instead of slabs in portals (2430)
5836 - don't create duplicate records when a failover MDS is present (2442)
5837 - remove unnecessary mount age check (2332)
5838 - don't remove directory inodes from locks prematurely (2451)
5839 - don't break if MDS service name is the same as hostname (2103)
5840 - fix races in client write RPC generation when cache full (2482)
5842 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5845 - remove now-unused request->rq_obd (278)
5846 - if an allocation fails, print out how much memory we've used (1933)
5847 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5848 - add the "configurable stack size" patch to most series files (1256)
5849 - ability to write large log records, for 100+ OST configs (2306)
5850 - fix NULL deref when filter_prep fails (2314)
5851 - fix operator precedence error in filter_sync
5852 - dynamic allocation of socknal TX descriptors (2315)
5853 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5854 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5855 - ensure that CWARN messages in Portals always get to the syslog
5856 - __init/__exit are not for prototype decls (ldlm_init/exit)
5857 - x86-64 compile warning fixes
5858 - fix gateway LMC keyword conflict (2318)
5859 - fix MDS lock inversions in getattr/reint paths (1844)
5860 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5861 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5862 - DLM race condition prevented some lock evictions (2328)
5863 - ENOMEM detection and retry on socknal sends (2230)
5864 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5865 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5866 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5867 - better error messages when a client is rejected during recovery (1505)
5868 - avoid cancelling locks which were never granted, after failure (2330)
5869 - fix i_sem/journal inversion in mds_client_add (2333)
5870 - fix truncate/getattr lock cycle deadlock (2334)
5871 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5872 - fix two rare exit paths which could leak an l_lock() ref (2321)
5873 - fencepost error in MDS/OST orphan recovery (2226)
5874 - make log record alignment 8 bytes (1988)
5875 - lstripe now fails when requested offset > ost_count (2237)
5876 - ensure that all kernel series have a complete list.h (1607)
5877 - fix crashes in special-file operations (2316)
5878 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5880 - allow configurable automake binary, for testing new versions
5881 - small update to the lfs documentation
5883 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5885 * fix negative export reference count in fsfilt_sync (2312)
5887 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5888 * release candidate 0.9.1
5890 - orphans are moved into the PENDING directory for possible recovery
5891 - replayed opens now open by fid for orphan/rename safety (1042)
5892 - last close of an orphan inode generates a transno (683)
5893 - chdir() and mount() now pin the directory entry (1020)
5894 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5895 - discard very old requests without processing them (1502)
5896 - remove some common, well-understood CERRORs (1505)
5897 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5898 - clear "grant" flags in OST replies until OST grant code lands (1644)
5899 - fix read performance by not clobbering i_blksize on client (1598)
5900 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5901 - make LRU size a /proc tunable, clears locks when reduced (707)
5902 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5903 - server locks take references on exports now (1558)
5904 - build fixes for 2.4.20-rh trees (1663)
5905 - return an error from lov_create if all OSCs are inactive (1751)
5906 - fix import levels when a reconnect happens without a timeout (1597)
5907 - exit early from mds_open if we get a lookup error (1749)
5908 - partial page read at EOF wouldn't wait for disk before sending (1642)
5909 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5910 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5911 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5912 - write/truncate lock inversion (1639)
5913 - Don't auto-load obdclass, portals modules during cleanup (1495)
5914 - fix timestamps from jumping to "now" (1763)
5915 - extra journal assertions (1648)
5916 - add an extra multiunlink test (1771)
5917 - fix read_record/write_record API (1776)
5918 - fix leak of offset_extent, possible incorrect i_size later (1772)
5919 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5920 - Debugging for the kqswnal_get_idle_tx problems (1820)
5921 - Allow recovery to be attempted multiple times (1536)
5922 - Write out MDS last_rcvd file after it is first created (1600)
5923 - Fix tx_descriptor leak in failed transmit situations (1827)
5924 - ext3 journaling fixes for assertion failure after IO error (1871)
5925 - class_export_put() on freed export after completion AST error (1896)
5926 - Fix revalidate looping in VFS (1322)
5927 - Don't access a freed export during MDS_REINT timeout (1521)
