3 * Server support for kernels:
4 2.6.18-238.19.1.el5 (RHEL5)
5 2.6.32-131.6.1.el6 (RHEL6)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-238.19.1.el5 (RHEL5)
9 2.6.32-131.6.1.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 --------------------------------------------------------------------------------
22 09-16-2011 Whamcloud, Inc.
24 * Server support for kernels:
25 2.6.18-238.19.1.el5 (RHEL5)
26 2.6.32-131.6.1.el6 (RHEL6)
27 * Client support for unpatched kernels:
28 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
29 2.6.18-238.19.1.el5 (RHEL5)
30 2.6.32-131.6.1.el6 (RHEL6)
31 2.6.29.4-167.fc11 (FC11)
32 2.6.16.60-0.69.1 (SLES10)
34 * Recommended e2fsprogs version: 1.41.90.wc3
35 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
36 Lustre filesystem with 4K stack may cause a stack overflow. For more
37 information, please refer to bugzilla 17630.
38 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
39 mounted with "32bitapi" mount option
41 Severity : enhancement
42 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
44 Severity : enhancement
45 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
47 Severity : enhancement
48 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
51 Frequency : only with SLES10
53 Description: Use OFED "KMP" provided by Novell
54 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
55 Lustre is now built against this package. That means you need to
56 install the ofed-kmp package from Novell for the patchless client.
57 Note that the ofed-kmp that Novell ships may not exactly match the
58 kernel version but should still be compatible.
60 Severity : enhancement
62 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
63 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
65 --------------------------------------------------------------------------------
67 2010-07-15 Oracle, Inc.
69 * Server support for kernels:
70 2.6.18-164.11.1.el5 (RHEL 5)
71 2.6.18-164.11.1.0.1.el5 (OEL 5)
72 * Client support for unpatched kernels:
73 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
74 2.6.18-164.11.1.el5 (RHEL 5),
75 2.6.18-164.11.1.0.1.el5 (OEL 5)
76 2.6.16.60-0.42.8 (SLES 10),
78 2.6.29.4-167.fc11 (FC11)
79 * Recommended e2fsprogs version: 1.41.10-sun2
81 Severity : enhancement
83 Description: Distribute lfs_migrate script with lustre
84 Details : The lfs_migrate mentioned in the manual is now included in the
85 lustre tree in lustre/scripts and will distributed in lustre RPMs.
87 Severity : enhancement
89 Description: Failover nid list added to import proc file.
91 Severity : enhancement
93 Description: "som_preview" mount option required to use SOM
94 Details : The clients will need to use som_preview mount option
98 Description: Lustre does not do 1MB IOs to HW RAID
99 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
101 Severity : enhancement
103 Description: lfs quota output is not very convenient for awk/sed-parsing
104 Details : Some positions in lfs quota output table could be empty or
105 non-empty which made it hard to parse it with scripts, now
106 a dash is put instead of space where there is not supposed
111 Description: Update to OFED 1.5.1
112 Details : OFED 1.5.1 will be used in environments that don't use
113 vendor supplied OFED.
117 Description: Remove lnet. prefix from lctl params display
118 Details : Also renamed lnet.memused to lnet_memused to
123 Description: Use conf_param -d to remove permanant settings.
127 Description: Add list_param -R
128 Details : lctl list_param now supports recursive mode.
131 Bugzilla : 21829, 22312
132 Description: Fix llobdstat.
133 Details : 1. Now updating the write_bytes counter
134 2. Port script fixes from 1.8.
138 Description: decrease the usage of memory on clients.
139 Details : 1. On clients, recycle dentries and inodes unused.
140 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
145 Description: llobdstat fix and enhancement
146 Details : add a counter to set a limit to how many samples will be returned
147 fix a wildcard in the path to limit to obdfilter stats only
151 Description: lfs check only allowed for root
152 Details : Code cleanups and added a regresion test of non-root use of
157 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
158 Details : Fix hr->hr_index overload.
162 Description: Fix FIEMAP functionality.
166 Description: FC11 patchless client support added
170 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
171 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
175 Description: New acc-sm script and various test-framework enhancments.
179 Description: Previously landed patch was reverted.
183 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
187 Description: Add ldiskfs maxdirsize mount option.
191 Description: Fix mkfs.lustre support for 16TB LUNs.
195 Description: Print mmp_check_interval
196 Details : print mmp_check_interval and make it possible to abort
197 mount operation in case it takes too long.
201 Description: Add cascading_rw to lustre mpi tests.
205 Description: Store SOM data in EA.
206 Details : Also several SOM related cleanups.
210 Description: Add refcount to md_open_data
214 Description: Use refcount on lov_request_set to prevent memory corruption.
218 Description: Statahead: use parent dir's i_mutex for synchronization.
222 Description: loadgen improvements.
226 Description: Enable peer health detection in ksocklnd and ko2iblnd.
230 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
234 Description: Fix race on obd_devcie:md_stats
238 Description: Fix SLES 11 client builds.
239 Details : Resolved uninitalized pointer gcc warnings.
243 Description: Allow non-root access for "lfs check".
244 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
248 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
250 Severity : enhancement
252 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
253 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
257 Description: Sanity test_27p fails.
258 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
259 Sanity tests 27 were cleaned up and improved.
263 Description: Unify req format on client/servers
264 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
265 buffers and swabbers approach..
270 Description: drop join file stuff.
272 Severity : enhancement
274 Description: Add 2.6.27 and SLES11 (patchless client) support
278 Description: Fix in ptlrpc_expire_one_request() to print the signed time
281 Severity : enhancement
283 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
286 Severity : enhancement
288 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
290 Severity : enhancement
292 Description: correctly handle big reply message.
293 Details : send LNet event if reply is bigger then buffer and adjust this buffer
298 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
300 Severity : enhancement
302 Description: Add support for OEL5.
306 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
308 Severity : enhancement
310 Description: "weak-modules" support
311 Details : Implement "weak-modules" support which enables kernel modules
312 to be used with any kernel that implements the same kABI. In
313 order to achieve this modules are now installed in
314 /lib/modules/$(uname -r)/updates/kernel on all distributions.
318 Description: client could not reconnect to OST because of an active request.
319 Details : The client now retries to reconnect to the same server, if a
320 connect request failed with EBUSY or -EAGAIN. Test case 35b
321 conf-sanity.sh has been created.
325 Description: Don't allow make backward step on assignin osc next id.
326 Details : race between allocation next id and ll_sync thread can be cause
327 of set wrong osc next id and can be kill valid ost objects.
331 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
335 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
339 Description: Error handling in osc_statfs_interpret() has been improved.
340 Details : Check in osc_statfs_interpret() for EBADR.
344 Description: Do not send statfs() requests to OSTs disabled by administrator.
345 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
347 Severity : enhancement
349 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
351 Severity : enhancement
353 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
355 Severity : enhancement
357 Description: Enhancement for lfs(1) command to use numeric uid/gid.
361 Description: Improved cleanup procedure for conf-sanity test_46a.
365 Description: Do not start bulk transfer, if client tried to reconnect.
366 Details : Do not start bulk transfer, if client tried to reconnect,
367 but failed due to active rpcs. Check for exp_abort_active_req
368 has been implemented.
372 Description: Prevent inconsistences between linux and lustre mount structures.
373 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
374 Make the sleep interruptible.
376 Severity : enhancement
378 Description: provide server to client comms path
379 Details : server to client communications path via new LDLM_SET_INFO rpc
383 Description: truncate starts GFP_FS allocation under transaction causing deadlock
384 Details : ldiskfs_truncate calls grab_cache_page which may start page
385 allocation under an open transaction. This may lead to
386 calling prune_icache with consequent lustre reentrance.
390 Description: send correctly lsm on open replay
391 Details : MDS is trust to LSM size on replay open, but client can set wrong size
396 Description: enable client interface failover
397 Details : When a child reconnects from another NID, properly update export
398 nid hash position and ldlm reverse import.
400 Severity : enhancement
402 Description: Communicate OST degraded/readonly state via statfs to MDS
403 Details : Flags in the statfs returned from OSTs indicate whether the
404 OST is in a degraded RAID state, or if the filesystem has
405 turned read-only after a filesystem error is detected.
410 Description: don't panic if EPROTO was hit when reading symlink
411 Details : correctly handling request reference in error cases.
413 Severity : enhancement
415 Description: Add LustreNetLink, a kernel-userspace communcation path.
417 Severity : enhancement
419 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
424 Description: access to llog context before init.
425 Details : move handling CATALOGS file at osc layer and forbid access to llog
429 Frequency : rare, only when client is writing to OST on local node
431 Description: Avoid deadlock for local client writes
432 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
433 memory freeing context. This allows OST threads to set the
434 PF_MEMALLOC flag on task structures in order to allocate memory
435 from reserved pools and complete IO.
436 Use GFP_HIGHUSER for OST allocations for non-local client writes,
437 so that the OST threads generate memory pressure and allow
438 inactive pages to be reclaimed.
440 Severity : enhancement
442 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
447 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
448 Details : move ldlm namespace creation in setup phase to avoid grab
449 _lprocfs_lock with cli_sem held.
453 Description: Temporarily disable grant shrink.
454 Details : Disable the feature for debugging.
457 Frequency : only when formatting test filesystems
459 Description: Unable to run several mkfs.lustre on loop devices at the same time
460 Details : mkfs.lustre returns error 256 on the concurrent loop devices
461 formatting. The solution is to proper handle the error.
463 Severity : enhancement
465 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
467 Severity : enhancement
469 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
474 Description: resolve race between obd_disconnect and class_disconnect_exports
475 Details : if obd_disconnect will be called to already disconnected export he
476 forget release one reference and osc module can't unloaded.
478 Severity : enhancement
480 Description: Allow tuning service thread via /proc
481 Details : For each service a new
482 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
483 created that can be used to set min/max thread counts, and get the
484 current number of running threads.
489 Descriptoin: don't return error if have partially created objects for file.
490 Details : lov_update_create_set uses set->set_success as index for created
491 objects, so if some requests will be failed, they will have hole
492 at end of array and qos_shrink_lsm for allocate correct lsm.
494 Severity : enhancement
496 Description: Update OFED support to 1.4.1
501 Description: Reduce small size read RPC
502 Details : Set read-ahead limit for every file and only do read-ahead when
503 available read-ahead pages are bigger than 1M to avoid small size
506 Severity : enhancement
508 Description: move AT tunable parameters for more consistent usage
509 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
511 Severity : enhancement
513 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
514 Details : allow skip disconnected ost for send statfs request and hide error
520 Description: fix racy locking of mballoc block bitmaps causing BUG
521 Details : The locking of the mballoc buddy bitmap and the in-memory
522 block bitmap was using two different spin locks in some
523 cases. This made it possible to incorrectly access the
524 mballoc bitmap while another process was modifying it,
525 causing a sanity assertion to fail. While no on-disk corruption
526 was reported, there was some risk of this happening.
529 Frequency : rare, on llog test 6
531 Descriptoin: don't allow connect to already connected import
532 Details : allowing connect to already connected import is hide connecting problem.
537 Descriptoin: allow kill process which wait statahead result
538 Details : for some reasons 'ls' can stick in waiting result from statahead,
539 in this case need way for kill this process.
541 Severity : enhancement
543 Description: Add state history info file, enhance import info file
544 Details : Track import connection state changes in a new osc/mdc proc file;
545 add overview-type data to the osc/mdc import proc file.
547 Severity : enhancement
549 Description: MDS create should not wait for statfs RPC while holding DLM lock.
551 Severity : enhancement
553 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
558 Descriptoin: OST grant too much space to client even there are not enough space.
559 Details : Client will shrink its grant cache to OST if there are no write
560 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
561 this grant cache if there are already not enough avaible space
562 (left_space < total_clients * 32M).
565 Frequency : start MDS on uncleanly shutdowned MDS device
567 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
568 Details : stay in waiting mds<>ost recovery finished produce random bugs
569 due race between two ll_sync thread for one lov target. send
570 ACTIVATE event only if connect realy finished and import have
574 Frequency : rare, connect and disconnect target at same time
576 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
577 Details : don't call obd_disconnect under lov_lock. this long time
578 operation and can block ptlrpcd which answer to connect request.
583 Descriptoin: don't lose wakeup for imp_recovery_waitq
584 Details : recover_import_no_retry or invalidate_import and import_close can
585 both sleep on imp_recovery_waitq, but we was send only one wakeup
589 Frequency : always with long access acl
591 Descriptoin: mds can't pack reply with long acl.
592 Details : mds don't control size of acl but they limited by reint/getattr
595 Severity : enhancement
597 Description: Update to SLES10 kernel-2.6.16.60-0.33.
599 Severity : enhancement
601 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
604 Frequency : start MDS on uncleanly shutdowned MDS device
606 Descriptoin: aborting recovery hang on MDS
607 Details : don't throttle destroy RPCs for the MDT.
610 Frequency : on remount
612 Description: external journal device not working after the remount
613 Details : clear dev_rdonly flag for external journal devices in
619 Description: shutdown vs evict race
620 Details : client_disconnect_export vs connect request race.
621 if client will evicted at this time - we start invalidate
622 thread without referece to import and import can be freed
626 Frequency : rare, need acl's on inode.
628 Description: client can't handle ost additional correctly
629 Details : if ost was added after client connected to mds client can have
630 hit lnet_try_match_md ... to big messages to wide striped files.
631 in this case need teach client to handle config events about add
632 lov target and update client max ea size at that event.
634 Severity : enhancement
636 Description: Changelogs
637 Details : Changelogs are a lightweight mechanism to track filesystem
638 metadata and namespace changes. The changelog is recorded
639 permanently on the MDTs, and is periodically "consumed" / purged
640 when records are no longer needed.
642 Severity : enhancement
644 Description: compact fld format with extents
645 Details : Store range of seq rather than every seq in FLD. Seq
646 controller update FLD rather than clients. In Case of CMD, mdt0
647 has FLD, all other metadata server act as non persistent proxy
648 for FLD queries and cache fld entries in fld cache.
653 Description: don't skip ost target if they assigned to file
654 Details : Drop slow OSCs if we can, but not for requested start idx.
655 This means "if OSC is slow and it is not the requested
656 start OST, then it can be skipped, otherwise skip it only
657 if it is inactive/recovering/out-of-space.
661 Description: more cleanup in mds_lov
662 Details : not send LOV EA under replay, we can't know about they size at this
663 time. Don't allow client connect to mds before any ost connected,
664 for avoid problems with LOV EA size and returning EIO to client.
666 Severity : enhancement
668 Description: Interoperability at server side (Disk interoperability)
670 Severity : enhancement
672 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
674 Severity : enhancement
676 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
678 Severity : enhancement
680 Description: New client IO stack (CLIO).
682 Severity : enhancement
684 Description: Commit on sharing. Eliminate inter-client dependencies between
685 uncommitted transactions by doing transaction commits.
686 Thereby clients may recovery independently.
689 Frequency : Create a symlink file with a very long name
691 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
692 Details : If there is no extra space in the request for early cancels,
693 ldlm_req_handles_avail() returns 0 instead of a negative value.
695 Severity : enhancement
697 Description: Add /proc entry for import status
698 Details : The mdc, osc, and mgc import directories now have
699 an import directory that contains useful import data for debugging
702 Severity : enhancement
704 Description: Re-disable certain /proc logging
705 Details : Enable and disable client's offset_stats, extents_stats and
706 extents_stats_per_process stats logging on the fly.
709 Frequency : Only on FC kernels 2.6.22+
711 Description: oops in statahead
712 Details : Do not drop reference count for the dentry from VFS when lookup,
713 VFS will do that by itself.
715 Severity : enhancement
717 Description: Generic /proc file permissions
718 Details : Set /Proc file permissions in a more generic way to enable non-
719 root users operate on some /proc files.
723 Description: Hitting mdc_commit_close() ASSERTION
724 Details : Properly handle request reference release in
725 ll_release_openhandle().
729 Description: quota recovery deadlock during mds failover
730 Details : This patch includes att18982, att18236, att18237 in bz14840.
732 1. fix osts hang when mds does failover with quotaon
733 2. prevent watchdog storm when osts threads wait for the
738 Frequency : only patchless client
739 Description: add workaround for race between add/remove dentry from hash
741 Severity : enhancement
743 Description: Allow OST glimpses to return PW locks
747 Description: LBUG when llog conf file is full
748 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
752 Description: Prevent import from entering FULL state when server in recovery
756 Description: service mount cannot take device name with ":"
757 Details : Only when device name contains ":/" will mount treat it as
763 Description: replace ptlrpcd with the statahead thread to interpret the async
764 statahead RPC callback
768 Frequency : on recovery
769 Description: I/O failures after umount during fail back
770 Details : if client reconnected to restarted server we need join to recovery
771 instead of find server handler is changed and process self
772 eviction with cancel all locks.
774 Severity : enhancement
776 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
778 Severity : enhancement
780 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
782 Severity : enhancement
784 Description: Upcall on Lustre log has been dumped
785 Details : Allow for a user mode script to be called once a Lustre log has
786 been dumped. It passes the filename of the dumped log to the
787 script, the location of the script can be specified via
788 /proc/sys/lnet/debug_log_upcall.
793 Description: avoid idr_remove called for id which is not allocated.
794 Details : Move assigment s_dev for clustered nfs to end of initialization,
795 to avoid problem with error handling.
800 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
801 Details : When connection is reused this not moved from CONN_UNUSED_HASH
802 into CONN_USED_HASH and this prodice warning when put connection
803 again in unused hash.
806 Severity : enhancement
808 Description: File striping can now be set to use an arbitrary pool of OSTs.
810 Severity : enhancement
812 Description: Export bytes_read/bytes_write count on OSC/OST.
816 Description: Early reply size mismatch, MGC loses connection
817 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
818 the connect flags are properly negotiated.
823 Description: quotas are not honored with O_DIRECT
824 Details : all writes with the flag O_DIRECT will use grants which leads to
825 this problem. Now using OBD_BRW_SYNC to guard this.
829 Description: add quota statistics
830 Details : 1. sort out quota proc entries and proc code.
831 2. add quota statistics
833 Severity : enhancement
835 Description: enable quota support for HEAD.
839 Description: Properly propagate oinfo flags from lov to osc for statfs
840 Details : restore missing copy oi_flags to lov requests.
842 Severity : enhancement
844 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
845 : l_getgroups(8), lst(8), routerstat(8)
846 Details : included man pages for llobdstat(8), llstat(8),
847 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
849 Severity : enhancement
851 Description: configure's --enable-quota should check the
852 : kernel .config for CONFIG_QUOTA
853 Details : configure is terminated if --enable-quota is passed but
854 : no quota support is in kernel
858 Description: Remove portals compatibility
859 Details : Remove portals compatibility, not interoperable with releases
864 Description: Resolve device initialization race
865 Details : Prevent proc handler from accessing devices added to the
866 obd_devs array but yet be intialized.
868 Severity : enhancement
870 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
872 Severity : enhancement
874 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
879 Description: Using wrong pointer in osc_brw_prep_request
880 Details : Access to array[-1] can produce panic if kernel compiled with
881 CONFIG_PAGE_ALLOC enabled
885 Description: Client runs out of low memory
886 Details : Consider only lowmem when counting initial number of llap pages
890 Description: *optional* service tags registration
891 Details : if the "service tags" package is installed on a Lustre node
892 When the filesystem is mounted, a local-node service tag will
893 be created. See http://inventory.sun.com/ for more information
894 about the Service Tags asset management system.
898 Description: Kernel BUG tries to release flock
899 Details : Lustre does not destroy flock lock before last reference goes
900 away. So always drop flock locks when client is evicted and
901 perform unlock regardless of successfulness of speaking to MDS.
905 Description: add refcount protection for osc callbacks, avoid panic on shutdown
909 Description: sanity test 65a fails if stripecount of -1 is set
910 Details : handle -1 striping on filesystem in ll_dirstripe_verify
915 Description: ASSERTION(CheckWriteback(page,cmd)) failed
916 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
919 Severity : enhancement
921 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
925 Description: do not process already freed flock
926 Details : flock can possibly be freed by another thread before it reaches
927 to ldlm_flock_completion_ast.
931 Description: LBUG during stress test
932 Details : Need properly lock accesses the flock deadlock detection list.
936 Description: oops in page fault handler
937 Details : kernel page fault handler can return two special 'pages' in error
938 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
942 Description: timeout with invalidate import.
943 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
944 should be handled by ptlrpcd. This produce long age waiting and
945 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
947 Severity : enhancement
949 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
953 Description: enable MGS and MDT services start separately
954 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
955 a co-located MGS without starting the MGS, which is a complement
956 to 'nosvc' mount option.
960 Frequency : after recovery
961 Description: precreate to many object's after del orphan.
962 Details : del orphan st in oscc last_id == next_id and this triger growing
963 count of precreated objects. Set flag LOW to skip increase count
964 of precreated objects.
968 Frequency : rare, on clear nid stats
969 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
970 Details : when clean nid stats sometimes try destroy live entry,
971 and this produce panic in free.
975 Description: Stack overflow during MDS log replay
976 ease stack pressure by using a thread dealing llog_process.
980 Description: wait until IO finished before start new when do lock cancel.
981 Details : VM protocol want old IO finished before start new, in this case
982 need wait until PG_writeback is cleared until check dirty flag and
983 call writepages in lock cancel callback.
985 Severity : enhancement
987 Description: using special macro for print time and cleanup in includes.
991 Description: mds_mfd_close() ASSERTION(rc == 0)
992 Details : In mds_mfd_close(), we need protect inode's writecount change
993 within its orphan write semaphore to prevent possible races.
997 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1001 Frequency : rare, on shutdown ost
1002 Description: don't hit live lock with umount ost.
1003 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1004 use shrink_dcache_sb instead.
1008 Description: don't panic with use echo client
1009 Details : echo client pass NULL as client nid pointer and this produce null
1010 pointer dereference.