5928 - Add open-unlink recovery support on the MDS (1673,1764)
5929 - Return an error if no MDS data was read from last_rcvd (1946)
5930 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5931 - Don't setup a disk device that doesn't match exported UUID (317)
5932 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5933 - avoid committing NULL handle in force close
5934 - local.sh is now a one-stripe LOV configuration
5935 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5936 - fix ext3 htree duplicate directory entry corruption (1516)
5937 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5938 - update to new LOV EA format (2097)
5939 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5941 2003-06-15 Phil Schwan <phil@clusterfs.com>
5944 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5945 - per-import/export recovery handling (958, 931, 959)
5946 - multiple last-rcvd slots, for serving multiple FSes (949)
5947 - connections are again shared between multiple imp/exports (963, 964)
5948 - "umount -f" would hang if any requests needed to be sent (393, 978)
5949 - avoid pinning large req buffer by copying for queued messages (989)
5950 - add "uuid" to "lctl device" command to help upcalls (991)
5951 - "open" RPCs with transnos would confuse recovery counters (1037)
5952 - do proper endian conversion of all wire messages (288, 340, 891)
5953 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5954 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5955 - fix ptlrpc_connection leak in target_handle_connect (1174)
5956 - fix import refcounting bug in OST and MDS cleanup (1134)
5957 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5958 - fix very unlikely obd_types race condition (501)
5959 - remove osc_open hack for echo_client (1187)
5960 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5961 - a failure in read_inode2 leads to deadlock (1139)
5962 - cancel ack-locks as soon as transaction is committed (1072)
5963 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5964 - make sure to commitrw after any preprw to avoid deadlock (1162)
5965 - failing to execute a file in a lustre FS would lock inode (1203)
5966 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5967 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5968 - obd_uuid2tgt would walk off the end of the list (1255)
5969 - on IA64 the osc would give portals incorrect bulk size (1258)
5970 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5971 - fix lock inversion caused by new llite matching code (1282)
5972 - limit the number of dirty pages on a client to 10MB (1286)
5973 - timed out locks were not being corrected cancelled (1289)
5974 - fix O_DIRECT above 4GB on IA-32 (1292)
5975 * major user-visible changes
5976 - fail out/fail over policy now controlled by the upcall (993)
5978 - add OBD_PING to check server availability and failure (954)
5979 - lustre messages are now sent in sending host order (288, 340, 891)
5980 - add eadatalen to MDS getattr reply (340)
5981 - OST read replies may contain second buffer, with per-page status (593)
5983 2003-03-11 Phil Schwan <phil@clusterfs.com>
5986 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5987 - failed open()s could cause deadlock; fixed (867, 869)
5988 - stop cancelling OST locks when files are closed (481)
5989 - overlapping XID spaces caused network corruption (851, 853)
5990 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5991 - setattr_raw added, to do single-RPC, server-side setattrs
5992 - lmc/lconf syntax change for OST UUIDs
5993 - fix crashy race condition between ptlrpc_free_req and osc_close
5994 - don't use request in mdc_enqueue if we hit a timeout (889)
5995 - don't set the inode i_size for regular files from the MDS (896)
5996 - handle out of order completion AST (842)
5997 - don't LBUG if a lock request times out after receiving AST (913)
5998 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5999 - if a bad lock AST arrives, send an error instead of dropping entirely
6000 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6001 - fix leak in bulk IO when only partially completed (899, 900, 926)
6002 - fix O_DIRECT for ia64 (55)
6003 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6004 - C-z after timeout could hang a process forever; fixed (977)
6006 - client-side I/O cache (678, 924, 929, 941, 970)
6008 - READPAGE and SETATTRs which don't take server-side locks get
6011 2003-02-11 Phil Schwan <phil@clusterfs.com>
6014 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6015 - Fix fsfilt_extN_readpage() to read a full page of directory
6016 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6017 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6018 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6019 - Most intent operations were removed, in favour of a new RPC mode
6020 that does a single RPC to the server and bypasses most of the VFS
6021 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6022 - Aggressively cancel local locks on DLM servers
6023 - mds_reint_unlink sends EA to the client if it's the last nlink.
6024 client uses that EA to unlink OST objects.
6025 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6026 - recursive symlinks were fixed (439)
6027 - fixed NULL deref in DEBUG_REQ
6028 - filter_update_lastobjid no longer calls sync, which annoyed extN
6029 - fixed multi-client small-writes to a single file problem (445)
6030 - fixed mtime updates during file writes (607)
6031 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6032 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6033 - fixed Linux OST size reporting problem (444, 656)
6034 - OST now updates object mtime with writes or setattr (607, 619)
6035 - client verifies file size before zeroing page past EOF (445)
6036 - OST now writes last allocated objid to disk with allocation (108)
6037 - LOV on echo now works (409)
6039 - mds_reint_unlink sends a new buffer, with the EA included. this
6040 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6041 set if a regular file was being unlinked, and it was the last link
6042 - use PtlGet from the target for bulk writes (315)
6043 - OST now updates object mtime with writes or setattr (607, 619)
6044 - LDLM now has a grant-time callback to revalidate locked items, if
6046 - Many MDS operations were reorganized to combat race conditions
6048 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6049 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6051 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6054 - Fully reactivate OST imports after reconnection (512, others)
6055 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6056 - More graceful error handling for truncating on dead OST (515)
6057 - Don't error out unless we're actually accessing dead stripes (474)
6058 - Fix garbage sizes when stripes are missing (410)
6059 - LRU counters were broken, causing constant lock purge (433, 432)
6060 - garbage on read from stripes with failed OSTs (441)
6061 - mark OSCs as active before reconnecting during recovery (438)
6062 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6063 - lfind did not preserve OST order in output (443)
6064 - symlinks cause hung clients, incorrect data (439)
6065 - stop dereferencing request after dropping refcount (457)
6066 - don't LASSERT(spin_is_locked) on non-SMP (455)
6067 - fixes for many rename() bugs
6068 - fstat didn't correctly synchronize attributes (399)
6069 - server must handle lock cancellation during blocking AST prep (487)
6070 - bulk descriptors were free()d too soon (511)
6071 - fix paths in lconf, which would load incorrect modules (451, 507)
6072 - fix confusing lconf 'host not found' error message (386)
6073 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6074 - fix race condition in mdc_blocking_ast() for inode access (526)
6075 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6076 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6077 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6078 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6079 - fix lstripe in several ways (kernel side) (527)
6080 - fix request leak in ldlm_cli_enqueue (262)
6081 - incorrect OSC was marked inactive after OST failure
6082 - call mds_fs_cleanup before unmounting filesystem (524)
6083 - fix races between taking ns_lock and ldlm_lock_change_resource
6084 - fix races updating LOV export open file list
6085 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6086 - fix recovery NULL deref in ldlm_cli_cancel_unused
6087 - fix some DLM races by using new hash table for lock handles (419)
6088 - permit the client to specify desired inodes, at replay
6089 - duplicate requests when we queue them for replay reintegration
6090 - fix last_rcvd offset calculation
6091 - sync after each recovered transaction, so we always make progress
6092 - never, not always, ERESTART requests without transnos
6093 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6095 - skip replay if the MDS says that the client is already connected
6096 - don't check for a recovery-enabled export to match lctl's UUID
6097 - don't INC_USE_COUNT for phantom exports
6098 - don't crash when cleaning up phantom exports (567)
6099 - don't double-finish or set replay data for errored mdc_open requests
6100 - abort requests when they time out, so we don't get old replies
6101 - send/receive replies for AST messages again
6102 - if the client says that it doesn't have the lock, cancel it on the
6104 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6105 mark it as destroyed, it will all work out when decref is called