1014 Description: fix build on ppc32
1015 Details : compile code with -m64 flag produce wrong object file for ppc32.
1019 Description: add message levels for liblustreapi
1023 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1024 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1025 will result in an -EAGAIN and in the exit call path we call
1026 alloc_rr() which will with increasing aggressiveness attempt to
1027 aquire precreated objects on the minimum number of required OSCs.
1031 Description: Use old size assignment to avoid deadlock
1032 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1033 the scheduling while holding a spinlock. We do not need locking
1034 for size in ll_update_inode() because size is only updated from
1035 the MDS for directories or files without objects, so there is no
1036 other place to do the update, and concurrent access to such inodes
1037 are protected by the inode lock.
1041 Description: resolve "_IOWR redefined" build error on SLES10
1045 Description: dump the memory debugging after all modules are unloaded to
1046 suppress false negative in conf_sanity test 39
1048 Severity : enhancement
1050 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1055 Description: fix SLES kernel versioning
1056 Details : the kernel version for our SLES 10 kernel did not include a "-"
1057 before the "smp" at the end. while this was not a problem in
1058 general, it did mean that software trying to use the kernel
1059 version to try to detect a vendor specific kernel would fail.
1060 this was most evident by the OFED build scripts.
1064 Description: Don't update lov_desc members until making sure they are valid
1065 Details : When updating lov_desc members via proc fs, need fix their
1066 validities before doing the real update.
1070 Description: don't put request into delay list while invalidate in flight.
1071 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1072 import in flight. this produce timeout for invalidate and sometimes
1073 can cause stale data.
1077 Frequency : on ppc only
1078 Description: not convert ost objects for directory because it's not exist.
1079 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1083 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1085 Severity : enhancement
1087 Description: Update to support 2.6.22.14 vanilla kernel.
1091 Frequency : rare, on recovery
1092 Description: read procfs can produce deadlock in some situation
1093 Details : Holding lprocfs lock which send rpc can produce block for destroy
1094 obd objects and this also block reconnect with -EALREADY. This
1095 isn't fix all lprocfs bugs - but make it rare.
1097 Severity : enhancement
1099 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1102 Frequency : frequent on X2 node
1104 Description: mdc_set_open_replay_data LBUG
1105 Details : Set replay data for requests that are eligible for replay.
1109 Description: lustre_mgs: operation 101 on unconnected MGS
1110 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1111 MGC, and late on MGC cannot successfully connect to MGS and a lot
1112 of the error messages complaining that MGS is not connected.
1116 Frequency : on network error
1117 Description: panic with double free request if network error
1118 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1119 but it's true only for synchronus enqueue, for async enqueue
1120 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1123 Severity : enhancement
1125 Description: client-side metadata stat-ahead during readdir(directory readahead)
1126 Details : perform client-side metadata stat-ahead when the client detects
1127 readdir and sequential stat of dir entries therein
1130 Frequency : on start mds
1132 Description: Implement get_info(last_id) in obdfilter.
1135 Frequency : occasional
1137 Description: Correctly check stale fid, not start epoch if ost not support SOM
1138 Details : open with flag O_CREATE need set old fid in op_fid3 because
1139 op_fid2 was overwritten with new generated fid, but mds can answer
1140 with one of these two fids and both is not stale. Setattr
1141 incorrectly started an epoch and assume will be called
1142 done_writing, but without SOM done_writing ever being called.
1145 Frequency : rare, depends on device drivers and load
1147 Description: MDS or OSS nodes crash due to stack overflow
1148 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1149 In some cases, in conjunction with device drivers that use a lot
1150 of stack the MDS (or possibly OSS) service threads could overflow
1151 the stack. One change which was identified to consume additional
1152 stack has been reworked to avoid the extra stack usage.
1155 Frequency : occasional
1157 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1158 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1159 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1160 oscc_has_objects and see OSC has no objests and call
1161 oscc_internal_create to resend create request.
1163 Severity : enhancement
1165 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1167 Severity : enhancement
1169 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1172 Frequency : very rare
1174 Description: Fix replay if there is an un-replied request and open
1175 Details : In some cases, older replay request will revert the
1176 mcd->mcd_last_xid on MDS which is used to record the client's
1177 latest sent request.
1179 Severity : enhancement
1181 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1183 Severity : enhancement
1185 Description: Add rhel5 support to HEAD.
1187 Serverity : enhancement
1189 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1194 Description: Don't allow skipping OSTs if index has been specified.
1195 Details : Don't allow skipping OSTs if index has been specified, make
1196 locking in internal create lots better.
1200 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1201 Details : access to bitfield in structure is always rounded to long
1202 and this produce problem with not atomic change any bit.
1206 Description: Lustre make rpms failed.
1207 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1208 builds Lustre rpms from tarball.
1211 Frequency : rare on shutdown ost
1213 Description: If llog cancel was not send before clean_exports phase, this can
1214 produce deadlock in llog code.
1215 Details : If llog thread has last reference to obd and call class_import_put
1216 this produce deadlock because llog_cleanup_commit_master wait when
1217 last llog_commit_thread exited, but this never success because was
1218 called from llog_commit_thread.
1222 Description: allow userland application know is lost one of stripes.
1223 Details : fill lvb_blocks with error code on ost and return it to
1224 application if error flag found.
1228 Description: NULL lov_tgts causing MDS oops
1229 Details : more safe checks for NULL lov_tgts for avoid oops.
1231 Severity : enhancement
1233 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1237 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1238 Details : If an OST is down the MDS will hang indefinitely in
1239 obd_statfs_rqset() waiting for the statfs data. While for
1240 MDS QOS usage of statfs, it should not stuck in waiting.
1242 Severity : enhancement
1244 Description: remote_acl support
1245 Details : Support ACL-based permission check for remote user.
1246 Support setfacl/getfacl for remote user with the utils
1247 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1248 the system "{s,g}etfacl" utils.
1250 Severity : enhancement
1252 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1254 Severity : enhancement
1256 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1260 Description: make mgs_setparam() handle fsname containing dash
1261 Details : fsname containing a dash does not work with lctl conf_param
1263 Severity : enhancement
1265 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1267 Severity : enhancement
1269 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1273 Frequency : rare, at shutdown
1274 Description: access already free / zero obd_namespace.
1275 Details : if client_disconnect_export was called without force flag set,
1276 and exist connect request in flight, this can produce access to
1277 NULL pointer (or already free pointer) when connect_interpret
1278 store ocd flags in obd_namespace.
1282 Frequency : only at startup
1283 Description: not alloc memory with spinlock held.
1284 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1285 if any spinlock held.
1287 Severity : enhancement
1289 Description: make lustre randomly fail allocating memory
1290 Details : Make lustre randomly failed allocating memory for testing purpose.
1292 Severity : enhancement
1294 Description: lost problems with lov objid file
1295 Details : Fixes some scability and access to not inited memory problems
1296 in work with lov objdid file.
1301 Description: lfs find does not continue on file error
1302 Details : Continue other files processing when a file/dir is absent.
1306 Description: Inconsistent usage of lustre_pack_reply()
1307 Details : Standardize the usage of lustre_pack_reply() such that it
1308 always generate a CERROR on failure.
1311 Frequency : occasional
1313 Description: MDS hang or stay in waiting lock
1314 Details : If client receive lock with CBPENDING flag ldlm need send lock
1315 cancel as separate rpc, to avoid situation when cancel request
1316 can't processed due all i/o threads stay in wait lock.
1320 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1324 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1328 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1332 Description: Update to RHEL4 latest kernel.
1334 Severity : enhancement
1336 Description: Build SLES10 patchless client fails
1337 Details : The configure was broken by run ./configure with
1338 --with-linux-obj=.... argument for patchless client. When the
1339 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1340 can't search header adequately. Use absolute path such as
1341 -I($LINUX)/include instead.
1345 Description: interrupt oig_wait produce painc on resend.
1346 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1347 private set, and this produce situation when rq_ptlrpcd_data not
1348 copyed to new allocated request and triggered LBUG on assert
1349 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1350 ptlrpcd set if request is changed and can be safety changed to use
1355 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1356 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1357 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1358 these macros and replace them with their difinitions instead.
1362 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1363 Details : Update which_patch & target file for SLES10 latest kernel.
1366 Description: Remove client patches from SLES 10 kernel.
1367 Details : This causes SLES 10 clients to behave as patchless clients
1368 even on a Lustre-patched (server) kernel.
1370 Severity : enhancement
1372 Description: self-adjustable client's lru lists
1373 Details : use adaptive algorithm for managing client cached locks lru
1374 lists according to current server load, other client's work
1375 pattern, memory activities, etc. Both, server and client
1376 side namespaces provide number of proc tunables for controlling
1379 Severity : enhancement
1381 Description: light-weight GSS support
1382 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1383 and reduce performance overhead.
1385 Severity : enhancement
1387 Description: Linux keyring support
1388 Details : Support using service of Linux keyring for Lustre GSS internal
1389 context refresh/cache mechanism.
1393 Description: Fix errors in lfs documentation
1394 Details : Fixes man pages
1398 Description: don't use GFP_* in generic Lustre code.
1399 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1403 Description: obdclass is limited by single OBD_ALLOC(idarray)
1404 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1408 Description: warnings with build patchless client with vanila 2.6.19 and up
1409 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1410 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1414 Frequency : Only for SLES
1415 Description: sanity_quota fail test_1
1416 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1417 file, which makes TSTID[2] inunique.
1421 Frequency : Only for SLES9
1422 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1423 Details : Update target/ChangeLog/which_patch .
1427 Frequency : Only for test_10 in sanity-quota.sh
1428 Description: a bug in quota test code
1429 Details : checking a test flag in a wrong place
1433 Frequency : Only for test_13 in sanity-quota.sh
1434 Description: a bug in quota test script
1435 Details : shouldn't check overfull allocated quota
1439 Description: jbd statistics
1440 Details : Port older jbd statistics patch for sles10
1442 Severity : enhancement
1444 Description: Add printing inode info into message about error in writepage.
1449 Description: reply_lock_interpret crash due to race with it and lock cancel.
1450 Details : Do not replay locks that are being cancelled. Do not reference
1451 locks by their address during replay, just by their handle.
1455 Frequency : When flocks are used.
1456 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1457 Details : Flock locks might destroy just granted lock if it could be merged
1458 with another existing flock, this is done in completion handler,
1459 so teach ldlm_cli_enquque_fini that this is a valid case for
1465 Description: Oops in read and write path when failing to allocate lock.
1466 Details : Check if lock allocation failed and return error back.
1470 Description: lstripe command fails for valid OST index
1471 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1472 instead of lov->desc.ld_active_tgt_count.
1476 Description: sanity.sh failed test 103
1477 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1482 Description: df doesn't work properly if diskfs blocksize != 4K
1483 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1487 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1488 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1490 Severity : enhancement
1492 Description: Adaptive timeouts
1493 Details : RPC timeouts adapt to changing server load and network
1494 conditions to reduce resend attempts and improve recovery time.
1498 Description: llapi_file_create() does not allow some changes
1499 Details : add llapi_file_open() that allows specifying the mode and
1500 open flags, and also returns an open file handle.
1504 Description: merge and cleanup kernel patches.
1505 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1509 Description: Add journal checksum support.(Kernel part)
1510 Details : The journal checksum feature adds two new flags i.e
1511 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1512 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1513 indicates that the commit block contains the checksum for
1514 the blocks described by the descriptor blocks. Now commit
1515 record can be sent to disk without waiting for descriptor
1516 blocks to be written to disk. This behavior is controlled
1517 using JBD2_FEATURE_ASYNC_COMMIT flag.
1521 Description: OSS needs mutliple precreate threads
1522 Details : Add ability to start more than one create thread per OSS.
1526 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1527 Details : execute lfs setstripe on client
1531 Description: mds_obd_create error creating tmp object
1532 Details : When the user sets quota on root, llog will be affected and can't
1533 create files and write files.
1536 Frequency : Always on ia64 patchless client, and possibly others.
1538 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1539 Details : This allows the patchless client to be loaded on architectures
1540 without this export.
1544 Description: RedHat Update kernel for RHEL5
1545 Details : Add the kernel config file for RHEL5.
1549 Description: RedHat Update kernel for RHEL5
1550 Details : Modify the kernel config file more closer RHEL5.
1554 Description: Build failure against Centos5 (RHEL5)
1555 Details : Define PAGE_SIZE when it isn't present.
1559 Description: test_7 on sanity-quota.sh can't be used on separate machines
1560 Details : add facet to handle it
1564 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1565 Details : with NFS, the anon dentry's parent was set to itself in
1566 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1567 corresponding dentry other than use rec->ur_name.
1569 Severity : enhancement
1571 Description: lfs setstripe enhancement
1572 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1576 Description: enable data checksumming by default
1577 Details : enable checksum by default, allow --disable-checksum
1578 configure option and "-o nochecksum" mount option. Checksums
1579 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1583 Description: lustre support for RHEL5
1584 Details : Add support for RHEL5.
1588 Description: Client eviction due to failover config
1589 Details : after a connection loss, the lustre client should attempt to
1590 reconnect to the last active server first before trying the
1591 other potential connections.
1595 Description: when mds and osts use different quota unit(32bit and 64bit),
1596 quota will be released repeatly.
1597 Details : void sending multiple quota reqs to mds, which will keep the
1598 status between the reqs.
1602 Description: osts not allocated evenly to files
1603 Details : change the condition to increase offset_idx
1606 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1608 Description: Data corruption for OSTs that are formatted larger than 2TB
1610 Details : When generating the bio request for lustre file writes the
1611 sector number would overflow a temporary variable before being
1612 used for the IO. The data reads correctly from Lustre (which
1613 will overflow in a similar manner) but other file data or
1614 filesystem metadata may be corrupted in some cases.
1618 Description: Tune the kernel for good SCSI performance.
1619 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1620 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1625 Description: rewrite ext2-derived code in obdclass/uuid.c
1626 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1627 from scratch preserving functionality.
1631 Description: "lfs find -obd UUID" prints directories
1632 Details : "lfs find -obd UUID" will return all directory names instead
1633 of just file names. It is incorrect because the directories
1634 do not reside on the OSTs.
1638 Description: use i_size_read and i_size_write in 2.6 port
1639 Details : replace inode->i_size access with i_size_read/write()
1642 Frequency : when removing large files
1644 Description: scheduling issue during removal of large Lustre files
1645 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1646 It causes scheduling issues when removing large files (17TB in the
1650 Frequency : only with liblustre clients on XT3
1652 Description: evictions taking too long
1653 Details : allow llrd to evict clients directly on OSTs
1656 Frequency : only on ppc
1658 Description: /proc/fs/lustre/devices broken on ppc
1659 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1660 We should make sure the type of 'index' is loff_t and then cast
1661 explicitly as needed below. Do not assign an explicitly cast
1665 Frequency : only for rhel5
1667 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1668 Details : Modify the target file & which_kernel.
1670 Severity : enhancement
1672 Description: omit set fsid for export NFS
1673 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1674 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1678 Frequency : Always, for kernels after 2.6.16
1679 Description: Fix warning idr_remove called for id=.. which is not allocated.
1680 Details : Last kernels save old s_dev before kill super and not allow
1681 to restore from callback - restore it before call kill_anon_super.
1685 Description: buffer overruns could theoretically occur
1686 Details : llapi_semantic_traverse() modifies the "path" argument by
1687 appending values to the end of the origin string, and a
1688 overrun may occur. Adding buffer overrun check in liblustreapi.
1692 Description: Fix error on 'ls .' at the top of the Lustre mount.
1693 Details : Don't revalidate dentry if it is a root dentry.
1697 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1698 Details : Modify vm-tunables-rhel4.patch.
1702 Description: Kernel config for 2.6.18-vanilla.
1703 Details : Modify targets/2.6-vanilla.target.in.
1704 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1705 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1706 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1707 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1711 Description: improve handling recoverable errors
1712 Details : if request processig with error which can be recoverable on server
1713 request should be resend, otherwise page released from cache and
1718 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1719 Details : Modify target file & which_patch.
1720 A flaw was found in the IA32 system call emulation provided
1721 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1722 value could be stored in the %RAX register, which could trigger an
1723 out-of-bounds system call table access. An untrusted local user
1724 could exploit this flaw to run code in the kernel
1725 (ie a root privilege escalation). (CVE-2007-4573).
1729 Description: O_DIRECT bypasses client statistics.
1730 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1731 read_ahead_stats were updated. copied the stats section from
1732 osc_send_oap_rpc() into async_internal().
1736 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1740 Description: change order of libsysio includes
1741 Details : '#include sysio.h' should always come before '#include xtio.h'
1745 Description: handle "serious error: objid * already exists" more gracefully
1746 Details : If LAST_ID value on disk is smaller than the objects existing in
1747 the O/0/d* directories, it indicates disk corruption and causes an
1748 LBUG(). If the object is 0-length, then we should use the existing
1749 object. This will help to avoid a full fsck in most cases.
1751 Severity : enhancement
1753 Description: adapt the lustre_config script to support the upgrade case
1754 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1755 server targets to 1.6 in parallel.
1759 Description: To avoid grant space > avaible space when the disk is almost
1760 full. Without this patch you might see the error "grant XXXX >
1761 available" or some LBUG about grant, when the disk is almost
1763 Details : In filter_check_grant, for non_grant cache write, we should
1764 check the left space by if (*left > ungranted + bytes), instead
1765 of (*left > ungranted), because only we are sure the left space
1766 is enough for another "bytes", then the ungrant space should be
1767 increase. In client, we should update cl_avail_grant only there
1768 is OBD_MD_FLGRANT in the reply.
1772 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1773 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1774 described in CVE-2007-4573. This problem could allow untrusted
1775 local users to gain root access.
1778 Frequency : when using O_DIRECT and quotas
1780 Description: Incorrect file ownership on O_DIRECT output files
1781 Details : block usage reported by 'lfs quota' does not take into account
1782 files that have been written with O_DIRECT.
1787 Description: touch file failed when fs is not full
1788 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1789 otherwise we can get ENOSP while fs is not full.
1793 Description: parallel lock callbacks
1794 Details : Instead of sending blocking and completion callbacks as separated
1795 requests, adding them to a set and sending in parallel.
1798 Frequency : only for Cray XT3
1799 Bugzilla : 12829/13455
1800 Description: Changing primary group doesn't change the group lustre assigns to
1802 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1803 overridden with the primary group provided by the group upcall,
1804 whereas the supplied fsgid can be trusted if it is in the list of
1805 supplementary groups returned by the group upcall.
1807 Severity : enhancement
1809 Description: Allow masking D_WARNING, D_ERROR messages from console
1810 Details : Console messages can now be disabled via lnet.printk.
1814 Description: User code with malformed file open parameter crashes client node
1815 Details : Before packing join_file req, all the related reference should be
1816 checked carefully in case some malformed flags cause fake
1817 join_file req on client.
1821 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1822 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1823 mode is equal to the granted mode. However, it should also check
1824 that LDLM_FL_CP_REQD is not set.
1828 Description: Heavy nfs access might result in deadlocks
1829 Details : After ELC code landed, it is now improper to enqueue any mds
1830 locks under och_sem, because enqueue might want to decide to
1831 cancel open locks for same inode we are holding och_sem for.
1835 Description: Client eviction while running blogbench
1836 Details : A lot of unlink operations with concurrent I/O can lead to a
1837 deadlock causing evictions. To address the problem, the number of
1838 oustanding OST_DESTROY requests is now throttled to
1839 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1840 callbacks are processed in priority.
1844 Description: enable ACLs on MDS by default
1845 Details : ACLs must be enabled on MDS by default.
1848 Frequency : PPC/PPC64 only
1850 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1851 Details : fix duplicated definitions between asm-ppc64/types.h and
1852 lustre_types.h on PPC.
1855 Frequency : PPC/PPC64 only
1857 Description: asm-ppc/segment.h does not exist
1858 Details : fix compile issue on PPC.
1862 Description: better handle error messages in extents code
1865 Frequency : RHEL4 only
1867 Description: mkfs is very slow on IA64/RHEL4
1868 Details : A performance regression has been discovered in the MPT Fusion
1869 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1870 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1871 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1873 Severity : enhancement
1875 Description: SNMP support enhancement
1876 Details : Adding total number of sampled request for an MDS node in snmp
1879 Severity : enhancement
1881 Description: Optimize ldlm waiting list processing for PR extent locks
1882 Details : When processing waiting list for read extent lock and meeting read
1883 lock that is same or wider to it that is not contended, skip
1884 processing rest of the list and immediatelly return current
1885 status of conflictness, since we are guaranteed there are no
1886 conflicting locks in the rest of the list.
1890 Description: Time out and refuse to reconnect
1891 Details : When the failover node is the primary node, it is possible
1892 to have two identical connections in imp_conn_list. We must
1893 compare not conn's pointers but NIDs, otherwise we can defeat
1894 connection throttling.
1898 Description: port llog fixes from b1_6 into HEAD
1899 Details : Port llog reference couting and some llog cleanups from b1_6
1900 (bug 10800) into HEAD, for protect from panic and access to already
1901 free llog structures.
1905 Description: Detect stride IO mode in read-ahead
1906 Details : When a client does stride read, read-ahead should detect that and
1907 read-ahead pages according to the detected stride pattern.
1911 Description: data checksumming impacts single node performance
1912 Details : add support for several checksum algorithm. Currently, only CRC32
1913 and Adler-32 are supported. The checksum type can be changed on
1914 the fly via /proc/fs/lustre/osc/*/checksum_type.
1918 Description: use adler32 for page checksums
1919 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1924 Description: build for x2 fails
1925 Details : fix compile issue on Cray systems.
1929 Description: Properly match for duplicate locks
1930 Details : Due to different lock order from skiplists code, we need to
1931 traverse entire list for now
1934 Frequency : only on PPC/SLES10
1936 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1937 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1938 be defined. Add a hack in mkfs_lustre.c to work around this compile
1943 Description: LASSERT on MDS when client holding flock lock dies
1944 Details : ldlm pool logic depends on number of granted locks equal to
1945 number of released locks which is not true for flock locks, so
1946 just exclude such locks from consideration.
1950 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1951 Details : Use fsfilt_sync() for both the whole filesystem sync and
1952 individual file sync to eliminate dangerous inode locking
1953 with I_LOCK that can lead to a deadlock.
1957 Description: performance in 1.6.3
1958 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1959 sure that 1MB requests are not fragmented by the block layer.
1961 Severity : enhancement
1963 Description: organize the server-side client stats on per-nid basis
1964 Details : Change the structure of stats under obdfilter and mds to
1972 The "uuid"s file would list the uuids of _active_ exports.
1973 And the clear entry is to clear all stats and stale nids.
1975 Severity : enhancement
1977 Description: eliminate client locks in face of contention
1978 Details : file contention detection and lockless i/o implementation
1979 for contended files.
1983 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1986 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1988 Description: MDS may lose file striping (and hence file data) in some cases
1989 Details : If there are additional extended attributes stored on the MDS,
1990 in particular ACLs, SELinux, or user attributes (if user_xattr
1991 is specified for the client mount options) then there is a risk
1992 of attribute loss. Additionally, the Lustre file striping
1993 needs to be larger than default (e.g. striped over all OSTs),
1994 and an additional attribute must be stored initially in the
1995 inode and then increase in size enough to be moved to the
1996 external attribute block (e.g. ACL growing in size) for file
2001 Description: skiplist implementation simplification
2002 Details : skiplists are used to group compatible locks on granted list
2003 that was implemented as tracking first and last lock of each
2004 lock group the patch changes that to using doubly linked lists
2008 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2009 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2010 because LASSERT on both the data supplied by a client, and the
2011 data on disk is dangerous and incorrect.
2013 Severity : enhancement
2015 Description: Slow truncate/writes to huge files at high offsets.
2016 Details : Directly associate cached pages to lock that protect those pages,
2017 this allows us to quickly find what pages to write and remove
2018 once lock callback is received.
2022 Description: more ldlm soft lockups
2023 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2024 starve other threads from the resource lock for a long time in
2025 case of long waiting queue, so change the debug level from
2026 D_OTHER to the less frequently used D_INFO.
2028 Severity : enhancement
2030 Description: add -gid, -group, -uid, -user options to lfs find
2034 Description: Hung threads in invalidate_inode_pages2_range
2035 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2036 once one is completed. As a result, some RPCs are stuck in the
2037 queue and are never sent.
2041 Description: filter threads hungs on waiting journal commit
2042 Details : Cleanup filter group llog code, then only filter group llog will
2043 be only created in the MDS/OST syncing process.
2047 Description: Procfs and llog threads access destoryed import sometimes.
2048 Details : Sync the import destoryed process with procfs and llog threads by
2049 the import refcount and semaphore.
2051 Severity : enhancement
2053 Description: openlock cache of b1_6 port to HEAD
2058 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2059 Details : If insertion of an extent fails, then discard the inode
2060 preallocation and free data blocks else it can lead to duplicate
2065 Description: don't always update ctime in ext3_xattr_set_handle()
2066 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2067 In some cases the ctime should not be updated, for example for
2068 2.0->1.8 compatibility it is necessary to delete an xattr and it
2069 should not update the ctime.
2073 Bugzilla : 15713/16362
2074 Description: Assertion in iopen_connect_dentry in 1.6.3
2075 Details : looking up an inode via iopen with the wrong generation number can
2076 populate the dcache with a disconneced dentry while the inode
2077 number is in the process of being reallocated. This causes an
2078 assertion failure in iopen since the inode's dentry list contains
2079 both a connected and disconnected dentry.
2083 Description: assertion failure in ldlm_handle2lock()
2084 Details : fix a race between class_handle_unhash() and class_handle2object()
2085 introduced in lustre 1.6.5 by bug 13622.
2090 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2091 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2096 Description: oops in sd_iostats_seq_show()
2097 Details : unloading/reloading the scsi low level driver triggers a kernel
2098 bug when trying to access the sd iostat file.
2103 Description: Kernel panics during QLogic driver reload
2104 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2105 patch, causing memory corruption.
2110 Description: journal_dev option does not work in b1_6
2111 Details : pass mount option during pre-mount.
2113 Severity : enhancement
2115 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2116 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2117 extent information of a file. It can be used to map logical blocks
2118 in a file to physical blocks in the block device.
2122 Description: LDLM soft lockups - improvement
2123 Details : It is be possible to send the lock handle along with each read
2124 or write request because the client is already doing a lock match
2125 itself so there isn't any reason the OST should have to re-do that
2129 Frequency : only on Cray X2
2131 Description: X2 build failures
2132 Details : fix build failures on Cray X2.
2136 Description: xid & resent requests
2137 Details : Initialize RPC XID from clock at startup (randomly if clock is
2140 Severity : enhancement
2142 Description: Add lustre_start utility to start or stop multiple Lustre servers
2147 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2148 Details : In case of memory pressure, list_del() can be called twice on
2149 req->rq_history_list, causing a kernel oops.
2153 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2154 Details : fix stack overflow in the distributed lock manager by defering
2155 export eviction after a failed AST to the elt thread instead of
2156 handling it in the dlm interpret routine.
2160 Description: Convert some comments to new format.
2161 Details : Update documenting comments to match doxygen conventions.
2165 Description: Grammar fixes.
2166 Details : A couple of trivial spelling fixes.
2170 Description: OSD_COUNTERS-mandatory
2171 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2172 to simplify the code.
2176 Description: simplify cmm_device freeing logic.
2177 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2181 Description: Add lockdep support to dt_object_operations locking interface.
2182 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2183 parameter indicating lock ordering. Update mdd code to use new
2188 Description: Introduce failloc constants for lockless IO tests.
2189 Details : Add two new failloc constants to test lockless IO. Only one of
2190 them in implemented---another is checked in yet to be landed
2195 Description: Add lockdep support for inode mutex.
2196 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2197 in the situations where Lustre has to lock more than one inode
2202 Description: Add optional invariants checking support.
2203 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2204 configure switch. This macro is to be used for consistency and
2205 sanity checks that are too expensive to be left in `production'
2210 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2211 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2216 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2217 Details : Use spin_lock_nested() in (the only) situation where more than
2218 one ldlm_lock is locked simultaneously. Also, fix possible
2219 dead-lock in ldlm_lock_change_resource() by enforcing particular
2224 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2225 Details : Instead of specifying each ldlm_lock call-back through separate
2226 parameter, wrap them into struct ldlm_callback_suite.
2230 Description: Kill join_lru obd method and its callers.
2231 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2232 in memory---a requirement implemented through ->o_join_lru() obd
2233 method in HEAD. Remove this method and its users.
2237 Description: Add asynchronous ldlm ENQUEUE completion handler.
2238 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2239 case that stock ldlm_completion_ast() cannot handle as it waits
2240 until lock is granted. Introduce new ldlm_completion_ast_async()
2241 for this. Also comment ldlm_completion_ast().
2245 Description: ldlm_error <-> errno conversion.
2246 Details : Add functions to map (rather arbitrary) between LDLM error codes
2247 and standard errno values. CLIO needs this to prevent LDLM specific
2248 constants from escaping ldlm and osc.
2252 Description: Kill unused ldlm_handle2lock_ns() function.
2253 Details : Kill unused ldlm_handle2lock_ns() function.
2257 Description: Add lu_ref support to ldlm_lock
2258 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2259 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2260 and ldlm_resource. LDLM interface has to be changed, because code that
2261 releases a reference on a lock, has to "know" what reference this is.
2262 In the most frequent case
2264 lock = ldlm_handle2lock(handle);
2266 LDLM_LOCK_PUT(lock);
2268 no changes are required. When any other reference (received _not_ from
2269 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2270 instead of LDLM_LOCK_PUT().
2272 Arguably, changes are pervasive, and interface requires some discipline
2273 for proper use. On the other hand, it was very instrumental in finding
2274 a few leaked lock references.
2278 Description: Add ldlm_lock_addref_try().
2279 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2280 attempts to addref a lock that might be being canceled
2285 Description: Add ldlm_weigh_callback().
2286 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2287 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2288 hard-coded `number of pages' logic.
2292 Description: Add lockdep annotations to llog code.
2293 Details : Use appropriately tagged _nested() locking calls in the places
2294 where llog takes more than one ->lgh_lock lock.
2298 Description: Add loi_kms_set().
2299 Details : Wrap kms updates into a helper function.
2303 Description: Constify instances of struct lsm_operations.
2304 Details : Constify instances of struct lsm_operations.
2308 Description: lu_conf support.
2309 Details : On a server, a file system object is uniquely identified
2310 by a fid, which is sufficient to locate and load all object
2311 state (inode). On a client, on the other hand, more data are
2312 necessary instantiate an object. Change lu_object_find() and
2313 friends to take additional `lu_conf' argument describing object.
2314 Typically this includes layout information.
2318 Description: lu_context fixes.
2319 Details : Introduce new lu_context functions that are needed on the client
2320 side, where some system threads (ptlrpcd) are shared by multiple
2321 modules, and so cannot be stopped during module shutdown.
2325 Description: Add start and stop methods to lu_device_type_operations.
2326 Details : Introduce two new methods in lu_device_type_operations, that are
2327 invoked when first instance of a given type is created and last one
2328 is destroyed respectively. This is need by CLIO.
2332 Description: Add lu_ref support to struct lu_device.
2333 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2334 track leaked references.
2338 Description: Introduce lu_kmem_descr.
2339 Details : lu_kmem_descr and its companion interface allow to create
2340 and destroy a number of kmem caches at once.
2344 Description: Fix lu_object finalization race.
2345 Details : Fix a race between lu_object_find() finding an object and its
2346 concurrent finalization. This race is (most likely) not possible
2347 on the server, but might happen on the client.
2351 Description: Introduce lu_ref interface.
2352 Details : lu_ref is a debugging module allowing to track references to
2353 a given object. It is quite cpu expensive, and has to be
2354 explicitly enabled with --enable-lu_ref. See usage description
2359 Description: Factor lu_site procfs stats into a separate function.
2360 Details : Separate lu_site stats printing code into a separate function
2361 that can be reused on a client.
2365 Description: Constify instances of struct {lu,dt,md}_device_operations.
2366 Details : Constify instances of struct {lu,dt,md}_device_operations.
2370 Description: Introduce struct md_site and move meta-data specific parts of
2371 struct lu_site here.
2372 Details : Move md-specific fields out of struct lu_site into special struct
2373 md_site, so that lu_site can be used on a client.
2377 Description: Kill mdd_lov_destroy().
2378 Details : Remove unused mdd code.
2382 Description: Add st_block checking to multistat.c.
2383 Details : Add st_block checking to multistat.c.
2387 Description: Add lu_ref support to struct obd_device.
2388 Details : Add lu_ref tracking to obd_device.
2392 Description: Kill obd_set_fail_loc().
2393 Details : Remove unused code.
2397 Description: Add special type for ptlrpc_request interpret functions.
2398 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2399 there. Actual usage will be in CLIO.
2403 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2404 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2405 doesn't work with lockdep.
2409 Description: Add rwv.c test program.
2410 Details : New testing program exercising readv(2) and writev(2) (Qian).
2414 Description: Add sendfile.c test program.
2415 Details : New testing program exercising sendfile(2) (Jay).
2419 Description: Ratelimit a message that can be very frequent.
2420 Details : Ratelimit a memory allocation failure message that can
2425 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2426 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2430 Description: Liblustre build fixes.
2431 Details : Liblustre build fixes.
2435 Description: libcfs: add cfs_{need,cond}_resched() interface.
2436 Details : libcfs: add cfs_{need,cond}_resched() definition and
2437 implementations for Linux, NT, and liblustre.
2439 Severity : enhancement
2441 Description: More exported tunables for mballoc
2442 Details : Add support for tunable preallocation window and new tunables for
2443 large/small requests
2447 Description: Detect corruption of block bitmap and checking for preallocations
2448 Details : Checks validity of on-disk block bitmap. Also it does better
2449 checking of number of applied preallocations. When corruption is
2450 found, it turns filesystem readonly to prevent further corruptions.
2454 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2455 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2456 window to 0. If it does hit cache miss, and read-pattern is still
2457 stride-io mode, does not reset the stride window, but also does not
2458 increase the stride window length in this case.
2462 Frequency : only for big-endian servers
2463 Description: Check if system is big-endian while mounting fs with extents feature
2464 Details : Mounting a filesystem with extents feature will fail on big-endian
2465 systems since ext3-based ldiskfs is not supported on big-endian
2466 systems. This can be over-riden with "bigendian_extents" mount option.
2468 Severity : enhancement
2470 Description: The root squash functionality
2471 Details : A security feature, which is to prevent users from being able
2472 to mount lustre on their desktop, run as root, and delete
2473 all of the files in the filesystem. The goal is accomplished by
2474 remapping user id (UID) and group id (GID) of the root user to
2475 a UID and GID specified by the system administartor via Lustre
2476 configuration management server (MGS). The functionality also
2477 allows to specify sets of clients for which the remapping does
2482 Description: Excessive recovery window
2483 Details : With AT enabled, the recovery window can be excessively long (6000+
2484 seconds). To address this problem, we no longer use
2485 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2486 timeout no longer depends on the service time, it is set to
2487 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2488 time via pb_service_time.
2492 Description: Watchdog triggered on MDS failover
2493 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2494 the OSTs know that the MDS "UUID" can be reused for the same export
2495 from a different NID, so we do not need to wait for the export to be
2499 Frequency : rare, only if using MMP with Linux RAID
2501 Description: MMP doesn't work with Linux RAID
2502 Details : While using HA for Lustre servers with Linux RAID, it is possible
2503 that MMP will not detect multiple mounts. To make this work we
2504 need to unplug the device queue in RAID when the MMP block is being
2505 written. Also while reading the MMP block, we should read it from
2506 disk and not the cached one.
2508 Severity : enhancement
2510 Description: open file using fid
2511 Details : A file can be opened using just its fid, like
2512 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2515 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2518 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2519 Details : add proper locking for creation of jbd_4k slab cache
2523 Description: MMP check in ext3_remount() fails without displaying any error
2524 Details : When multiple mount protection fails during remount, proper error
2527 Severity : enhancement
2529 Description: Allow stripe size to be up to 4G-64k
2530 Details : Fix math logic to allow large stripe sizes.
2534 Description: add check for >8TB ldiskfs filesystems
2535 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2536 Don't allow >8TB ldiskfs filesystems to be mounted without
2537 force_over_8tb mount option
2542 Descriptoin: OSS complains "too many blocks" and crashes
2543 Details : In filter_commitrw_write():
2544 1. lquota_chkquota();
2545 2. obdo_from_inode();
2546 3. lquota_pending_commit();
2547 "2." has changed uid and gid in this case, then caused
2548 lquota_chkquota() and lquota_pending_commit() to use
2549 different uid and gid. That is the root of the bug.
2553 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2554 Details : Detect and print out the 'stripe count = -1' case properly.
2558 Description: Out or order replies might be lost on replay
2559 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2560 request with tid smaller then one currently being added, add it
2561 to the start, not end of the list.
2563 --------------------------------------------------------------------------------
2565 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2567 * Support for kernels:
2568 2.6.5-7.283 (SLES 9),
2569 2.6.9-55.EL (RHEL 4),
2570 2.6.16.46-0.14 (SLES 10),
2571 2.6.18.8 vanilla (kernel.org)
2572 * Client support for unpatched kernels:
2573 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2574 2.6.16 - 2.6.22 vanilla (kernel.org)
2575 * Due to recently discovered recovery problems, we do not recommend
2576 using patchless RHEL 4 clients with this or any earlier release.
2577 * Recommended e2fsprogs version: 1.39.cfs8
2578 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2579 * Starting with this release, the ldiskfs backing filesystem required
2580 by Lustre is now in its own package, lustre-ldiskfs. This package
2581 should be installed. It is versioned separately from Lustre and
2582 may be released separately in future.
2586 Description: block reactivating mgc import until all deactivates complete
2587 Details : Fix race when failing back MDT/MGS to itself (testing)
2590 Frequency : at statup only
2592 Description: mds_lov_synchronize race leads to various problems
2593 Details : simultaneous MDT->OST connections at startup can cause the
2594 sync to abort, leaving the OSC in a bad state.
2596 Severity : enhancement
2598 Description: add optional extra BUILD_VERSION info
2599 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2600 to override the lustre version.
2603 Frequency : 2.6.18 servers only
2605 Description: ll_kern_mount() doesn't release the module reference
2606 Details : The ldiskfs module reference count never drops down to 0
2607 because ll_kern_mount() doesn't release the module reference.
2612 Description: server LBUG when using old ost_num_threads parameter
2613 Details : Accept the old ost_num_threads parameter but warn that it
2614 is deprecated, and fix an off-by-one error that caused an LBUG.
2619 Description: Transient SCSI error results in persistent IO issue
2620 Details : iobuf->dr_error is not reinitialized to 0 between two
2624 Frequency : sometimes when underlying device returns I/O errors
2626 Description: OSTs not going read-only during write failures
2627 Details : OSTs are not remounted read-only when the journal commit threads
2628 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2629 instead of the ext3 wrappers.
2633 Description: poor connect scaling with increasing client count
2634 Details : Don't run filter_grant_sanity_check for more than 100 exports
2635 to improve scaling for large numbers of clients.
2638 Frequency : SLES10 only
2640 Description: sanity-quota.sh quotacheck failed: rc = -22
2641 Details : Quotas cannot be enabled on SLES10.
2644 Frequency : liblustre clients only
2646 Description: getdirentries does not give error when run on compute nodes
2647 Details : getdirentries does not fail when the size specified as an argument
2648 is too small to contain at least one entry
2650 Severity : enhancement
2652 Description: Add LNET router traceability for debug purposes
2653 Details : If a checksum failure occurs with a router as part of the
2654 IO path, the NID of the last router that forwarded the bulk data
2655 is printed so it can be identified.
2660 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2661 Details : Due to a race condition, liblustre clients were occasionally
2662 evicted incorrectly.
2664 Severity : enhancement
2666 Description: lfs setstripe use optional parameters instead of postional
2669 Severity : enhancement
2671 Description: Nanosecond timestamp support for ldiskfs
2672 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2673 resolution timestamps. There is not yet support for this at
2674 the Lustre filesystem level.
2677 Frequency : during server recovery
2679 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2680 Details : request with rq_no_resend flag not awake l_wait_event if they get
2684 Frequency : nfs export on patchless client
2686 Description: connectathon hang when test nfs export over patchless client
2687 Details : Disconnected dentry cannot be found with lookup, so we do not need
2688 to unhash it or make it invalid
2691 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2692 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2693 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2698 Description: open req refcounting wrong on reconnect
2699 Details : If reconnect happened between getting open reply from server and
2700 call to mdc_set_replay_data in ll_file_open, we will schedule
2701 replay for unreferenced request that we are about to free.
2702 Subsequent close will crash in variety of ways.
2703 Check that request is still eligible for replay in
2704 mdc_set_replay_data().
2709 Description: disable writes to filesystem when reading health_check file
2710 Details : the default for reading the health_check proc file has changed
2711 to NOT do a journal transaction and write to disk, because this
2712 can cause reads of the /proc file to hang and block HA state
2713 checking on a healthy but otherwise heavily loaded system. It
2714 is possible to return to the previous behaviour during configure
2715 with --enable-health-write.
2717 Severity : enhancement
2719 Description: 64-bit inode version
2720 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2721 to the inode. This will be required for version-based recovery.
2726 Description: MDS fails to start if a duplicate client export is detected
2727 Details : in some rare cases it was possible for a client to connect to
2728 an MDS multiple times. Upon recovery the MDS would detect this
2729 and fail during startup. Handle this more gracefully.
2731 Severity : enhancement
2733 Description: Add -o localflock option to simulate old noflock
2735 Details : This will achieve local-only flock/fcntl locks
2741 Description: log_commit_thread vs filter_destroy race leads to crash
2742 Details : Take import reference before releasing llog record semaphore
2747 Description: Wrong request locking in request set processing
2748 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2749 imp_delayed_list, in this place should be used imp_lock.
2752 Frequency : when reconnection
2754 Description: Grant Leak when osc reconnect to OST
2755 Details : When osc reconnect ost, OST(filter) should check whether it
2756 should grant more space to client by comparing fed_grant and
2757 cl_avail_grant, and return the granted space to client instead
2758 of "new granted" space, because client will call osc_init_grant
2759 to update the client grant space info.
2762 Frequency : when client reconnect to OST
2764 Description: Grant Leak when osc do resend and replay bulk write
2765 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2766 bulk write request, because the grant info while be sync between
2767 OSC and OST when reconnect, and we should ignore the grant info
2768 these of resend/replay write req.
2773 Description: Grant space more than avaiable left space sometimes.
2774 Details : When then OST is about to be full, if two bulk writing from
2775 different clients came to OST. Accord the avaliable space of the
2776 OST, the first req should be permitted, and the second one
2777 should be denied by ENOSPC. But if the seconde arrived before
2778 the first one is commited. The OST might wrongly permit second
2779 writing, which will cause grant space > avaiable space.
2782 Frequency : when client is evicted
2784 Description: Grant might be wrongly erased when osc is evicted by OST
2785 Details : when the import is evicted by server, it will fork another
2786 thread ptlrpc_invalidate_import_thread to invalidate the
2787 import, where the grant will be set to 0. While the original
2788 thread will update the grant it got when connecting. So if
2789 the former happened latter, the grant will be wrongly errased
2790 because of this race.
2795 Description: Checking Stale with correct fid
2796 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2797 to check whether it is stale, because sometimes, we want the
2798 enqueue happened anyway, and op_data.fid2 will not be initialized.