6106 - fix module use counts (22, 581)
6108 - ASTs now expect a reply (server cancels lock on error reply)
6110 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6113 - fix many simultaneous client startup (392)
6114 - fix dentry->d_it clobbering
6115 - credentials weren't being shipped for readdir/getattr operations
6116 - remove invalid assertions triggered during some concurrent MD
6118 - proper Lustre versions added (336, 389)
6119 - fix memory leak for create error case (398)
6120 - fix LOV locking bug that would get cli/srv out of sync
6121 - fix echo client over LOV (409)
6122 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6123 - fix double-O_EXCL intent crash (424)
6124 - avoid sending multiple lock CANCELs (352)
6126 - MDS can do multi-client recovery (modulo bugs in new code)
6128 - many updates, edits, cleanups
6130 2002-11-18 Phil Schwan <phil@clusterfs.com>
6133 - fix null d_it dereference (346)
6134 - fix full OST/dbench hang (333)
6135 - fix permission problem with file removal (286)
6136 - fix removal of OSCs from LOV when they fail
6137 - fix NULL deref during bulk timeout (214)
6138 - fix problems related to multiple filesystems on one MDS (241)
6139 - fixed serious subtle metadata locking bugs
6140 - free locks on clients when inodes are removed due to memory
6142 - fix inode pointer in lock data (285)
6143 - partial support for multiple MDS on a single host (241)
6144 - data locks weren't cancelled at clear_inode time (290, 311)
6145 - intent locks could lead to unbounded lock growth (205)
6146 - added a maximum lock count, an LRU list, and a flusher
6147 - fix multiple rename (365)
6148 - properly abstracted the echo client
6149 - OSC locked 1 byte too many; fixed
6150 - rewrote brw callback code:
6151 - fixed recovery bugs related to LOVs (306)
6152 - fixed too-many-pages-in-one-write crash (191)
6153 - fixed (again) crash in sync_io_timeout (214)
6154 - probably fixed callback-related race (385)
6156 - Add capability to MDS protocol
6157 - LDLM cancellations and callbacks on different portals
6159 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6162 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6163 - defer open object destruction to close time (601981, 138)
6164 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6165 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6166 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6167 - don't call obd_setattr in ll_file_release for destroyed objects
6168 * protocol change to lustre_msg: move |version| and add |flags|
6169 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6170 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6171 * added replay of create, unlink, link and rename operations during
6172 MDS failover; recovery should be much more robust now
6173 * remove failed OSCs from LOVs (only lov_create uses this so far)
6174 * the lustre-HOWTO was brought (more) up to date (582544)
6176 2002-10-23 Phil Schwan <phil@clusterfs.com>
6179 - in-use dentries weren't being reused properly (617851)
6180 - prevent multiple LDLM setup (599178)
6181 - fix LOV size calculations for truncate (617853)
6182 - fix client handling of MDS intent errors (POSIX)
6183 - fix permission bug in lovstripe.c test (624321)
6184 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6185 - truncate past end of file could corrupt data
6186 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6187 - a race in recovery could return ETIMEDOUT to apps (623947)
6188 - building outside the source directory was fixed
6189 * the lustre-HOWTO was brought (more) up to date (582544)
6190 * major progress was made on recovery functionality
6192 2002-10-10 Phil Schwan <phil@clusterfs.com>
6195 - recovery deadlock fix
6196 - rm -rf causes LBUG fix (617817)
6197 - file open by multiple tasks fix (618962)
6198 - directory permissions bugs (602707 and 620007)
6199 - journal_stop fixed with locking (611313)
6200 - O_APPEND failures resolved (618273, perhaps 614459)
6201 - lconf PATH fix (619770)
6202 - IA64 build fix (621450)
6203 - RPC buffer sizes scale with amount of memory
6205 2002-10-01 Phil Schwan <phil@clusterfs.com>
6208 - locks would be cancelled without throwing away data pages,
6209 resulting in inconsistent data (605627)
6210 - inode attributes were not always being refreshed (605627, 612449)
6211 - lconf now continues to cleanup after lctl reports an error
6212 - MDS now enforces user permissions (602707)
6213 - lprocfs cleanup fixed, but not yet enabled (614157)