2800 Severity : enhancement
2802 Description: update patchless client
2803 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2806 Frequency : only with 2.4 kernel
2808 Description: random memory corruption
2809 Details : size of struct ll_inode_info is to big for union inode.u and this
2810 can be cause of random memory corruption.
2815 Description: Memory leak in recovery
2816 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2817 It should also check obd_fail before fsfilt_start, otherwise if
2818 fsfilt_start return -EROFS,(failover mds during mds recovery).
2819 then the req will return with repmsg->transno = 0 and rc = EROFS.
2820 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2821 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2822 be freed no matter whether fsfilt_commit success or not.
2825 Frequency : only with huge count clients
2827 Description: Prevents from taking the superblock lock in llap_from_page for
2829 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2830 for llap_from_page call in ll_removepage prevents from taking the
2831 superblock lock for a soon died page.
2836 Description: Not check open intent error before release open handle
2837 Details : in some rare cases, the open intent error is not checked before
2838 release open handle, which may cause
2839 ASSERTION(open_req->rq_transno != 0), because it tries to release
2840 the failed open handle.
2845 Description: Set cat log bitmap only after create log success.
2846 Details : in some rare cases, the cat log bitmap is set too early. and it
2847 should be set only after create log success.
2851 Description: Accessing a block bevice can re-enable I/O when Lustre is
2852 tearing down a device.
2853 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2858 Description: service threads may hog cpus when there are a lot of requests
2859 Details : Insert cond_resched to give other threads a chance to use some CPU
2864 Description: the cat log was not initialized in recovery
2865 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2866 unlink log on mds will not be initialized until mds post
2867 recovery. And also in mds post recovery, the unlink log will
2868 initialization will be done asynchronausly, so there will be race
2869 between add unlink log and unlink log initialization.
2873 Description: brw_stats were being printed incorrectly
2874 Details : brw_stats were being printed as log2 but all of them were not
2875 recorded as log2. Also remove some code duplication arising from
2876 filter_tally_{read,write}.
2880 Frequency : rare, only in recovery.
2881 Description: ASSERTION(req->rq_type != LI_POISON) failed
2882 Details : imp_lock should be held while iterating over imp_sending_list for
2883 prevent destroy request after get timeout in ptlrpc_queue_wait.
2887 Description: replay-single.sh test 52 fails
2888 Details : A lock's skiplist need to be cleanup when it being unlinked
2889 from its resource list.
2893 Description: Short directio read returns full requested size rather than
2895 Details : Direct I/O operations should return actual amount of bytes
2896 transferred rather than requested size.
2898 Severity : enhancement
2900 Description: metadata RPC reduction (e.g. for rm performance)
2901 Details : decrease the amount of synchronous RPC between clients and servers
2902 by canceling conflicing lock before the operation on the client
2903 and packing thier handles into the main operation RPC to server.
2905 Severity : enhancement
2907 Description: Async OSC create to avoid the blocking unnecessarily.
2908 Details : If a OST has no remain object, system will block on the creating
2909 when need to create a new object on this OST. Now, ways use
2910 pre-created objects when available, instead of blocking on an
2911 empty osc while others are not empty. If we must block, we block
2912 for the shortest possible period of time.
2916 Description: improve handling recoverable errors
2917 Details : if request processig with error which can be recoverable on server
2918 request should be resend, otherwise page released from cache and
2921 Severity : enhancement
2923 Description: refine locking for avoid write wrong info into lov_objid file
2924 Details : fix possible races with add new target and write/update data in
2927 --------------------------------------------------------------------------------
2929 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2934 Frequency : on some architectures
2936 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2937 Details : Uninitialized flags sometimes cause configuration commands to
2941 Frequency : patchless clients only
2943 Description: missing __iget() symbol export
2944 Details : The __iget() symbol export is missing. To avoid the need for
2945 this on patchless clients the deathrow inode reaper is turned
2946 off, and we depend on the VM to clean up old inodes. This
2947 dependency was during via the fix for bug 12181.
2949 --------------------------------------------------------------------------------
2951 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2953 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2954 INTEROPERATE with older versions automatically. In many cases a
2955 special upgrade step is needed. Please read the
2956 user documentation before upgrading any part of a live system.
2957 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2958 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2959 * WARNING: Lustre configuration and startup changes are required with
2960 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2964 Severity : enhancement
2966 Description: MountConf
2967 Details : Lustre configuration is now managed via mkfs and mount
2968 commands instead of lmc and lconf. New obd types (MGS, MGC)
2969 are added for dynamic configuration management. See
2970 https://mail.clusterfs.com/wikis/lustre/MountConf for
2973 Severity : enhancement
2975 Description: dynamic OST addition
2976 Details : OSTs can now be added to a live filesystem
2978 Severity : enhancement
2980 Description: startup order invariance
2981 Details : MDTs and OSTs can be started in any order. Clients only
2982 require the MDT to complete startup.
2984 Severity : enhancement
2986 Description: parallel, asynchronous orphan cleanup
2987 Details : orphan cleanup is now performed in separate threads for each
2988 OST, allowing parallel non-blocking operation.
2990 Severity : enhancement
2992 Description: optimized stripe assignment
2993 Details : stripe assignments are now made based on ost space available,
2994 ost previous usage, and OSS previous usage, in order to try
2995 to optimize storage space and networking resources.
2997 Severity : enhancement
2999 Description: Permanently set tunables
3000 Details : All writable /proc/fs/lustre tunables can now be permanently
3001 set on a per-server basis, at mkfs time or on a live system.
3003 Severity : enhancement
3005 Description: Lustre message v2
3006 Details : Add lustre message format v2.
3008 Severity : enhancement
3010 Description: client OST exclusion list
3011 Details : Clients can be started with a list of OSTs that should be
3012 declared "inactive" for known non-responsive OSTs.
3016 Description: SPEC SFS validation failure on NFS v2 over lustre.
3017 Details : Changes the blocksize for regular files to be 2x RPC size,
3018 and not depend on stripe size.
3020 Severity : enhancement
3022 Description: Multiple MD RPCs in flight.
3023 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3024 To avoid overly-overloading MDS, introduce a limit on number of
3025 MDS RPCs in flight for a single client and add /proc controls
3026 to adjust this limit.
3028 Severity : enhancement
3030 Description: client read/write statistics
3031 Details : Add client read/write call usage stats for performance
3032 analysis of user processes.
3033 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3034 file access. extents_stats shows chunk size distribution.
3035 extents_stats_per_process show chunk size distribution per
3038 Severity : enhancement
3040 Description: mds statistics
3041 Details : Add detailed mds operations statistics in
3042 /proc/fs/lustre/mds/*/stats.
3046 Description: Failure of copying files with lustre special EAs.
3047 Details : Client side always return success for setxattr call for lustre
3048 special xattr (currently only "trusted.lov").
3053 Description: Refcount LNET uuids
3054 Details : The global LNET uuid list grew linearly with every startup;
3055 refcount repeated list entries instead of always adding to
3058 Severity : enhancement
3060 Description: Dynamic service threads
3061 Details : Within a small range, start extra service threads
3062 automatically when the request queue builds up.
3065 Frequency : mixed-endian client/server environments
3067 Description: mixed-endian crashes
3068 Details : The new msg_v2 system had some failures in mixed-endian
3071 Severity : enhancement
3073 Description: Easy OST removal
3074 Details : OSTs can be permanently deactivated with e.g. 'lctl
3075 conf_param lustre-OST0001.osc.active=0'
3077 Severity : enhancement
3079 Description: MGS proc entries
3080 Details : Added basic proc entries for the MGS showing what filesystems
3083 Severity : enhancement
3085 Description: provide MGS failover
3086 Details : Added config lock reacquisition after MGS server failover.
3088 Severity : enhancement
3090 Description: add Linux 2.4 support
3091 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3095 Description: a large application tries to do I/O to the same resource and dies
3096 in the middle of it.
3097 Details : Check the req->rq_arrival time after the call to
3098 ost_brw_lock_get(), but before we do anything about
3099 processing it & sending the BULK transfer request. This
3100 should help move old stale pending locks off the queue as
3101 quickly as obd_timeout.
3104 Frequency : when an incorrect nid is specified during startup
3106 Description: ptlrpc connect to non-existant node causes kernel crash
3107 Details : LNET can't be re-entered from an event callback, which
3108 happened when we expire a message after the export has been
3109 cleaned up. Instead, hand the zombie cleanup off to another
3112 Severity : enhancement
3114 Description: plain/inodebits lock performance improvement
3115 Details : Grouping plain/inodebits in granted list by their request modes
3116 and bits policy, thus improving the performance of search through
3120 Frequency : only if OST filesystem is corrupted
3122 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3123 Details : for a short time RPCs with bulk IO are in the replay list,
3124 but replay of bulk IOs is unimplemented. If the OST filesystem
3125 is corrupted due to disk cache incoherency and then replay is
3126 started it is possible to trip an assertion. Avoid putting
3127 committed RPCs into the replay list at all to avoid this issue.
3130 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3132 Description: System hang on startup
3133 Details : This bug allowed the liblustre (e.g. catamount) client to
3134 return to the app before handling all startup RPCs. This
3135 could leave the node unresponsive to lustre network traffic
3136 and manifested as a server ptllnd timeout.
3138 Severity : enhancement
3140 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3141 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3142 boolean to control whether to print peer debug info when a
3143 client's RPC times out.
3146 Frequency : only for kernels with patches from Lustre below 1.4.3
3148 Description: Remove old rdonly API
3149 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3152 Frequency : only for devices with external journals
3154 Description: Set external device read-only also
3155 Details : During a commanded failover stop, we set the disk device
3156 read-only while the server shuts down. We now also set any
3157 external journal device read-only at the same time.
3160 Frequency : when upgrading from 1.4 while trying to change parameters
3162 Description: The wrong (new) MDC name was used when setting parameters for
3163 upgraded MDT's. Also allows changing of OSC (and MDC)
3164 parameters if --writeconf is specified at tunefs upgrade time.
3167 Frequency : when setting specific ost indicies
3169 Description: QOS code breaks on skipped indicies
3170 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3171 created with --index need not be sequential.
3173 Severity : enhancement
3175 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3176 Details : The uninit_groups feature works in conjunction with the kernel
3177 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3178 up the pass1 processing of e2fsck. This is a read-only feature
3179 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3180 mount filesystems that have had this feature enabled.
3182 Severity : enhancement
3184 Description: Improve multi-block allocation algorithm to avoid fragmentation
3185 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3186 allocation locality and avoid filesystem fragmentation.
3188 ------------------------------------------------------------------------------
3190 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3192 * Support for kernels:
3193 2.4.21-47.0.1.EL (RHEL 3)
3194 2.6.5-7.283 (SLES 9)
3195 2.6.9-42.0.10.EL (RHEL 4)
3196 2.6.12.6 vanilla (kernel.org)
3197 2.6.16.27-0.9 (SLES 10)
3198 * Recommended e2fsprogs version: 1.39.cfs5
3200 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3204 Frequency : occasional, depends on client load and configuration
3205 Bugzilla : 12181, 12203
3206 Description: data loss for recently-modified files
3208 Details : In some cases it is possible that recently written or created
3209 files may not be written to disk in a timely manner (this should
3210 normally be within 30s unless client IO load is very high).
3211 The problem appears as zero-length files or files that are a
3212 multiple of 1MB in size after a client crash or client eviction
3213 that are missing data at the end of the file.
3215 This problem is more likely to be hit on clients where files are
3216 repeatedly created and unlinked in the same directory, clients
3217 have a large amount of RAM, have many CPUs, the filesystem has
3218 many OSTs, the clients are rebooted frequently, and/or the files
3219 are not accessed by other nodes after being written.
3221 The presence of the problem can be detected by looking at
3222 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3223 smaller than the second (nr_unused) then dirty files will not
3224 be flushed automatically to disk. "sync; sleep 10" should be
3225 run several times on the node before unmounting it to update
3226 Lustre (this is also safe to run on nodes without this problem).
3228 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3229 that can cause this same problem, so customers using that kernel
3230 also need to update the kernel in addition to Lustre. In order
3231 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3233 It is normal that files written just before a client crash (less
3234 than 30s) may not yet have been flushed to disk, even for local
3238 Frequency : frequent on thin XT3 nodes
3240 Description: UUID collision on thin XT3 Linux nodes
3241 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3242 randomly, since we relied on an insufficiently-seeded PRNG.
3247 Description: OSS hangs after "All ost request buffers busy"
3248 Details : A deadlock between quota and journal operations caused OSS
3249 hangs after printing "All ost request buffers busy."
3252 Frequency : always on liblustre builds
3254 Description: Cleanup compiler warnings on liblustre
3257 Frequency : always on liblustre builds on XT3
3259 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3263 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3264 > 32000 subdirectories
3265 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3266 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3267 subdirectory count crosses 32000. This will aid e2fsck to
3268 correctly handle more than 32000 subdirectories.
3271 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3273 Description: System hang on startup
3274 Details : This bug allowed the liblustre (e.g. catamount) client to
3275 return to the app before handling all startup RPCs. This
3276 could leave the node unresponsive to lustre network traffic
3277 and manifested as a server ptllnd timeout.
3279 Severity : enhancement
3281 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3282 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3283 boolean to control whether to print peer debug info when a
3284 client's RPC times out.
3289 Description: make O_SYNC working on 2.6 kernels
3290 Details : 2.6 kernels use different method for mark pages for write,
3291 so need add a code to lustre for O_SYNC work.
3296 Description: Failure to close file and release space on NFS
3297 Details : Put inode details into lock acquired in ll_intent_file_open.
3298 Use mdc_intent_lock in ll_intent_open to properly
3299 detect all kind of errors unhandled by mdc_enqueue.
3304 Description: proc file read during shutdown sometimes raced obd removal,
3306 Details : Add lock to prevent obd access after proc file removal.
3309 Frequency : Only for files larger than 4GB on 32-bit clients.
3311 Description: improperly doing page alignment of locks
3312 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3313 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3318 Description: RPCs being resent when they shouldn't be
3319 Details : Some RPCs that should not be resent are being resent. This
3320 can cause inconsistencies in the RPC state machine. Do not
3321 resend such requests.
3324 Frequency : rare, only with NFS export
3326 Description: Crash on NFS re-export node
3327 Details : under very unusual load conditions an assertion is hit in
3328 ll_intent_file_open()
3331 Frequency : only if OST filesystem is corrupted
3333 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3334 Details : for a short time RPCs with bulk IO are in the replay list,
3335 but replay of bulk IOs is unimplemented. If the OST filesystem
3336 is corrupted due to disk cache incoherency and then replay is
3337 started it is possible to trip an assertion. Avoid putting
3338 committed RPCs into the replay list at all to avoid this issue.
3343 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3344 Details : Large single O_DIRECT read and write calls can fail to allocate
3345 a sufficiently large buffer to process the request. In case of
3346 allocation failure the allocation is retried with a smaller
3347 buffer and broken into smaller requests.
3349 Severity : enhancement
3351 Description: Add -o localflock option to simulate old noflock behaviour.
3352 Details : This will achieve local-only flock/fcntl locks coherentness.
3357 Description: versioning check is incomplete
3358 Details : Checking the version difference of client vs. server, report
3359 error if the gap is too big.
3364 Description: add support PG_writeback bit
3365 Details : add support for PG_writeback bit for Lustre, for more carefull
3366 work with page cache in 2.6 kernel. This also fix some deadlocks
3367 and remove hack for work O_SYNC with 2.6 kernel.
3369 Severity : enhancement
3371 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3372 Details : The uninit_groups feature works in conjunction with the kernel
3373 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3374 up the pass1 processing of e2fsck. This is a read-only feature
3375 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3376 mount filesystems that have had this feature enabled.
3378 Severity : enhancement
3380 Description: Improve multi-block allocation algorithm to avoid fragmentation
3381 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3382 allocation locality and avoid filesystem fragmentation.
3384 ------------------------------------------------------------------------------
3386 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3388 * Support for kernels:
3389 2.6.9-42.0.3.EL (RHEL 4)
3390 2.6.5-7.276 (SLES 9)
3391 2.4.21-47.0.1.EL (RHEL 3)
3392 2.6.12.6 vanilla (kernel.org)
3393 2.6.16.21-0.8 (SLES10)
3394 * Recommended e2fsprogs version: 1.39.cfs2-0
3396 * The backwards-compatible /proc/sys/portals symlink has been removed
3397 in this release. Before upgrading, please ensure that you change
3398 any configuration scripts or /etc/sysctl.conf files that access
3399 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3400 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3401 in advance of the upgrade on any system running Lustre 1.4.6 or
3402 newer, since /proc/sys/lnet was added in that version.
3403 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3407 Frequency : only when quota is used
3409 Description: avoid scanning export list for quota master
3410 Details : Change the algorithms to avoid scanning export list in order
3411 to improve the efficiency.
3414 Frequency : MDS failover only, very rarely
3416 Description: "went back in time" messages on mds failover
3417 Details : The greatest transno may be lost when the current operation
3418 finishes with an error (transno==0) and the client's last_rcvd
3419 record is over-written. Save the greatest transno in the
3420 mds_last_transno for this case.
3423 Frequency : always for specific kernels and striping counts
3425 Description: client may get "Matching packet too big" without ACL support
3426 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3427 when trying to access files in certain configurations. The
3428 clients should in fact be denied when mounting because they do
3429 not understand ACLs.
3432 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3434 Description: many clients connecting with IO in progress causes connect timeouts
3435 Details : Avoid synchronous journal commits to avoid delays caused by many
3436 clients connecting/disconnecting when bulk IO is in progress.
3437 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3438 OST_IO_PORTAL to avoid delays behind potentially many pending
3442 Frequency : occasionally with multiple writers to a single file
3444 Description: shared writes to file may result in wrong size reported by stat()
3445 Details : Allow growing of kms when extent lock is cancelled
3448 Frequency : always with random mmap IO to multi-striped file
3450 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3451 Details : Only if the hole is at the end of OST object so that kms is too
3452 small. Fix is to increase kms accordingly in ll_nopage.
3455 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3457 Description: writes to a missing object would leak memory on the OST
3458 Details : If there is an inconsistency between the MDS and OST filesystems,
3459 such that the MDS references an object that doesn't exist, writes
3460 to that object will leak memory due to incorrect cleanup in the
3461 error handling path, eventually running out of memory on the OST.
3466 Description: Creating too long symlink causes lustre errors
3467 Details : Check symlink and name lengths before sending requests to MDS.
3470 Frequency : only if flock is enabled (not on by default)
3472 Description: posix locks not released on fd closure on 2.6.9+
3473 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3474 kernels, this caused such locks not to be released on fd close and
3475 then assertions on fs unmount about still used locks.
3478 Frequency : MDS failover only, very rarely
3480 Description: clients may get ASSERTION(granted_lock != NULL)
3481 Details : When request was taking a long time, and a client was resending
3482 a getattr by name lock request. The were multiple lock requests
3483 with the same client lock handle and
3484 mds_getattr_name->fixup_handle_for_resent_request found one of the
3485 lock handles but later failed with ASSERTION(granted_lock != NULL).
3490 Description: handle->h_buffer_credits > 0, assertion failure
3491 Details : h_buffer_credits is zero after truncate, causing assertion
3492 failure. This patch extends the transaction or creates a new
3496 Frequency : NFS re-export or patchless client
3497 Bugzilla : 11179, 10796
3498 Description: Crash on NFS re-export node (__d_move)
3499 Details : We do not want to hash the dentry if we don't have a lock.
3500 But if this dentry is later used in d_move, we'd hit uninitialised
3501 list head d_hash, so we just do this to init d_hash field but
3502 leave dentry unhashed.
3505 Frequency : NFS re-export or patchless client
3507 Description: NFS exports has problem with symbolic link
3508 Details : lustre client didn't properly install dentry when re-exported
3509 to NFS or running patchless client.
3512 Frequency : NFS re-export or patchless client
3514 Description: Various nfs/patchless fixes.
3515 Details : fixes reuse disconected alias for lookup process - this fixes
3516 warning "find_exported_dentry: npd != pd",
3517 fix permission error with open files at nfs.
3518 fix apply umask when do revalidate.
3521 Frequency : occasional
3523 Description: Crash on NFS re-export node
3524 Details : calling clear_page() on the wrong pointer triggered oops in
3525 generic_mapping_read().
3528 Frequency : rarely, using O_DIRECT IO
3530 Description: unaligned directio crashes client with LASSERT
3531 Details : check for unaligned buffers before trying any requests.
3534 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3536 Description: stale data returned from RAID cache
3537 Details : If only a small amount of IO is done to the RAID device before
3538 reading it again it is possible to get stale data from the RAID
3539 cache instead of reading it from disk.
3542 Frequency : always for sles10 kernel
3544 Description: sles10 support
3545 Details : ll_follow_link: compile fixes and using of nd_set_link
3546 under newer kernels.
3549 Frequency : depends on arch, kernel and compiler version, always on sles10
3552 Description: recursive or deep enough symlinks cause stack overflow
3553 Details : getting rid of large stack-allocated variable in
3557 Frequency : depends on hardware
3559 Description: lustre write performance loss in the SLES10 kernel
3560 Details : the performance loss is caused by using of write barriers in the
3561 ext3 code. The SLES10 kernel turns barrier support on by
3562 default. The fix is to undo that change for ldiskfs.
3564 ------------------------------------------------------------------------------
3566 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3568 * Support for kernels:
3569 2.6.9-42.0.3EL (RHEL 4)
3570 2.6.5-7.276 (SLES 9)
3571 2.4.21-47.0.1.EL (RHEL 3)
3572 2.6.12.6 vanilla (kernel.org)
3576 Frequency : quota enabled and large files being deleted
3578 Description: releasing more than 4GB of quota at once hangs OST
3579 Details : If a user deletes more than 4GB of files on a single OST it
3580 will cause the OST to spin in an infinite loop. Release
3581 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3586 Description: statfs data retrieved from /proc may be stale or zero
3587 Details : When reading per-device statfs data from /proc, in the
3588 {kbytes,files}_{total,free,avail} files, it may appear
3589 as zero or be out of date.