6214 - fixed infinite server hang, should a client not respond to an AST
6215 - avoid going into recovery if user calls readlink() with a buffer
6216 that's too small (613941)
6217 - AST RPCs no longer require replies (614867) -- this may be changed
6218 - don't crash server if client sends an IOV that's too big (611336)
6219 - fixed lock conversion deadlock (611892)
6220 - fixed the following of symlinks (614622)
6221 * recovery: the server can remove locks from a client that dies, other
6222 clients can make progress
6223 * more extN patch fixes
6224 * compile-time configurable ptlrpc buffer allocations
6226 - collaborative read cache document
6227 - Lustre Lite Performance CDR document-in-progress
6229 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6232 - fix typo in patch-2.4.18
6234 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6237 - clear ptlrpc request each time in handle_incoming_request()
6238 - unlink of files now destroys the object on the OST
6240 2002-09-19 Peter Braam <braam@clusterfs.com>
6242 * add hard link support
6243 * change obdfile creation method
6244 * kernel patch changed
6246 2002-09-19 Peter Braam <braam@clusterfs.com>
6249 - stack overflow bug in extN fixed
6251 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6253 * documentation updates
6254 - add man pages for config tools
6255 - update tests/README to describe testing with new config tools
6256 - finish metadata API descriptions
6257 * bug fixes and cleanups
6258 - statfs workaround for 16TB limit
6259 - LOV stripe allocation improved, can stripe on subset of OSTs
6260 - LOV file size/IO offset was wrong for files > 4GB in size
6261 - object EA data was being dropped, caused files to be unreadable
6262 - memory overflow with non-LOV OST caused memory corruption
6263 - fixed regression tests to work with new config tools, obdfilter
6264 - fixed bug when directory size became larger than 1 block
6265 - fixed bug (for single client case) when PWD was deleted
6266 - invalidate local directory pages when doing intent-based ops
6267 - avoid LDLM oops when lock callback contained bad data
6269 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6271 * documentation updates
6272 * bug fixes and cleanups
6273 - configuration tools
6276 - 64-bit compile warnings
6277 - 64-bit internal statfs data
6279 * test_brw on persistent OST devices
6281 * lprocfs (disabled)
6283 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6285 * documentation updates
6286 * bug fixes and cleanups
6287 * configuration tools
6289 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6292 * many small fixes to 0_5_4
6293 * io/network handling
6294 * thinkos in MDS operations
6296 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6299 * crucial basic fixes to 0.5.3
6300 * IOR, Iozone work over Elan
6303 2002-08-07 Phil Schwan <phil@clusterfs.com>
6304 * version 0_5_3, our first alpha
6305 * we use the new Portals iovs
6306 * documentation updates
6307 * bug fixes and cleanups
6308 * small changes in the DLM wire protocol
6310 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6311 * version 0_5_1 with some initial stability,
6312 * locking on MD and file I/O.
6313 * documentation updates
6314 * several bug fixes since 0.5.0
6315 * small changes in wire protocol
6317 2002-07-18 Phil Schwan <phil@clusterfs.com>
6319 * delivered as Lustre Light Alpha
6320 * fixed a crash after handling invalid MDS requests
6321 * fixed directory pages for architectures with non-4k pages sizes
6323 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6324 * release version v0_4_4
6325 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6327 * Fixes a number of interruption problems with OST operations.
6328 * Update documentation for portals header changes
6329 * Move all wire protocol structs/defines to lustre_idl.h
6330 * Fixes symlink length bug.
6331 * Add tcpdump to repository.
6333 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6334 * release version v0_4_3
6335 * Fixes statfs for inodes on extN.
6336 * Fixes bug in runtests which would delete /etc/hosts.
6337 * Use 64-bit object IDs wherever possible (not into VFS though)
6338 Remove ost_get_info, which is unused by lustre, and out of date.
6340 2002-07-03 Peter Braam <braam@clusterfs.com>
6341 * release version v0_4_2 Fixes a lookup error (type not passed)
6342 * move forward to head of Portals
6343 * move forward to latest Lustre kernel
6345 2002-06-25 Peter Braam <braam@clusterfs.com>
6346 * release version v0_4_1. Hopefully stable on single node use.