3592 Frequency : systems with MD RAID1 external journal devices
3594 Description: lconf's call to blkid is confused by RAID1 journal devices
3595 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3596 of returning all block devices that match the journal UUID.
3599 Frequency : always, for aggregate stripe size over 4GB
3601 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3602 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3603 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3606 Frequency : always if "lfs find" used on a local file/directory
3608 Description: "lfs find" segfaults if used on a local file/directory
3609 Details : The case where a directory component was not specified wasn't
3610 handled correctly. Handle this properly.
3613 Frequency : always on ppc64
3615 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3616 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3617 perform endianness conversion.
3620 Frequency : rarely (truncate to non-zero file size after write under load)
3621 Bugzilla : 10730, 10687
3622 Description: Files padded with zeros to next 4K multiple
3623 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3624 it is possible that files that are truncated to a non-zero size
3625 immediately after being written are filled with zero bytes beyond
3626 the truncated size. No file data is lost.
3628 Severity : enhancement
3630 Description: Allow recovery/failover for liblustre clients.
3631 Details : liblustre clients were unaware of failover configurations until
3634 Severity : enhancement
3636 Description: user file locks should fail when not mounting with flock option
3637 Details : Set up an error-returning stub in ll_file_operations.lock field
3638 to prevent incorrect behaviour when client is mounted without
3639 flock option. Also, set up properly f_op->flock field for
3643 Frequency : always on ia64
3645 Description: "lfs df" loops on printing out MDS statfs information
3646 Details : The obd_ioctl_data was not initialized and in some systems
3647 this caused a failure during the ioctl that did not return
3648 an error. Initialize the struct and return an error on failure.
3651 Frequency : SLES 9 only
3653 Description: Error of copying files with lustre special EAs as root
3654 Details : Client side always return success for setxattr call for lustre
3655 special xattr (currently only "trusted.lov").
3658 Frequency : rarely on clusters with both ia64+i386 clients
3660 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3661 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3662 clients are doing concurrent writes to the same file it is
3663 possible that the ia64 clients may LASSERT because the OST
3664 extent locks are not PAGE_SIZE aligned. Ensure that grown
3665 locks are always aligned on the request boundary.
3668 Frequency : specific use, occasional
3670 Description: Overwriting in use executable truncates on-disk binary image
3671 Details : If one node attempts to overwrite an executable in use by
3672 another node, we now correctly return ETXTBSY instead of
3673 truncating the file.
3675 Severity : enhancement
3677 Description: Async OSC create to avoid the blocking unnecessarily.
3678 Details : If a OST has no remain object, system will block on the creating
3679 when need to create a new object on this OST. Now, ways use
3680 pre-created objects when available, instead of blocking on an
3681 empty osc while others are not empty. If we must block, we block
3682 for the shortest possible period of time.
3687 Description: chmod on Lustre root is propagated to other clients
3688 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3689 invalid by the follow_mount time.
3692 Frequency : liblustre clients only
3694 Description: Race in 'instant cancel' lock handling could lead to such locks
3695 never to be granted in case of SMP MDS
3696 Details : Do not destroy not yet granted but cbpending locks in
3700 Frequency : replay/resend of open
3702 Description: non null lock assetion failure in mds_intent_policy
3703 Details : Trying to replay/resend lockless open requests resulted in
3704 mds_open() returning 0 with no lock. Now it sets a flag if
3705 a lock is going to be returned.
3707 Severity : enhancement
3709 Description: Checksum enhancements
3710 Details : New checksum enhancements allow for resending RPCs that failed
3713 Severity : enhancement
3715 Description: Tunables on number of dirty pages in cacche
3716 Details : Allow to set limit on number of dirty pages cached.
3721 Description: client crash on unmount - lock still has references
3722 Details : In some error handling cases it was possible to leak a lock
3723 reference on a client while accessing a file. This was not
3724 harmful to the client during operation, but would cause the
3725 client to crash when the filesystem is unmounted.
3728 Frequency : specific case, rare
3730 Description: ETXTBSY on mds though file not in use by client
3731 Details : ETXTBSY is no longer incorrectly returned when attempting to
3732 chmod or chown a directory that the user previously tried to
3733 execute or a currently-executing binary.
3736 Frequency : extremely rare except on liblustre-based clients
3738 Description: Lustre space not freed when files are deleted
3739 Details : Clean up open-unlinked files after client eviction. Previously
3740 the unlink was skipped and the files remained as orphans.
3745 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3746 Details : In some cases it was possible to send a blocking callback to
3747 a client doing a glimpse, even though that client didn't get
3748 a lock granted. When the glimpse lock is cancelled on the OST
3749 the freed lock is left on the waiting list and corrupted the list.
3752 Frequency : all core dumps
3754 Description: Broke core dumps to lustre
3755 Details : Negative dentry may be unhashed if parent does not have UPDATE
3756 lock, but some callers, e.g. do_coredump, expect dentry to be
3757 hashed after successful create, hash it in ll_create_it.
3759 ------------------------------------------------------------------------------
3761 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3763 * Support for kernels:
3764 2.6.9-42.0.2.EL (RHEL 4)
3765 2.6.5-7.276 (SLES 9)
3766 2.4.21-40.EL (RHEL 3)
3767 2.6.12.6 vanilla (kernel.org)
3771 Frequency : always on RHEL 3
3773 Description: Number of open files grows over time
3774 Details : The number of open files grows over time, whether or not
3775 Lustre is started. This was due to a filp leak introduced
3776 by one of our kernel patches.
3778 ------------------------------------------------------------------------------
3780 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3782 * Support for kernels:
3783 2.6.9-42.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)
3791 Bugzilla : 5719, 9635, 9792, 9684
3792 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3793 Details : If a server is under heavy load and cannot reply to new
3794 connection requests before the client resends the (re)connect,
3795 the connection handling code can behave badly if two service
3796 threads are concurrently handing separate (re)connections from
3797 the same client. Add better locking to the connection handling
3798 code, and ensure that only a single connection will be processed
3799 for a given client UUID, even if the lock is dropped.
3801 Severity : enhancement
3803 Description: add TCP zero-copy support to kernel
3804 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3805 sends if the hardware supports scatter-gather and checksumming.
3806 This allows socklnd to do client-write and server-read more
3807 efficiently and reduce CPU utilization from skbuf copying.
3810 Frequency : only if NFS exporting from client
3812 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3813 Details : If setting attributes on a file created under NFS that had
3814 never been opened it would be possible to oops the client
3815 if the file had no objects.
3818 Frequency : always for liblustre
3820 Description: liblustre client does MDS+OSTs setattr RPC for each write
3821 Details : When doing a write from a liblustre client, the client
3822 incorrectly issued an RPC to the MDS and each OST the file was
3823 striped over in order to update the timestamps. When writing
3824 with small chunks and many clients this could overwhelm the MDS
3825 with RPCs. In all cases it would slow down the write because
3826 these RPCs are unnecessary.
3828 Severity : enhancement
3830 Description: allow number of MDS service threads to be changed at module load
3831 Details : It is now possible to change the number of MDS service threads
3832 running. Adding "options mds mds_num_threads={N}" to the MDS's
3833 /etc/modprobe.conf will set the number of threads for the next
3834 time Lustre is restarted (assuming the "mds" module is also
3835 reloaded at that time). The default number of threads will
3836 stay the same, 32 for most systems.
3841 Description: OST crash if filesystem is unformatted or corrupt
3842 Details : If an OST is started on a device that has never been formatted
3843 or if the filesystem is corrupt and cannot even mount then the
3844 error handling cleanup routines would dereference a NULL pointer.
3849 Description: NULL pointer deref in llap_from_page.
3850 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3851 incorrect) mapping if the page was truncated/reclaimed while it was
3852 searched for. Check for this condition and skip such pages when
3853 doing readahead. Introduce extra check to llap_from_page() to
3854 verify page->mapping->host is non-NULL (so page is not anonymous).
3857 Frequency : Sometimes when using sys_sendfile
3859 Description: "page not covered by a lock" warnings from ll_readpage
3860 Details : sendfile called ll_readpage without right page locks present.
3861 Now we introduced ll_file_sendfile that does necessary locking
3862 around call to generic_file_sendfile() much like we do in
3866 Frequency : with certain MDS communication failures at client mount time
3868 Description: NULL pointer deref after failed client mount
3869 Details : a client connection request may delayed by the network layer
3870 and not be sent until after the PTLRPC layer has timed out the
3871 request. If the client fails the mount immediately it will try
3872 to clean up before the network times out the request. Add a
3873 reference from the request import to the obd device and delay
3874 the cleanup until the network drops the request.
3877 Frequency : occasionally during client (re)connect
3879 Description: assertion failure during client (re)connect
3880 Details : processing a client connection request may be delayed by the
3881 client or server longer than the client connect timeout. This
3882 causes the client to resend the connection request. If the
3883 original connection request is replied in this interval, the
3884 client may trip an assertion failure in ptlrpc_connect_interpret()
3885 which thought it would be the only running connect process.
3888 Frequency : only with obd_echo servers and clients that are rebooted
3890 Description: kernel BUG accessing uninitialized data structure
3891 Details : When running an obd_echo server it did not start the ping_evictor
3892 thread, and when a client was evicted an uninitialized data
3893 structure was accessed. Start the ping_evictor in the RPC
3894 service startup instead of the OBD startup.
3896 Severity : enhancement
3897 Bugzilla : 10193 (patchless)
3898 Description: Remove dependency on various unexported kernel interfaces.
3899 Details : No longer need reparent_to_init, exit_mm, exit_files,
3900 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3903 Frequency : rare (only users of deprecated and unsupported LDAP config)
3905 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3906 Details : LDAP apparently contains 'lustreName' attributes instead of
3907 'name'. A simple remapping of the name is sufficient.
3910 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3912 Description: waiting_locks_callback trips kernel BUG if client is evicted
3913 Details : Running with the dump_on_timeout debug flag turned on makes
3914 it possible that the waiting_locks_callback() can try to dump
3915 the Lustre kernel debug logs from an interrupt handler. Defer
3916 this log dumping to the expired_lock_main() thread.
3918 Severity : enhancement
3920 Description: Support NFS exporting on 2.6 kernels.
3921 Details : Implement non-rawops metadata methods for NFS server to use without
3922 changing NFS server code.
3925 Frequency : very rare (synthetic metadata workload only)
3927 Description: two racing renames might cause an MDS thread to deadlock
3928 Details : Running the "racer" program may cause one MDS thread to rename
3929 a file from being the source of a rename to being the target of
3930 a rename at exactly the same time that another thread is doing
3931 so, and the second thread has already enqueued these locks after
3932 doing a lookup of the target and is trying to relock them in
3933 order. Ensure that we don't try to re-lock the same resource.
3936 Frequency : only very large systems with liblustre clients
3938 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3939 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3940 all OSTs in parallel. This allows the eviction of stale liblustre
3941 clients to proceed much faster than if they were done in series,
3942 and also offers similar improvements for other set_info RPCs.
3947 Description: excessive CPU usage during initial read phase on client
3948 Details : During the initial read phase on a client, it would agressively
3949 retry readahead on the file, consuming too much CPU and impacting
3950 performance (since 1.4.5.8). Improve the readahead algorithm
3951 to avoid this, and also improve some other common cases (read
3952 of small files in particular, where "small" is files smaller than
3953 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3958 Description: MDS crash when receiving packet with unknown intent.
3959 Details : Do not LBUG in unknown intent case, just return -EFAULT
3961 Severity : enhancement
3962 Bugzilla : 9293, 9385
3963 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3964 Details : Do not serialize getattr (non-intent version) and statfs.
3967 Frequency : occasional, when OST network is overloaded/intermittent
3969 Description: client evicted by OST after bulk IO timeout
3970 Details : If a client sends a bulk IO request (read or write) the OST
3971 may evict the client if it is unresposive to its data GET/PUT
3972 request. This is incorrect if the network is overloaded (takes
3973 too long to transfer the RPC data) or dropped the OST GET/PUT
3974 request. There is no need to evict the client at all, since
3975 the pinger and/or lock callbacks will handle this, and the
3976 client can restart the bulk request.
3979 Frequency : Always when mmapping file with no objects
3981 Description: client crashes when mmapping file with no objects
3982 Details : Check that we actually have objects in a file before doing any
3983 operations on objects in ll_vm_open, ll_vm_close and
3989 Description: Request leak when working with deleted CWD
3990 Details : Introduce advanced request refcount tracking for requests
3991 referenced from lustre intent.
3993 Severity : Enhancement
3995 Description: Cache open file handles on client.
3996 Details : MDS now will return special lock along with openhandle, if
3997 requested and client is allowed to hold openhandle, even if unused,
3998 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3999 closing files for every read/write openration.
4001 Severity : Enhancement
4003 Description: Cache open negative dentries on client when possible.
4004 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4005 negative dentries on lock revocation.
4010 Description: Remounting a client read-only wasn't possible with a zconf mount
4011 Details : It wasn't possible to remount a client read-only with llmount.
4013 Severity : enhancement
4014 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4015 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4016 MPICH is included to add Lustre-specific ADIO interfaces.
4017 This is based closely on the UFS ADIO layer and only differs
4018 in file creation, in order to allow the OST striping to be set.
4019 This is user-contributed code and not supported by CFS.
4024 Description: extended inode attributes (immutable, append-only) work improperly
4025 when 2.4 and 2.6 kernels are used on client/server or vice versa
4026 Details : Introduce kernel-independent values for these flags.
4028 Severity : enhancement
4031 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4032 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4033 granularity. Now, Fractional MB granularity is be supported,
4034 this is very useful for low memory system.
4036 Severity : enhancement
4038 Description: Getattr by fid
4039 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4040 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4041 have locks and avoids getattr rpc requests for it on every stat.
4044 Frequency : Always, for filesystems larger than 2TB
4046 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4047 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4048 filesystems larger than 2TB to be mounted with mballoc enabled.
4051 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4053 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4054 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4055 filesystems larger than 2TB is mounted then the kernel will
4056 silently corrupt the start of the filesystem. CONFIG_LBD is
4057 enabled for all CFS-supported kernels, but the possibility of
4058 this happening with a modified kernel config exists.
4060 Severity : enhancement
4062 Description: add client O_DIRECT support for 2.6 kernels
4063 Details : It is now possible to do O_DIRECT reads and writes to files
4064 in the Lustre client mountpoint on 2.6 kernel clients.
4066 Severity : enhancement
4068 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4069 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4070 parallel, not waiting for response from every OST before sending
4071 a rpc to the next OST.
4076 Description: setattr vs write race when updating file timestamps
4077 Details : Client processes that update a file timestamp into the past
4078 right after writing to the file (e.g. tar) it is possible that
4079 the updated file modification time can be reset to the current
4080 time due to a race between processing the setattr and write RPC.
4082 Severity : enhancement
4084 Description: Bring 'lfs find' closer in line with regular Linux find.
4085 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4086 -print0 options and obtains all the needed info through the lustre
4089 Severity : enhancement
4091 Description: support up to 1024 configured devices on one node
4092 Details : change obd_dev array from statically allocated to dynamically
4093 allocated structs as they are first used to reduce memory usage
4098 Description: Flush dirty partially truncated pages during truncate
4099 Details : Immediatelly flush partially truncated pages in filter_setattr,
4100 this way we completely avoid having any pages in page cache on OST
4101 and can retire ugly workarounds during writes to flush such pages.
4106 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4107 Details : protect inode from truncation within vfs_unlink() context
4108 just take a reference before calling vfs_unlink() and release it
4109 when parent's i_sem is free.
4114 Description: last_id value checked outside lock on OST caused LASSERT failure
4115 Details : If there were multiple MDS->OST object precreate requests in
4116 flight, it was possible that the OST's last object id was checked
4117 outside a lock and incorrectly tripped an assertion. Move checks
4118 inside locks, and discard old precreate requests.
4121 Frequency : always, if extents are used on OSTs
4123 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4124 Details : OSTs using the extents format would not zero the high 16 bits of
4125 the index physical block number. This is not a problem for any
4126 OST filesystems smaller than 16TB, and no kernels support ext3
4127 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4128 new/modified files) and can be fixed for existing filesystems
4129 with e2fsprogs-1.39-cfs1.
4134 Description: import connection selection may be incorrect if timer wraps
4135 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4136 import connections to be ignored if the 32-bit jiffies counter
4137 wraps. Use a 64-bit jiffies counter.
4140 Frequency : very large clusters immediately after boot
4142 Description: LNET request buffers exhausted under heavy short-term load
4143 Details : If a large number of client requests are generated on a service
4144 that has previously never seen so many requests it is possible
4145 that the request buffer growth cannot keep up with the spike in
4146 demand. Instead of dropping incoming requests, they are held in
4147 the LND until the RPC service can accept more requests.
4150 Frequency : Sometimes during replay
4152 Description: Assertion failure in ll_local_open after replay.
4153 Details : If replay happened on an open request reply before we were able
4154 to set replay handler, reply will become not swabbed tripping the
4155 assertion in ll_local_open. Now we set the handler right after
4156 recognising of open request
4159 Frequency : very rare
4161 Description: kernel reports "badness in vsnprintf"
4162 Details : Reading from the "recovery_status" /proc file in small chunks
4163 may cause a negative length in lprocfs_obd_rd_recovery_status()
4164 call to vsnprintf() (which is otherwise harmless). Exit early
4165 if there is no more space in the output buffer.
4167 Severity : enhancement
4169 Description: clear OBD RPC statistics by writing to them
4170 Details : It is now possible to clear the OBD RPC statistics by writing
4171 to the "stats" file.
4176 Description: Client mtime is not the same on different clients after utimes
4177 Details : In some cases, the client was using the utimes() syscall on
4178 a file cached on another node. The clients now validate the
4179 ctime from the MDS + OSTs to determine which one is right.
4184 Description: Inability to activate failout mode
4185 Details : lconf script incorrectly assumed that in pythong string's numeric
4186 value is used in comparisons.
4189 Frequency : always with multiple stripes per file
4191 Description: Inefficient object allocation for mutli-stripe files
4192 Details : When selecting which OSTs to stripe files over, for files with
4193 a stripe count that divides evenly into the number of OSTs,
4194 the MDS is always picking the same starting OST for each file.
4195 Return the OST selection heuristic to the original design.
4200 Description: mount failures may take full timeout to return an error
4201 Details : Under some heavy load conditions it is possible that a
4202 failed mount can wait for the full obd_timeout interval,
4203 possibly several minutes, before reporting an error.
4204 Instead return an error as soon as the status is known.
4206 Frequency : quota enabled and large files being deleted
4208 Description: releasing more than 4GB of quota at once hangs OST
4209 Details : If a user deletes more than 4GB of files on a single OST it
4210 will cause the OST to spin in an infinite loop. Release
4211 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4216 Description: statfs data retrieved from /proc may be stale or zero
4217 Details : When reading per-device statfs data from /proc, in the
4218 {kbytes,files}_{total,free,avail} files, it may appear
4219 as zero or be out of date.
4222 Frequency : systems with MD RAID1 external journal devices
4224 Description: lconf's call to blkid is confused by RAID1 journal devices
4225 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4226 of returning all block devices that match the journal UUID.
4229 Frequency : always, for aggregate stripe size over 4GB
4231 Description: assertion fails when trying to use 4GB stripe size
4232 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4233 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4236 Frequency : always on ppc64
4238 Description: the first write on an ext3 filesystem with mballoc got stuck
4239 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4240 perform endianness conversion.
4242 ------------------------------------------------------------------------------
4244 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4246 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4247 INTEROPERATE with older versions automatically. Please read the
4248 user documentation before upgrading any part of a live system.
4249 * WARNING: Lustre networking configuration changes are required with
4250 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4253 * Support for kernels:
4254 2.6.9-22.0.2.EL (RHEL 4)
4255 2.6.5-7.244 (SLES 9)
4256 2.6.12.6 vanilla (kernel.org)
4259 Severity : enhancement
4260 Bugzilla : 7981/8208
4261 Description: Introduced Lustre Networking (LNET)
4262 Details : LNET is new networking infrastructure for Lustre, it includes
4263 a reorganized network configuration mode (see the user
4264 documentation for full details) as well as support for routing
4265 between different network fabrics. Lustre Networking Devices
4266 (LNDs) for the supported network fabrics have also been
4267 created for this new infrastructure.
4269 Severity : enhancement
4270 Description: Introduced Access control lists
4271 Details : clients can set ACLs on files and directories in order to have
4272 more fine-grained permissions than the standard Unix UGO+RWX.
4273 The MDS must be started with the "-o acl" mount option.
4275 Severity : enhancement
4276 Description: Introduced filesystem quotas
4277 Details : Administrators may now establish per-user quotas on the
4280 Severity : enhancement
4282 Description: Configuration change for the XT3
4283 The PTLLND is now used to run Lustre over Portals on the XT3
4284 The configure option(s) --with-cray-portals are no longer used.
4285 Rather --with-portals=<path-to-portals-includes> is used to
4286 enable building on the XT3. In addition to enable XT3 specific
4287 features the option --enable-cray-xt3 must be used.
4292 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4293 Details : A race between allocating a new llcd and re-getting the llcd_lock
4294 allowed another thread to grab newly-allocated llcd.
4296 Severity : enhancement
4298 Description: 2.6 OST async journal commit and locking fix to improve performance
4299 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4300 2.6 kernels are now async as they already were in 2.4 kernels so
4301 that they can commit concurrently with the network bulk transfer.
4302 For block-allocated files the filter allocation semaphore is held
4303 to avoid filesystem fragmentation during allocation. BKL lock
4304 removed for 2.6 xattr operations where it is no longer needed.
4309 Description: lconf incorrectly determined whether two IP networks could talk
4310 Details : In some more complicated routing and multiple-network
4311 configurations, lconf will avoid trying to make a network
4312 connection to a disjoint part of the IP space. It was doing the
4313 math incorrectly for one set of cases.
4318 Description: Fix for potential infinite loop processing records in an llog.
4319 Details : If an llog record is corrupted/zeroed, it is possible to loop
4320 forever in llog_process(). Validate the llog record length
4321 and skip the remainder of the block on error.
4324 Frequency : occasional (liblustre only)
4326 Description: liblustre could not open files whose last component is a symlink
4327 Details : sysio_path_walk() would incorrectly pass the open intent to
4328 intermediate path components.
4331 Frequency : rare (liblustre only with non-standard tuning)
4332 Bugzilla : 7201 (7350)
4333 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4334 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4335 after it has been released, only do this for referenced locks
4337 Severity : enhancement
4339 Description: specifying an (invalid) directory default stripe_size of -1
4340 would reset the directory default striping
4341 Details : stripe_size -1 was used internally to signal directory stripe
4342 removal, now use "all default" to signal dir stripe removal
4343 as a directory striping of "all default" is not useful
4346 Frequency : common for large clusters running liblustre clients
4348 Description: doing an ls when liblustre clients are running is slow
4349 Details : sending a glimpse AST to a liblustre client waits for every AST
4350 to time out, as liblustre clients will not respond. Since they
4351 cannot cache data we refresh the OST lock LVB from disk instead.
4353 Severity : enhancement
4355 Description: doing an ls at the same time as file IO can be slow
4356 Details : enqueue and other "small" requests can be blocked behind many
4357 large IO requests. Create a new OST IO portal for non-IO
4358 requests so they can be processed faster.
4361 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4363 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4364 created file with mode 000
4365 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4366 allow an ftruncate() as a non-root user to a file with mode 000.
4367 Lustre now respects this flag to disable mode checks when
4368 truncating a file owned by the user
4371 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4374 Description: Revoking locks from clients that went dead or catatonic might take
4376 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4377 cancellation of such locks instant on servers without waiting for
4378 any reply from clients. Clients drops these locks when cancel
4379 notification from server is received without replying.
4382 Frequency : liblustre-only, when liblustre client dies or becomes busy
4384 Description: Doing ls on Linux clients can take a long time with active
4386 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4387 granting such locks to it in the first place if possible. Locks
4388 are taken by proxy on the OST during the read or write and
4389 dropped immediately afterward. Add connect flags handling, do
4390 not grant locks to liblustre clients for glimpse ASTs.
4392 Severity : enhancement
4394 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4395 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4396 sequential reads and behaves badly with applications doing random
4397 reads. Improve it by reading ahead at least read region, and
4398 avoiding excessive large RPC for small reads.
4400 Severity : enhancement
4402 Description: Creating more than 1000 files for a single job may cause a load
4403 imbalance on the OSTs if there are also a large number of OSTs.
4404 Details : qos_prep_create() uses an OST index reseed value that is an
4405 even multiple of the number of available OSTs so that if the
4406 reseed happens in the middle of the object allocation it will
4407 still utilize the OSTs as uniformly as possible.
4412 Description: OST or MDS may oops in ping_evictor_main()
4413 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4414 but doesn't restart at beginning of obd_exports_timed list
4417 Severity : enhancement
4419 Description: improve by-nid export eviction on the MDS and OST
4420 Details : allow multiple exports with the same NID to be evicted at one
4421 time without re-searching the exports list.
4424 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4426 Description: MDS may oops in groups_free()
4427 Details : in rare race conditions a newly allocated group_info struct is
4428 freed again, and this can be NULL. The 2.4 compatibility code
4429 for groups_free() checked for a NULL pointer, but 2.6 did not.
4432 Frequency : common for liblustre clients doing little filesystem IO
4433 Bugzilla : 9352, 7313
4434 Description: server may evict liblustre clients accessing contended locks
4435 Details : if a client is granted a lock or receives a completion AST
4436 with a blocking AST already set it would not reply to the AST
4437 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4440 Frequency : lfs setstripe, only systems with more than 160 OSTs
4442 Description: unable to set striping with a starting offset beyond OST 160
4443 Details : llapi_create_file() incorrectly limited the starting stripe
4444 index to the maximum single-file stripe count.
4447 Frequency : LDAP users only
4449 Description: lconf did not handle in-kernel recovery with LDAP properly
4450 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4452 Severity : enhancement
4454 Description: bind OST threads to NUMA nodes to improve performance
4455 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4456 node and do their allocations there to localize memory access
4458 Severity : enhancement
4460 Description: llmount can determine client NID directly from Myrinet (GM)
4461 Details : the client NID code from gmnalnid was moved directly into
4462 llmount, removing the need to use this or specifying the
4463 client NID explicitly when mounting GM clients with zeroconf
4466 Frequency : if client is started with down MDS
4468 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4469 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4470 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4471 ptlrpc_import_delayed_req() because we didn't check if the
4472 request was interrupted, and we also didn't break out of the
4473 event loop if there was no timeout
4478 Description: data loss during non-page-aligned writes to a single file from
4479 both multiple nodes and multiple threads on one node at same time
4480 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4481 inconsistency led to false short-reads, that were cached and later
4482 used by ->prepare_write() to fill in partially written page,
4483 leading to data loss.
4486 Frequency : always, if lconf --abort_recovery used
4488 Description: lconf --abort_recovery fails with 'Operation not supported'
4489 Details : lconf was attempting to abort recovery on the MDT device and not
4492 Severity : enhancement
4494 Description: remove cleanup logs
4495 Details : replace lconf-generated cleanup logs with lustre internal
4496 cleanup routines. Eliminates the need for client-cleanup and
4499 Severity : enhancement
4501 Description: add support for EAs (user and system) on lustre filesystems
4502 Details : it is now possible to store extended attributes in the Lustre
4503 client filesystem, and with the user_xattr mount option it
4504 is possible to allow users to store EAs on their files also
4506 Severity : enhancement
4508 Description: Add possibility (config option) to show minimal available OST free
4510 Details : When compiled with --enable-mindf configure option, statfs(2)
4511 (and so, df) will return least minimal free space available from
4512 all OSTs as amount of free space on FS, instead of summary of
4513 free spaces of all OSTs.
4515 Severity : enhancement
4517 Description: do not expand extent locks acquired on OST-side
4518 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4519 by server: they are not cached anyway.
4522 Frequency : when mmap is used/binaries executed from Lustre
4524 Description: Unmmap pages before throwing them away from read cache.
4525 Details : llap_shrink cache now attempts to unmap pages before discarding
4526 them (if unmapping failed - do not discard). SLES9 kernel has
4527 extra checks that trigger if this unmapping is not done first.
4532 Description: lconf didn't resolve symlinks before checking to see whether a
4533 given mountpoint was already in use
4536 Frequency : when migrating failover services
4537 Bugzilla : 6395, 9514
4538 Description: When migrating a subset of services from a node (e.g. failback
4539 from a failover service node) the remaining services would
4540 time out and evict clients.
4541 Details : lconf --force (implied by --failover) sets the global obd_timeout
4542 to 5 seconds in order to quickly disconnect, but this caused
4543 other RPCs to time out too quickly. Do not change the global
4544 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4546 Severity : enhancement
4547 Frequency : if MDS is started with down OST
4548 Bugzilla : 9439,5706
4549 Description: Allow startup/shutdown of an MDS without depending on the
4550 availability of the OSTs.
4551 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4552 Add appropriate locking and lov-osc refcounts for safe
4553 cleaning. Add osc abort_inflight calls in case the
4554 synchronize never started.
4557 Frequency : occasional (Cray XT3 only)
4559 Description: root not authorized to access files in CRAY_PORTALS environment
4560 Details : The client process capabilities were not honoured on the MDS in
4561 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4562 been accessed by an authorized user then root was able to access
4563 the file on the local client also. The root user capabilities
4564 are now allowed on the MDS, as this environment has secure UID.
4567 Frequency : occasional
4569 Description: ldiskfs "too long searching" message happens too often
4570 Details : A debugging message (otherwise harmless) prints too often on
4571 the OST console. This has been reduced to only happen when
4572 there are fragmentation problems on the filesystem.
4577 Description: Division by zero in statfs when all OSCs are inactive
4578 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4579 lov_statfs divides by value returned by lov_get_stripecnt().
4583 Bugzilla : 9489, 3273
4584 Description: First write from each client to each OST was only 4kB in size,
4585 to initialize client writeback cache, which caused sub-optimal
4586 RPCs and poor layout on disk for the first writen file.
4587 Details : Clients now request an initial cache grant at (re)connect time
4588 and so that they can start streaming writes to the cache right
4589 away and always do full-sized RPCs if there is enough data.
4590 If the OST is rebooted the client also re-establishes its grant
4591 so that client cached writes will be honoured under the grant.
4596 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4597 Details : Now I/O RPCs go to different portal number and (presumably) fast
4598 lock requests (and glimses) and other RPCs get their own service
4599 threads pool that should be able to service those RPCs
4602 Severity : enhancement
4604 Description: Ability to exchange lustre version between client and servers and
4605 issue warnings at client side if client is too old. Also for
4606 liblustre clients there is ability to refuse connection of too old
4608 Details : New 'version' field is added to connect data structure that is
4609 filled with version info. That info is later checked by server and
4613 Frequency : rare, liblustre only.
4614 Bugzilla : 9296, 9581
4615 Description: Two simultaneous writes from liblustre at offset within same page
4616 might proceed at the same time overwriting eachother with stale
4618 Details : I/O lock withing llu_file_prwv was released too early, before data
4619 actually was hitting the wire. Extended lock-holding time until
4620 server acknowledges receiving data.
4623 Frequency : extremely rare. Never observed in practice.
4625 Description: avoid generating lustre_handle cookie of 0.
4626 Details : class_handle_hash() generates handle cookies by incrementing
4627 global counter, and can hit 0 occasionaly (this is unlikely, but
4628 not impossible, because initial value of cookie counter is
4629 selected randonly). Value of 0 is used as a sentinel meaning
4630 "unassigned handle" --- avoid it. Also coalesce two critical
4631 sections in this function into one.
4633 Severity : enhancement
4635 Description: allow liblustre clients to delegate truncate locking to OST
4636 Details : To avoid overhead of locking, liblustre client instructs OST to
4637 take extent lock in ost_punch() on client's behalf. New connection
4638 flag is added to handle backward compatibility.
4640 Severity : enhancement
4641 Bugzilla : 4928, 7341, 9758
4642 Description: allow number of OST service threads to be specified
4643 Details : a module parameter allows the number of OST service threads
4644 to be specified via "options ost ost_num_threads={N}" in the
4645 OSS's /etc/modules.conf or /etc/modprobe.conf.
4649 Bugzilla : 6146, 9635, 9895
4650 Description: servers crash with bad pointer in target_handle_connect()
4651 Details : In rare cases when a client is reconnecting it was possible that
4652 the connection request was the last reference for that export.
4653 We would temporarily drop the export reference and get a new
4654 one, but this may have been the last reference and the export
4655 was just destroyed. Get new reference before dropping old one.
4657 Severity : enhancement
4658 Frequency : if client is started with failover MDS
4660 Description: Allow multiple MDS hostnames in the mount command
4661 Details : Try to read the configuration from all specified MDS
4662 hostnames during a client mount in case the "primary"
4665 Severity : enhancement
4667 Description: Stop sending data to evicted clients as soon as possible.
4668 Details : Check if the client we are about to send or are sending data to
4669 was evicted already. (Check is done every second of waiting,
4670 for which l_wait_event interface was extended to allow checking
4671 of exit condition at specified intervals).
4674 Frequency : rare, normally only when NFS exporting is done from client
4676 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4678 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4681 Frequency : always, if srand() is called before liblustre initialization
4683 Description: Liblustre uses system PRNG disturbing its usage by user application
4684 Details : Introduce internal to lustre fast and high-quality PRNG for
4685 lustre usage and make liblustre and some other places in generic
4686 lustre code to use it.
4688 Severity : enhancement
4689 Bugzilla : 9477, 9557, 9870
4690 Description: Verify that the MDS configuration logs are updated when xml is
4691 Details : Check if the .xml configuration logs are newer than the config
4692 logs stored on the MDS and report an error if this is the case.
4693 Request --write-conf, or allow starting with --old_conf.
4695 Severity : enhancement
4697 Description: Handle symlinks in the path when checking if Lustre is mounted.
4698 Details : Resolve intermediate symlinks when checking if a client has
4699 mounted a filesystem to avoid duplicate client mounts.
4704 Description: lconf can hit an error exception but still return success.
4705 Details : The lconf command catches the Command error exception at the top
4706 level script context and will exit with the associated exit
4707 status, but doesn't ensure that this exit status is non-zero.
4712 Description: failure of ptlrpc thread startup can cause oops
4713 Details : Starting a ptlrpc service thread can fail if there are a large
4714 number of threads or the server memory is very fragmented.
4715 Handle this without oopsing.
4718 Frequency : always, only if liblustre and non-default acceptor port was used
4720 Description: liblustre cannot connect to servers with non-default acceptor port
4721 Details : tcpnal_set_default_params() was not called and was therefore
4722 ignoring the environment varaible TCPNAL_PORT, as well as other
4723 TCPNAL_ environment variables
4728 Description: two objects could be created on the same OST for a single file
4729 Details : If an OST is down, in some cases it was possible to create two
4730 objects on a single OST for a single file. No problems other
4731 than potential performance impact and spurious error messages.
4735 Bugzilla : 5681, 9562
4736 Description: Client may oops in ll_unhash_aliases
4737 Details : Client dcache may become inconsistent in race condition.
4738 In some cases "getcwd" can fail if the current directory is
4744 Description: Inode refcounting problems in NFS export code
4745 Details : link_raw functions used to call d_instantiate without obtaining
4746 extra inode reference first.
4750 Bugzilla : 9942, 9903
4751 Description: Referencing freed requests leading to crash, memleaks with NFS.
4752 Details : We used to require that call to ll_revalidate_it was always
4753 followed by ll_lookup_it. Also with revalidate_special() it is
4754 possible to call ll_revalidate_it() twice for the same dentry
4755 even if first occurence returned success. This fix changes semantic
4756 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4757 reference on a request referred from the intent.
4758 ll_intent_release() then releases such a request.
4761 Frequency : rare, normally benchmark loads only
4763 Description: unlinked inodes were kept in memory on the client
4764 Details : If a client is repeatedly creating and unlinking files it
4765 can accumulate a lot of stale inodes in the inode slab cache.
4766 If there is no other client load running this can cause the
4767 client node to run out of memory. Instead flush old inodes
4768 from client cache that have the same inode number as a new inode.
4771 Frequency : SLES9 2.6.5 kernel and long filenames only
4772 Bugzilla : 9969, 10379
4773 Description: utime reports stale NFS file handle
4774 Details : SLES9 uses out-of-dentry names in some cases, which confused
4775 the lustre dentry revalidation. Change it to always use the
4779 Frequency : rare, unless heavy write-truncate concurrency is continuous
4780 Bugzilla : 4180, 6984, 7171, 9963, 9331
4781 Description: OST becomes very slow and/or deadlocked during object unlink
4782 Details : filter_destroy() was holding onto the parent directory lock
4783 while truncating+unlinking objects. For very large objects this
4784 may block other threads for a long time and slow overall OST
4785 responsiveness. It may also be possible to get a lock ordering
4786 deadlock in this case, or run out of journal credits because of
4787 the combined truncate+unlink. Solution is to do object truncate
4788 first in one transaction without parent lock, and then do the
4789 final unlink in a new transaction with the parent lock. This
4790 reduces the lock hold time dramatically.
4793 Frequency : rare, 2.4 kernels only
4795 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4796 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4797 before unmounting their filesystem in order to avoid deadlock.
4798 The kernel_locked() function in 2.4 kernels only checks whether
4799 the kernel lock is held, not whether it is this process that is
4800 holding it as 2.6 kernels do.
4805 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4806 Details : The client ptlrpc code may be trying to reconnect to a down
4807 server before a previous connection attempt has timed out.
4808 Increase the reconnect interval to be longer than the connection
4809 timeout interval to avoid sending duplicate connections to
4813 Frequency : echo_client brw_test command
4815 Description: fix echo_client to work with OST preallocated code
4816 Details : OST preallocation code (5137) didn't take echo_client IO path
4817 into account: echo_client calls filter methods outside of any
4818 OST thread and, hence, there is no per-thread preallocated
4819 pages and buffers to use. Solution: hijack pga pages for IO. As
4820 a byproduct, this avoids unnecessary data copying.
4824 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4825 Description: Client can oops in mdc_commit_close() after open replay
4826 Details : It was possible for the MDS to return an open request with no
4827 transaction number in mds_finish_transno() if the client was
4828 evicted, but without actually returning an error. Clients
4829 would later try to replay that open and may trip an assertion
4830 Simplify the client close codepath, and always return an error
4831 from the MDS in case the open is not successful.
4834 Frequency : rare, 2.6 OSTs only
4836 Description: OST may deadlock under high load on fragmented files
4837 Details : If there was a heavy load and highly-fragmented OST filesystems
4838 it was possible to have all the OST threads deadlock waiting on
4839 allocation of biovecs, because the biovecs were not released
4840 until the entire RPC IO was completed. Instead, release biovecs
4841 as soon as they are complete to ensure forward IO progress.
4843 Severity : enhancement
4845 Description: Support for specifying external journal device at mount
4846 Details : If an OST or MDS device is formatted with an external journal
4847 device, this device major/minor is stored in the ext3 superblock
4848 and may not be valid for failover. Allow detecting and
4849 specifying the external journal at mount time.
4854 Description: Mounting an MDS with pending unlinked files may cause oops
4855 Details : target_finish_recovery() calls mds_postrecov() which returned
4856 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4857 considers this an error and immediately begins cleaning up the
4858 lov, just after starting the mds_lov process
4860 Severity : enhancement
4862 Description: Implement 'lfs df' to report actual free space on per-OST basis
4863 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4864 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4865 usage of objects; '-h' to report in human readable format.
4867 ------------------------------------------------------------------------------
4869 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4876 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4877 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4878 at startup during orphan unlinking. Instead, skip buddy update
4879 before setup, as it will be regenerated after recovery is complete.
4884 Description: If an OST is inactive, its locks might reference stale inodes.
4885 Details : lov_change_cbdata() must iterate over all namespaces, even if
4886 they are inactive to clear inode references from the lock.
4888 Severity : enhancement
4889 Frequency : occasional, if non-standard max_dirty_mb used
4891 Description: Client will block write RPCs if not enough grant
4892 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4893 then the client will block writes while waiting for another RPC
4894 to complete instead of consuming its dirty limit. With change
4895 we get improved performance when max_dirty_mb is small.
4897 Severity : enhancement
4898 Bugzilla : 3389, 6253
4899 Description: Add support for supplementary groups on the MDS.
4900 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4901 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4902 lookups for user supplementary groups into a cache.
4906 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4907 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4908 for liblustre O_EXCL is enforced by MDS.
4912 Description: readdir never returns NULL in liblustre.
4913 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4914 dentry in struct dirent.
4918 Frequency : liblustre only, depends on application IO pattern
4919 Description: liblustre clients evicted if not contacting servers
4920 Details : Don't put liblustre clients into the ping_evictor list, so
4921 they will not be evicted by the pinger ever.
4923 Severity : enhancement
4925 Description: Add ability to evict clients by NID from MDS.
4926 Details : By echoing "nid:$NID" string into
4927 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4928 $NID would be instantly evicted from this MDS and from all active
4929 OSTs connected to it.
4933 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4934 Details : lookup_it_finish() used to query file size from OSTs that was not
4939 Description: service threads change working directory to that of init
4940 Details : Starting lustre service threads may pin the working directory
4941 of the parent thread, making that filesystem busy. Threads
4942 now change to the working directory of init to avoid this.
4946 Frequency : during shutdown only
4947 Description: shutdown with a failed MDS or OST can cause unmount to hang
4948 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4953 Frequency : 2.6 only
4954 Description: chmod/chown may include an extra supplementary group
4955 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4956 supplementary group and if none is specified this is used.
4959 Bugzilla : 5479 (6816)
4961 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4962 Details : If lookup is in progress on a file that is unlinked we might try
4963 to revalidate the inode and fail in revalidate after lookup is
4964 complete and ll_file_open() enqueues the open again but
4965 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4968 Frequency : always, if lconf --abort_recovery used
4970 Description: lconf --abort_recovery fails with 'Operation not supported'
4971 Details : lconf was attempting to abort recovery on the MDT device and not
4974 ------------------------------------------------------------------------------
4976 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4981 Frequency : rare (only unsupported configurations with a node running as an
4983 Bugzilla : 6514, 5137
4984 Description: Mounting a Lustre file system on a node running as an OST could
4986 Details : OSTs now preallocates memory needed to write out data at
4987 startup, instead of when needed, to avoid having to
4988 allocate memory in possibly low memory situations.
4989 Specifically, if the file system is mounted on on OST,
4990 memory pressure could force it to try to write out data,
4991 which it needed to allocate memory to do. Due to the low
4992 memory, it would be unable to do so and the node would
4993 become unresponsive.
4995 Severity : enhancement
4997 Description: Addition of lconf --service command line option
4998 Details : lconf now accepts a '--service <arg>' option, which is
4999 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5001 Severity : enhancement
5003 Description: Failover mode is now the default for OSTs.
5004 Details : By default, OSTs will now run in failover mode. To return to
5005 the old behaviour, add '--failout' to the lmc line for OSTs.
5007 Severity : enhancement
5009 Description: Health checks are now provided for MDS and OSTs
5010 Details : Additional detailed health check information on MSD and OSTs
5011 is now provided through the procfs health_check value.
5014 Frequency : occasional, depends on IO load
5016 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5017 after numerous create/delete cycles
5018 Details : The ext3 inode allocation policy would not allocate new inodes
5019 very well on the OSTs because there are no new directories
5020 being created. Instead we look for groups with free space if
5021 the parent directories are nearly full.
5026 Description: Network or server problems during mount may cause partially
5027 mounted clients instead of returning an error.
5028 Details : The config llog parsing code may overwrite the error return
5029 code during mount error handling, returning success instead
5035 Description: MDS can fail to allocate large reply buffers
5036 Details : After long uptimes the MDS can fail to allocate large reply
5037 buffers (e.g. zconf client mount config records) due to memory
5038 fragmentation or consumption by the buffer cache. Preallocate
5039 some large reply buffers so that these replies can be sent even
5040 under memory pressure.
5044 Frequency : rare (liblustre)
5045 Description: fsx running with liblustre complained that using truncate() to
5046 extend the file doesn't work. This patch corrects that issue.
5047 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5048 ATTR_SIZE and lsm use in llu_setattr_raw.
5052 Frequency : rare, only 2.6 kernels
5053 Description: Unusual file access patterns on the MDS may result in inode
5054 data being lost in very rare circumstances.
5055 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5056 code in the 2.6 kernel caused the inode and/or EA data not to
5057 be read from disk, causing single-file corruption.
5061 Frequency : rare, only 2.6 filesystems using extents
5062 Description: Heavy concurrent write and delete load may cause data corruption.
5063 Details : It was possible under high-load situations to have an extent
5064 metadata block in the block device cache from a just-unlinked
5065 file overwrite a newly-allocated data block. We now unmap any
5066 metadata buffers that alias just-allocated data blocks.
5070 Frequency : filesystems with default stripe_count larger than 77
5071 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5072 Details : lconf specifies an inode size of 4096 bytes when the default
5073 stripe_count is larger than 77. This conflicts with the default
5074 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5075 this case to avoid pinning too much memory for large EAs.
5077 ------------------------------------------------------------------------------
5079 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5084 Frequency : rare (extremely heavy IO load with hundreds of clients)
5086 Description: Client is evicted, gets IO error writing to file
5087 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5088 caused clients to be evicted for AST timeouts. The fixes in
5089 bug 5192 mean we no longer need to have such short AST timeouts
5090 so ldlm_timeout has been increased.
5093 Frequency : occasional during --force or --failover shutdown under load
5094 Bugzilla : 5949, 4834
5095 Description: Server oops/LBUG if stopped with --force or --failover under load
5096 Details : a collection of import/export refcount and cleanup ordering
5097 issues fixed for safer force cleanup
5100 Frequency : only filesystems larger than 120 OSTs
5101 Bugzilla : 5990, 6223
5102 Description: lfs getstripe would oops on a very large filesystem
5103 Details : lov_getconfig used kfree on vmalloc'd memory
5106 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5107 Bugzilla : 6242, 6243
5108 Description: reading from files that had been truncated to a non-zero size
5109 but never opened returned no data
5110 Details : ll_file_read() reads zeros from no-object files to EOF
5115 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5116 Details : There's little harm in aborting MDS/OSS recovery and letting it
5117 try again, so I removed the LASSERT and return an error instead.
5119 Severity : enhancement
5121 Description: New debugging infrastructure for tracking down data corruption
5122 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5123 (b) cover more of the client-side code path, and (c) try to narrow
5124 down where problems occurred
5128 Bugzilla : 3819, 4364, 4397, 6313
5129 Description: Racing close and eviction MDS could cause assertion in mds_close
5130 Details : It was possible to get multiple mfd references during close and
5131 client eviction, leading to one thread referencing a freed mfd.
5133 Severity: : enhancement
5134 Bugzilla : 3262, 6359
5135 Description: Attempts to reconnect to servers are now more aggressive.
5136 Details : This builds on the enhanced upcall-less recovery that was added
5137 in 1.4.2. When trying to reconnect to servers, clients will
5138 now try each server in the failover group every 10 seconds. By
5139 default, clients would previously try one server every 25 seconds.
5144 Description: After recovery, certain operations trigger a failed
5145 assertion on a client.
5146 Details : Failing over an mds, using lconf -d --failover, while a
5147 client was doing a readdir() call would cause the client to
5148 LBUG after recovery completed and the readdir() was resent.
5150 Severity : enhancement
5152 Description: Default groups are now added by lconf
5153 Details : You can now run lconf --group <servicename> without having to
5154 manually add groups with lmc.
5157 Frequency : occasional
5159 Description: Nodes with an elan id of 0 trigger a failed assertion
5162 Frequency : always when accessing e.g. tty/console device nodes
5164 Description: tty and some other devices nodes cannot be used on lustre
5165 Details : file's private_data field is used by device data and lustre
5166 values in there got lost. New field was added to struct file to
5167 store fs-specific private data.
5170 Frequency : when exporting Lustre via NFS
5172 Description: NFSD failed occasionally when looking up a path component
5173 Details : NFSD is looking up ".." which was broken in ext3 directories
5174 that had grown large enough to become hashed.
5177 Frequency : Clusters with multiple interfaces not on the same subnet
5179 Description: Nodes will repeatedly try to reconnect to an interface which it
5180 cannot reach and report an error to the log.
5181 Details : Extra peer list entries will be created by lconf with some peers
5182 unreachable. lconf now validates the peer before adding it.
5185 Frequency : Only if a default stripe is set on the filesystem root.
5187 Description: Setting a default stripe on the filesystem root prevented the
5188 filesystem from being remounted.
5189 Details : The client was sending extra request flags in the root getattr
5190 request and did not allocate a reply buffer for the dir EA.
5193 Frequency : occasional, higher if lots of files are accessed by one client
5194 Bugzilla : 6159, 6097
5195 Description: Client trips assertion regarding lsm mismatch/magic
5196 Details : While revalidating inodes the VFS looks up inodes with ifind()
5197 and in rare cases can find an inode that is being freed.
5198 The ll_test_inode() code will free the lsm during ifind()
5199 when it finds an existing inode and then the VFS later attaches
5200 this free lsm to a new inode.
5204 Bugzilla : 6422, 7030
5205 Description: MDS deadlock between mkdir and client eviction
5206 Details : Creating a new file via mkdir or mknod (starting a transaction
5207 and getting the ns lock) can deadlock with client eviction
5208 (gets ns lock and trying to finish a synchronous transaction).
5211 Frequency : occasional
5212 Description: While starting a server, the fsfilt_ext3 module could not be
5214 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5215 kernels. Previously, lconf would still use the ext3 name
5216 when trying to load modules. Now, it will correctly use
5217 ext3 on 2.4 and ldiskfs on 2.6.
5219 Severity : enhancement
5220 Description: The default stripe count has been changed to 1
5221 Details : The interpretation of the default stripe count (0, to lfs
5222 or lmc) has been changed to mean striping across a single
5223 OST, rather than all available. For general usage we have
5224 found a stripe count of 1 or 2 works best.
5226 Severity : enhancement
5227 Description: Add support for compiling against Cray portals.
5228 Details : Conditional compiling for some areas that are different
5232 Frequency : occasional
5233 Bugzilla : 6409, 6834
5234 Description: Creating files with an explicit stripe count may lead to
5235 a failed assertion on the MDS
5236 Details : If some OSTs are full or unavailable, creating files may
5237 trigger a failed assertion on the MDS. Now, Lustre will
5238 try to use other servers or return an error to the
5242 Frequency : occasional
5244 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5245 caused lock timeout during readahead (since 1.4.2).
5246 Details : Processes doing readahead might match a lock that hasn't been
5247 granted yet if there are overlapping and conflicting lock
5248 requests. The readahead process waits on ungranted lock
5249 (original lock is CBPENDING), while OST waits for that process
5250 to cancel CBPENDING read lock and eventually evicts client.
5252 Severity : enhancement
5254 Description: Initial enabling of flock support for clients
5255 Details : Implements fcntl advisory locking and file status functions.
5256 This feature is provided as an optional mount flag (default
5257 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5258 locking are implemented yet, and those that are are not guaranteed
5259 to be completely correct in production environments.
5260 mount -t lustre -o [flock|noflock] ...
5263 Frequency : occasional
5265 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5266 assertion in the ext3 JBD (journaling) layer.
5267 Details : The b_committed_data struct is protected by the big kernel lock
5268 in 2.4 kernels, serializing journal_commit_transaction() and
5269 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5270 access to this struct. In 2.6 kernels there is finer grained
5271 locking to improve SMP performance of the JBD layer.
5275 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5280 Description: merge and cleanup kernel patches.
5283 -----------------------------------------------------------------------------
5285 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5287 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5288 versions of Lustre. Please update all servers and clients to
5289 version 1.4.2 or later at the same time. You must also run
5290 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5291 to update the configuration logs.
5293 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5294 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5295 - swab llog records of type '0' so we get proper header size/idx (5861)
5296 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5297 - fix rename of one directory over another leaking an inode (5953)
5298 - avoid SetPageDirty on 2.6 (5981)
5299 - don't re-add just-being-destroyed locks to the waiting list (5653)
5300 - when creating new directories, inherit the parent's custom
5301 striping settings if present parent (3048)
5302 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5303 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5304 - don't hold client locks on temporary worklist from l_lru (5666)
5305 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5306 - automatically evict dead clients (5921)
5307 - Update file size properly in create+truncate+fstat case (6196)
5308 - Do not unhash mountpoint dentries, do not allow removal of
5310 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5311 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5312 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5313 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5314 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5315 - Correctly handle reads of files with no objects (6243)
5316 - lctl recover will also mark a device active if deactivate used (5933)
5318 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5319 - allow --write-conf on an MDS with different nettype than client (5619)
5320 - don't write config llogs to MDS for mounts not from that MDS (5617)
5321 - lconf should create multiple TCP connections from a client (5201)
5322 - init scripts are now turned off by default; run chkconfig --on
5323 lustre and chkconfig --on lustrefs to use them
5324 - upcalls are no longer needed for clients to recover to failover
5326 - add --abort-recovery option to lconf to abort recovery on device
5328 - add support for an arbitrary number of OSTs (3026)
5329 - Quota support protocol changes.
5330 - forward compatibility changes to wire structs (6007)
5331 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5332 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5333 - improved "lustre status" script
5334 - initialize blocksize for non-regular files (6062)
5335 - added --disable-server and --disable-client configure options (5782)
5336 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5337 - Vanilla 2.4.29 support
5338 - increase maximum number of obd devices to 520 (6242)
5339 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5340 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5341 SLES 9 2.6.5 kernels
5342 - limit stripes per file to 160 (the maximum EA size) (6093)
5344 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5347 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5348 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5349 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5350 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5351 - small ext3 extents cleanups and fixes (5733)
5352 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5353 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5354 - avoid needless client->OST connect, fix handle mismatch (5317)
5355 - fix DLM error path that led to out-of-sync client, long delays (5779)
5356 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5357 - fix several locking issues related to i_size (5492,5624,5654,5672)
5358 - don't move pending lock onto export if it is already evicted (5683)
5359 - fix kernel oops when creating .foo in unlinked directory (5548)
5360 - fix deadlock in obdfilter statistics vs. object create (5811)
5361 - use time_{before,after} to avoid timer jiffies wrap (5882)
5362 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5363 - Do not leak request if server was not able to process it (5154)
5364 - If mds_open unable to find parent dir, make that negative lookup(5154)
5365 - don't create new directories with extent-mapping (5909, 5936)
5367 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5368 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5369 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5370 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5371 - support for RHEL4 kernel on i686 (5773)
5372 - provide error messages when incompatible logs are encountered (5898)
5374 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5375 * version 1.4.0.10 (1.4.1 release candidate 1)
5377 - don't keep a lock reference when lock is not granted (4238)
5378 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5379 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5380 - drop import inflight refcount on signal_completed_replay error (5255)
5381 - unlock page after async write error during send (3677)
5382 - handle missing objects in filter_preprw_read properly (5265)
5383 - no transno return for symlink open, don't save no-trasno open (3440)
5384 - don't try to complete elan receive that already failed (4012)
5385 - free RPC server reply state on error (5406)
5386 - clean up thread from ptlrpc_start_thread() on error (5160)
5387 - readahead could read extra page into cache that wasn't ejected (5388)
5388 - prevent races in class_attach/setup/cleanup/detach (5260)
5389 - don't dereference de->d_inode after l_dput of de (5458)
5390 - use "int" for stripe value returned from lock_to_stripe (5544)
5391 - mballoc allocation and error-checking fixes in 2.6 (5504)
5392 - block device patches to fix I/O request sizes in 2.6 (5482)
5393 - look up hostnames for IB nals (5602)
5394 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5395 - don't start multiple acceptors for the same port (5277)
5396 - fix incorrect LASSERT in mds_getattr_name (5635)
5397 - export a proc file for general "ping" checking (5628)
5398 - fix "lfs check" to not block when the MDS is down (5628)
5400 - service request history (4965)
5401 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5402 - create an "evict_client" /proc entry on OSTs, like the MDS has
5403 - fix mount usage message, return errors per mount(8) (5168)
5404 - change grep [] to grep "[]" in tests so they work in more UMLs
5405 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5406 - remove ancient LOV_MAGIC_V0 EA support (5047)
5407 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5408 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5409 - allow readahead allocations to fail when low on memory (5383)
5410 - mmap locking landed again, after considerable improvement (2828)
5411 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5413 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5416 - send OST transaction number in read/write reply to free req (4966)
5417 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5418 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5419 - the watchdog thread now runs as interruptible (5246)
5420 - flock/lockf fixes (but it's still disabled, pending 5135)
5421 - don't use EXT3 constants in llite code (5094)
5422 - memory shortage at startup could cause assertion (5176)
5424 - reorganization of lov code
5425 - single portals codebase
5427 - add extents/mballoc support (5025)
5428 - direct I/O reads in the obdfilter (4048)
5429 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5431 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5434 - send OST transaction number in read/write reply to free req (4966)
5435 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5436 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5437 - the watchdog thread now runs as interruptible (5246)
5438 - handle missing objects in filter_preprw_read properly (5265)
5439 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5440 - drop import inflight refcount on signal_completed_replay error (5255)
5441 - unlock page after async write error during send (3677)
5442 - return original error code on reconstructed replies (3761)
5443 - no transno return for symlink open, don't save no-trasno open (3440)
5445 - add pid to ldlm debugging output (4922)
5446 - bump the watchdog timeouts -- we can't handle 30sec yet
5447 - extra debugging for orphan dentry/inode bug (5259)
5449 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5452 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5453 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5454 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5455 - fix NULL dereference in /proc/sys/portals/routes (4827)
5456 - allow failed mdc_close() operations to be interrupted (4561)
5457 - stop precreate on OST before MDS would time out on it (4778)
5458 - don't send partial-page writes before EOF from client (4410)
5459 - discard client grant for sub-page writes on large-page clients (4520)
5460 - don't free dentries not owned by NFS code, check generation (4806)
5461 - fix lsm leak if mds_create_objects() fails (4801)
5462 - limit debug_daemon file size, always print CERROR messages (4789)
5463 - use transno after validating reply (3892)
5464 - process timed out requests if import state changes (3754)
5465 - update mtime on OST during writes, return in glimpse (4829)
5466 - add mkfsoptions to LDAP (4679)
5467 - use ->max_readahead method instead of zapping global ra (5039)
5468 - don't interrupt __l_wait_event() during strace
5470 - add software watchdogs to catch hung threads quickly (4941)
5471 - make lustrefs init script start after nfs is mounted
5472 - fix CWARN/ERROR duplication (4930)
5473 - return async write errors to application if possible (2248)
5474 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5475 - print NAL number in %x format (4645)
5476 - update barely-supported suse-2.4.21-171 series (4842)
5477 - support for sles 9 %post scripts
5478 - support for building 2.6 kernel-source packages
5479 - support for sles km_* packages
5481 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5484 - ignore -ENOENT errors in osc_destroy (3639)
5485 - notify osc create thread that OSC is being cleaned up (4600)
5486 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5487 - reconstruct ost_handle() like mds_handle() (4657)
5488 - create a new thread to do import eviction to avoid deadlock (3969)
5489 - let lconf resolve symlinked-to devices (4629)
5490 - don't unlink "objects" from directory with default EA (4554)
5491 - hold socknal file ref over connect in case target is down (4394)
5492 - allow more than 32000 subdirectories in a single directory (3244)
5493 - fix blocks count for O_DIRECT writes (3751)
5494 - OST returns ENOSPC from object create when no space left (4539)
5495 - don't send truncate RPC if file size isn't changing (4410)
5496 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5497 - bind to privileged port in socknal and tcpnal (3689)
5499 - rate limit CERROR/CWARN console message to avoid overload (4519)
5500 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5501 - basic mmap support (3918)
5502 - kernel patch series update from b1_4 (4711)
5504 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5507 - avoid crash during MDS cleanup with OST shut down (2775)
5508 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5509 - don't use bad inodes on the MDS (3744)
5510 - dynamic object preallocation to improve recovery speed (4236)
5511 - don't hold spinlock over lock dumping or change debug flags (4401)
5512 - don't zero obd_dev when it is force cleaned (3651)
5513 - print grants to console if they go negative (4431)
5514 - "lctl deactivate" will stop automatic recovery attempts (3406)
5515 - look for existing locks in ldlm_handle_enqueue() (3764)
5516 - don't resolve lock handle twice in recovery avoiding race (4401)
5517 - revalidate should check working dir is a directory (4134)
5519 - don't always mark "slow" obdfilter messages as errors (4418)
5521 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5524 - don't close LustreDB during write_conf until it is done (3860)
5525 - fix typo in lconf for_each_profile (3821)
5526 - allow dumping logs from multiple threads at one time (3820)
5527 - don't allow multiple threads in OSC recovery (3812)
5528 - fix debug_size parameters (3864)
5529 - fix mds_postrecov to initialize import for llog ctxt (3121)
5530 - replace config semaphore with spinlock (3306)
5531 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5532 - don't allow enqueue to complete on a destroyed export (3822)
5533 - down write_lock before checking llog header bitmap (3825)
5534 - recover from lock replay timeout (3764)
5535 - up llog sem before sending rpc (3652)
5536 - reduce ns lock hold times when setting kms (3267)
5537 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5538 - fix NULL deref and obd_dev leak on setup error (3312)
5539 - replace some LBUG about llog ops with error handling (3841)
5540 - don't match INVALID dentries from d_lookup and spin (3784)
5541 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5542 - fix invalid assertion in ptlrpc_set_wait (3880)
5544 - add libwrap support for the TCP acceptor (3996)
5545 - add /proc/sys/portals/routes for non-root route listing (3994)
5546 - allow setting MDS UUID in .xml (2580)
5547 - print the stack of a process that LBUGs (4228)
5549 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5552 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5553 - make sure to unset replay flag from failed open requests (3440)
5554 - if default stripe count is 0, use OST count for inode size (3636)
5555 - update parent mtime/ctime on client for create/unlink (2611)
5556 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5557 - free recovery state on server during a forced cleanup (3571)
5558 - unregister_reply for resent reqs (3063)
5559 - loop back devices mounting and status check on 2.6 (3563)
5560 - fix resource-creation race that can provoke i_size == 0 (3513)
5561 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5562 - more debugging for page-accounting assertion (3746)
5563 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5564 - avoid deadlock after precreation failure (3758)
5565 - fix race and lock order deadlock in orphan handling (3450, 3750)
5566 - add validity checks when grabbing inodes from l_ast_data (3599)
5568 - add /proc/.../recovery_status to obdfilter (3428)
5569 - lightweight CDEBUG infrastructure, debug daemon (3668)
5570 - change default OSC RPC parameters to be better on small clusters
5571 - turn off OST read cache for files smaller than 32MB
5572 - install man pages and include them in rpms (3100)
5573 - add new init script for (un)mounting lustre filesystems (2593)
5574 - run chkconfig in %post for init scripts (3701)
5575 - drop scimac NAL (unmaintained)
5577 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5580 - clean kiobufs before and after use (3485)
5581 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5582 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5583 - change init script to fail more clearly if not run as root (1528)
5584 - allow clients to reconnect during replay (1742)
5585 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5586 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5587 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5588 - fix cond_resched (3554)
5589 - only evict unfinished clients after recovery (3515)
5590 - allow bulk resend, prevent data loss (3570)
5591 - dynamic ptlrpc request buffer allocation (2102)
5592 - don't allow unlinking open directory if it isn't empty (2904)
5593 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5594 - remove extraneous obd dereference causing LASSERT failure (3334)
5595 - don't use get_cycles() when creating temp. files on the mds (3156)
5596 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5597 - handle EEXIST for set-stripe, set proper directory name (3336)
5599 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5600 - fix ksocknal_fmb_callback() error messages (2918)
5602 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5605 - don't copy lvb into (possibly NULL) reply on error (2983)
5606 - don't deref dentry after dput, don't free lvb on error (2922)
5607 - use the kms to determine writeback rpc length (2947)
5608 - increment oti_logcookies when osc is inactive (2948)
5609 - update client's i_blocks count via lvb messages (2543)
5610 - handle intent open/close of special files properly (1557)
5611 - mount MDS with errors=remount-ro, like obdfilter (2009)
5612 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5613 - don't use cancelling-locks' kms values (2947)
5614 - use highest lock extent for kms, not last one (2925)
5615 - don't dereference ERR_PTR() dentry in error handling path (3107)
5616 - fix thread race in portals_debug_dumplog() (3122)
5617 - create lprocfs device entries at setup instead of at attach (1519)
5618 - common AST error handler, don't evict client on completion race (3145)
5619 - zero nameidata in detach_mnt in 2.6 (3118)
5620 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5621 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5622 - initialize RPC timeout timer earlier for 2.6 (3219)
5623 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5624 - print nal/nid for unknown nid (3258)
5625 - additional checks for oscc recovery before doing precreate (3284)
5626 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5627 - don't crash in mdc_close for bad permissions on open (3285)
5628 - zero i_rdev for non-device files (3147)
5629 - clear page->private before handing to FS, better assertion (3119)
5630 - tune the read pipeline (3236)
5631 - fix incorrect decref of invalidated dentry (2350)
5632 - provide read-ahead stats and refine rpc in flight stats (3328)
5633 - don't hold journal transaction open across create RPC (3313)
5634 - update atime on MDS at close time (3265)
5635 - close LDAP connection when recovering to avoid server load (3315)
5636 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5637 - don't leak open file on MDS after open resend (3325)
5638 - serialize filter_precreate and filter_destroy_precreated (3329)
5639 - loop device shouldn't call sync_dev() for nul device (3092)
5640 - clear page cache after eviction (2766)
5641 - resynchronize MDS->OST in background (2824)
5642 - refuse to mount the same filesystem twice on same mountpoint (3394)
5643 - allow llmount to create routes for mounting behind routers (3320)
5644 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5645 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5646 - fix rare problem with rename on htree directories (3417)
5648 - allow default OST striping configuration per directory (1414)
5649 - fix compilation for qswnal for 2.6 kernels (3125)
5650 - increase maximum number of MDS request buffers for large systems
5651 - change liblustreapi to be useful for external progs like lfsck (3098)
5652 - increase local configuration timeout for slow disks (3353)
5653 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5655 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5658 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5659 - don't overwrite extent policy data in reply if lock was blocked (2901)
5660 - drop filter export grants atomically with removal from device (2663)
5661 - del obd_self_export from work_list in class_disconnect_exports (2908)
5662 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5663 - swab reply message in mdc_close, other PPC fixes (2464)
5664 - fix destroying of named logs (2325)
5665 - overwrite old logs when running lconf --write_conf (2264)
5666 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5667 - fix race in target_handle_connect (2898)
5668 - mds_reint_create() should take same inode create lock (2926)
5669 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5670 - don't close files for self_export to avoid uninitialized obd (2936)
5671 - allow MDS with the same name as client node (2939)
5672 - hold dentry reference for closed log files for unlink (2325)
5673 - reserve space for all logs during transactions (2059)
5674 - don't evict page beyond end of stripe extent (2925)
5675 - don't oops on a deleted current working directory (2399)
5676 - handle hard links to targets without a parent properly (2517)
5677 - don't dereference NULL lock when racing during eviction (2867)
5678 - don't grow lock extents when lots of conflicting locks (2919)
5680 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5683 - account for cache space usage on clients to avoid data loss (974)
5684 - lfsck support in lustre kernel code (2349)
5685 - reduce journal credits needed for BRW writes (2370)
5686 - orphan handling to avoid losing space on client/server crashes
5687 - ptlrpcd can be blocked, stopping ALL progress (2477)
5688 - use lock value blocks to assist in proper KMS, faster stat (1021)
5689 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5690 - recovery for initial connections (2355)
5691 - fixes for mds_cleanup_orphans (1934)
5692 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5693 - block all file creations until orphan recovery completes (1901)
5694 - client remove rq_connection from request struct (2423)
5695 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5696 - recovery timer race (2670)
5697 - mdc_close recovey bug (2532)
5698 - ptlrpc cleanup bug (2710)
5699 - mds timeout on local locks (2588)
5700 - namespace lock held during RPCs (2431)
5701 - handle interrupted sync write properly (2503)
5702 - don't try to handle a message that hasn't been replied to (2699)
5703 - client assert failure during cleanup after abort recovery (2701)
5704 - leak mdc device after failed mount (2712)
5705 - ptlrpc_check_set allows timedout requests to complete (2714)
5706 - wait for inflight reqs when ptlrpcd finishes (2710)
5707 - make sure unregistered services are removed from the srv_list
5708 - reset bulk XID's when resending them (caught by 1138 test)
5709 - unregister_bulk after timeout
5710 - fix lconf error (2694)
5711 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5712 - readahead locks pages, leaves pending causing memory pressure (2673)
5713 - increase OST request buffers to 4096 on large machines (2729)
5714 - fix up permission of existing directories in simple_mkdir (2661)
5715 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5716 - don't assign transno to errored transactions (2742)
5717 - don't delete objects on OST if given a bogus objid from MDS (2751)
5718 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5719 - if rq_no_resend, then timeout request after recovery (2432)
5720 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5721 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5722 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5723 - MDS recovery shouldn't delete objects that it has given out (2730)
5724 - if enqueue arrives after completion, don't clobber LVB (2819)
5725 - don't unlock pages twice when trigger_group_io returns error (2814)
5726 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5727 - don't write pages to disk if there was an error (1450)
5728 - don't ping imports that have recovery disabled (2676)
5729 - take buffered bytes into account when balancing socknal conn (2817)
5730 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5731 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5732 - remove little-endian swabbing of llog records (1987)
5733 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5734 - retry reposting request buffers if they fail (1191)
5735 - grow extent at grant time to avoid granting a revoked lock (2809)
5736 - lock revoke doesn't evict page if covered by a second lock (2765)
5737 - disable VM readahead to avoid reading outside lock extents (2805)
5739 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5740 - updated kernel patches for hp-2.4.20 kernel (2681)
5742 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5745 - fix truncated write corruption (2366)
5746 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5748 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5749 - ensure len(uuid) < 37 in lmc (1171)
5750 - fix ia64 OOPS in llog_test (2255)
5751 - zero end of page at obdfilter for partial page writes (2648)
5752 - don't leave stale dentries around after renames (bug 2428)
5753 - fix timeouts when evicting a client with a single lock held (2642)
5754 - set deadline for the initial HELLO message to drain (2634)
5755 - print out dotted-quad IP addresses in the socknal (2302)
5757 - additional debugging for MDS client eviction problem (2443)
5758 - fix mkfsoptions support for osts (2603, 2604)
5760 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5763 - add series for the vanilla 2.6.0 kernel
5764 - add series for the vanilla 2.4.24 kernel
5765 - add series for a cray x86/64 UL kernel drop
5766 - fix xattr patches for the vanilla 2.4.19 series
5768 - generate true UUIDs in lmc (1171)
5769 - have portals stack dumping break in UML (2466)
5770 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5771 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5772 - wake callback waiting threads on client eviction (2460)
5773 - Add --ptldebug and --subsystem to lmc (1719)
5774 - update assertion to allow safe interrupt allocation
5775 - set rq_no_resend for cancel requests (2432)
5776 - recalculate ptlrpcd timeout after resend (2494)
5777 - call vfs_rmdir when removing pending directories (2368)
5778 - fix renaming a file to itself (2429)
5779 - lmc creates a default one-stripe lov (2454)
5780 - expand procfs space to handle large clusters (2326)
5781 - increase UML stack to avoid overflow
5782 - update lconf's list of debug and subsystem masks
5783 - fix lfs find --obd (2510)
5784 - /proc tunable for disabling filter read caching (2591)
5785 - stop rpm packages from altering slapd.conf (2301)
5786 - disable nagle in the socknal under 0conf (2578)
5787 - choose mds inode size based on stripe count (2572)
5788 - fix kernel-source rpm problems (2516)
5790 - add --disable-doc to avoid pdf generation (2421)
5791 - update documentation, tests, type-os, comments
5792 - avoid format warnings on ia64
5793 - remove the TOE NAL
5794 - tiny code cleanups by removing unused fields
5796 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5799 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5800 - remove the most blatant lies from BUILDING file (2371)
5801 - change default debug level to reasonable production setting
5802 - reduce client side cache size to reduce cache flush time
5803 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5804 - make TCP zerocopy and pinger support enabled by default (2476)
5805 - sync writes completed after process exits caused crashes (2319)
5806 - maintain correct mount count on the MDS (2356)
5807 - backout 1557, because 2316 wasn't really fixed
5808 - better file I/O statistics gathering in /proc
5809 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5810 - another kernel patch to fix zero-copy TCP function export
5811 - don't take duplicate lock when processing re-sent getattr (2420)
5812 - lctl uses obd_self_export instead of creating new conn (2353)
5813 - MDS/OST recovery case which requires object creation asserted (2425)
5814 - move lfs from /usr/sbin to /usr/bin in packages
5815 - fix race between mds_client_add and mds_client_free (2417)
5816 - use kmalloc instead of slabs in portals (2430)
5817 - don't create duplicate records when a failover MDS is present (2442)
5818 - remove unnecessary mount age check (2332)
5819 - don't remove directory inodes from locks prematurely (2451)
5820 - don't break if MDS service name is the same as hostname (2103)
5821 - fix races in client write RPC generation when cache full (2482)
5823 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5826 - remove now-unused request->rq_obd (278)
5827 - if an allocation fails, print out how much memory we've used (1933)
5828 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5829 - add the "configurable stack size" patch to most series files (1256)
5830 - ability to write large log records, for 100+ OST configs (2306)
5831 - fix NULL deref when filter_prep fails (2314)
5832 - fix operator precedence error in filter_sync
5833 - dynamic allocation of socknal TX descriptors (2315)
5834 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5835 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5836 - ensure that CWARN messages in Portals always get to the syslog
5837 - __init/__exit are not for prototype decls (ldlm_init/exit)
5838 - x86-64 compile warning fixes
5839 - fix gateway LMC keyword conflict (2318)
5840 - fix MDS lock inversions in getattr/reint paths (1844)
5841 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5842 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5843 - DLM race condition prevented some lock evictions (2328)
5844 - ENOMEM detection and retry on socknal sends (2230)
5845 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5846 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5847 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5848 - better error messages when a client is rejected during recovery (1505)
5849 - avoid cancelling locks which were never granted, after failure (2330)
5850 - fix i_sem/journal inversion in mds_client_add (2333)
5851 - fix truncate/getattr lock cycle deadlock (2334)
5852 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5853 - fix two rare exit paths which could leak an l_lock() ref (2321)
5854 - fencepost error in MDS/OST orphan recovery (2226)
5855 - make log record alignment 8 bytes (1988)
5856 - lstripe now fails when requested offset > ost_count (2237)
5857 - ensure that all kernel series have a complete list.h (1607)
5858 - fix crashes in special-file operations (2316)
5859 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5861 - allow configurable automake binary, for testing new versions
5862 - small update to the lfs documentation
5864 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5866 * fix negative export reference count in fsfilt_sync (2312)
5868 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5869 * release candidate 0.9.1
5871 - orphans are moved into the PENDING directory for possible recovery
5872 - replayed opens now open by fid for orphan/rename safety (1042)
5873 - last close of an orphan inode generates a transno (683)
5874 - chdir() and mount() now pin the directory entry (1020)
5875 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5876 - discard very old requests without processing them (1502)
5877 - remove some common, well-understood CERRORs (1505)
5878 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5879 - clear "grant" flags in OST replies until OST grant code lands (1644)
5880 - fix read performance by not clobbering i_blksize on client (1598)
5881 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5882 - make LRU size a /proc tunable, clears locks when reduced (707)
5883 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5884 - server locks take references on exports now (1558)
5885 - build fixes for 2.4.20-rh trees (1663)
5886 - return an error from lov_create if all OSCs are inactive (1751)
5887 - fix import levels when a reconnect happens without a timeout (1597)
5888 - exit early from mds_open if we get a lookup error (1749)
5889 - partial page read at EOF wouldn't wait for disk before sending (1642)
5890 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5891 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5892 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5893 - write/truncate lock inversion (1639)
5894 - Don't auto-load obdclass, portals modules during cleanup (1495)
5895 - fix timestamps from jumping to "now" (1763)
5896 - extra journal assertions (1648)
5897 - add an extra multiunlink test (1771)
5898 - fix read_record/write_record API (1776)
5899 - fix leak of offset_extent, possible incorrect i_size later (1772)
5900 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5901 - Debugging for the kqswnal_get_idle_tx problems (1820)
5902 - Allow recovery to be attempted multiple times (1536)
5903 - Write out MDS last_rcvd file after it is first created (1600)
5904 - Fix tx_descriptor leak in failed transmit situations (1827)
5905 - ext3 journaling fixes for assertion failure after IO error (1871)
5906 - class_export_put() on freed export after completion AST error (1896)
5907 - Fix revalidate looping in VFS (1322)
5908 - Don't access a freed export during MDS_REINT timeout (1521)
5909 - Add open-unlink recovery support on the MDS (1673,1764)
5910 - Return an error if no MDS data was read from last_rcvd (1946)
5911 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5912 - Don't setup a disk device that doesn't match exported UUID (317)
5913 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5914 - avoid committing NULL handle in force close
5915 - local.sh is now a one-stripe LOV configuration
5916 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5917 - fix ext3 htree duplicate directory entry corruption (1516)
5918 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5919 - update to new LOV EA format (2097)
5920 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5922 2003-06-15 Phil Schwan <phil@clusterfs.com>
5925 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5926 - per-import/export recovery handling (958, 931, 959)
5927 - multiple last-rcvd slots, for serving multiple FSes (949)
5928 - connections are again shared between multiple imp/exports (963, 964)
5929 - "umount -f" would hang if any requests needed to be sent (393, 978)
5930 - avoid pinning large req buffer by copying for queued messages (989)
5931 - add "uuid" to "lctl device" command to help upcalls (991)
5932 - "open" RPCs with transnos would confuse recovery counters (1037)
5933 - do proper endian conversion of all wire messages (288, 340, 891)
5934 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5935 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5936 - fix ptlrpc_connection leak in target_handle_connect (1174)
5937 - fix import refcounting bug in OST and MDS cleanup (1134)
5938 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5939 - fix very unlikely obd_types race condition (501)
5940 - remove osc_open hack for echo_client (1187)
5941 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5942 - a failure in read_inode2 leads to deadlock (1139)
5943 - cancel ack-locks as soon as transaction is committed (1072)
5944 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5945 - make sure to commitrw after any preprw to avoid deadlock (1162)
5946 - failing to execute a file in a lustre FS would lock inode (1203)
5947 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5948 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5949 - obd_uuid2tgt would walk off the end of the list (1255)
5950 - on IA64 the osc would give portals incorrect bulk size (1258)
5951 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5952 - fix lock inversion caused by new llite matching code (1282)
5953 - limit the number of dirty pages on a client to 10MB (1286)
5954 - timed out locks were not being corrected cancelled (1289)
5955 - fix O_DIRECT above 4GB on IA-32 (1292)
5956 * major user-visible changes
5957 - fail out/fail over policy now controlled by the upcall (993)
5959 - add OBD_PING to check server availability and failure (954)
5960 - lustre messages are now sent in sending host order (288, 340, 891)
5961 - add eadatalen to MDS getattr reply (340)
5962 - OST read replies may contain second buffer, with per-page status (593)
5964 2003-03-11 Phil Schwan <phil@clusterfs.com>
5967 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5968 - failed open()s could cause deadlock; fixed (867, 869)
5969 - stop cancelling OST locks when files are closed (481)
5970 - overlapping XID spaces caused network corruption (851, 853)
5971 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5972 - setattr_raw added, to do single-RPC, server-side setattrs
5973 - lmc/lconf syntax change for OST UUIDs
5974 - fix crashy race condition between ptlrpc_free_req and osc_close
5975 - don't use request in mdc_enqueue if we hit a timeout (889)
5976 - don't set the inode i_size for regular files from the MDS (896)
5977 - handle out of order completion AST (842)
5978 - don't LBUG if a lock request times out after receiving AST (913)
5979 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5980 - if a bad lock AST arrives, send an error instead of dropping entirely
5981 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5982 - fix leak in bulk IO when only partially completed (899, 900, 926)
5983 - fix O_DIRECT for ia64 (55)
5984 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5985 - C-z after timeout could hang a process forever; fixed (977)
5987 - client-side I/O cache (678, 924, 929, 941, 970)
5989 - READPAGE and SETATTRs which don't take server-side locks get
5992 2003-02-11 Phil Schwan <phil@clusterfs.com>
5995 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5996 - Fix fsfilt_extN_readpage() to read a full page of directory
5997 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5998 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5999 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6000 - Most intent operations were removed, in favour of a new RPC mode
6001 that does a single RPC to the server and bypasses most of the VFS
6002 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6003 - Aggressively cancel local locks on DLM servers
6004 - mds_reint_unlink sends EA to the client if it's the last nlink.
6005 client uses that EA to unlink OST objects.
6006 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6007 - recursive symlinks were fixed (439)
6008 - fixed NULL deref in DEBUG_REQ
6009 - filter_update_lastobjid no longer calls sync, which annoyed extN
6010 - fixed multi-client small-writes to a single file problem (445)
6011 - fixed mtime updates during file writes (607)
6012 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6013 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6014 - fixed Linux OST size reporting problem (444, 656)
6015 - OST now updates object mtime with writes or setattr (607, 619)
6016 - client verifies file size before zeroing page past EOF (445)
6017 - OST now writes last allocated objid to disk with allocation (108)
6018 - LOV on echo now works (409)
6020 - mds_reint_unlink sends a new buffer, with the EA included. this
6021 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6022 set if a regular file was being unlinked, and it was the last link
6023 - use PtlGet from the target for bulk writes (315)
6024 - OST now updates object mtime with writes or setattr (607, 619)
6025 - LDLM now has a grant-time callback to revalidate locked items, if
6027 - Many MDS operations were reorganized to combat race conditions
6029 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6030 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6032 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6035 - Fully reactivate OST imports after reconnection (512, others)
6036 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6037 - More graceful error handling for truncating on dead OST (515)
6038 - Don't error out unless we're actually accessing dead stripes (474)
6039 - Fix garbage sizes when stripes are missing (410)
6040 - LRU counters were broken, causing constant lock purge (433, 432)
6041 - garbage on read from stripes with failed OSTs (441)
6042 - mark OSCs as active before reconnecting during recovery (438)
6043 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6044 - lfind did not preserve OST order in output (443)
6045 - symlinks cause hung clients, incorrect data (439)
6046 - stop dereferencing request after dropping refcount (457)
6047 - don't LASSERT(spin_is_locked) on non-SMP (455)
6048 - fixes for many rename() bugs
6049 - fstat didn't correctly synchronize attributes (399)
6050 - server must handle lock cancellation during blocking AST prep (487)
6051 - bulk descriptors were free()d too soon (511)
6052 - fix paths in lconf, which would load incorrect modules (451, 507)
6053 - fix confusing lconf 'host not found' error message (386)
6054 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6055 - fix race condition in mdc_blocking_ast() for inode access (526)
6056 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6057 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6058 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6059 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6060 - fix lstripe in several ways (kernel side) (527)
6061 - fix request leak in ldlm_cli_enqueue (262)
6062 - incorrect OSC was marked inactive after OST failure
6063 - call mds_fs_cleanup before unmounting filesystem (524)
6064 - fix races between taking ns_lock and ldlm_lock_change_resource
6065 - fix races updating LOV export open file list
6066 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6067 - fix recovery NULL deref in ldlm_cli_cancel_unused
6068 - fix some DLM races by using new hash table for lock handles (419)
6069 - permit the client to specify desired inodes, at replay
6070 - duplicate requests when we queue them for replay reintegration
6071 - fix last_rcvd offset calculation
6072 - sync after each recovered transaction, so we always make progress
6073 - never, not always, ERESTART requests without transnos
6074 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6076 - skip replay if the MDS says that the client is already connected
6077 - don't check for a recovery-enabled export to match lctl's UUID
6078 - don't INC_USE_COUNT for phantom exports
6079 - don't crash when cleaning up phantom exports (567)
6080 - don't double-finish or set replay data for errored mdc_open requests
6081 - abort requests when they time out, so we don't get old replies
6082 - send/receive replies for AST messages again
6083 - if the client says that it doesn't have the lock, cancel it on the
6085 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6086 mark it as destroyed, it will all work out when decref is called
6087 - fix module use counts (22, 581)
6089 - ASTs now expect a reply (server cancels lock on error reply)
6091 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6094 - fix many simultaneous client startup (392)
6095 - fix dentry->d_it clobbering
6096 - credentials weren't being shipped for readdir/getattr operations
6097 - remove invalid assertions triggered during some concurrent MD
6099 - proper Lustre versions added (336, 389)
6100 - fix memory leak for create error case (398)
6101 - fix LOV locking bug that would get cli/srv out of sync
6102 - fix echo client over LOV (409)
6103 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6104 - fix double-O_EXCL intent crash (424)
6105 - avoid sending multiple lock CANCELs (352)
6107 - MDS can do multi-client recovery (modulo bugs in new code)
6109 - many updates, edits, cleanups
6111 2002-11-18 Phil Schwan <phil@clusterfs.com>
6114 - fix null d_it dereference (346)
6115 - fix full OST/dbench hang (333)
6116 - fix permission problem with file removal (286)
6117 - fix removal of OSCs from LOV when they fail
6118 - fix NULL deref during bulk timeout (214)
6119 - fix problems related to multiple filesystems on one MDS (241)
6120 - fixed serious subtle metadata locking bugs
6121 - free locks on clients when inodes are removed due to memory
6123 - fix inode pointer in lock data (285)
6124 - partial support for multiple MDS on a single host (241)
6125 - data locks weren't cancelled at clear_inode time (290, 311)
6126 - intent locks could lead to unbounded lock growth (205)
6127 - added a maximum lock count, an LRU list, and a flusher
6128 - fix multiple rename (365)
6129 - properly abstracted the echo client
6130 - OSC locked 1 byte too many; fixed
6131 - rewrote brw callback code:
6132 - fixed recovery bugs related to LOVs (306)
6133 - fixed too-many-pages-in-one-write crash (191)
6134 - fixed (again) crash in sync_io_timeout (214)
6135 - probably fixed callback-related race (385)
6137 - Add capability to MDS protocol
6138 - LDLM cancellations and callbacks on different portals
6140 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6143 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6144 - defer open object destruction to close time (601981, 138)
6145 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6146 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6147 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6148 - don't call obd_setattr in ll_file_release for destroyed objects
6149 * protocol change to lustre_msg: move |version| and add |flags|
6150 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6151 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6152 * added replay of create, unlink, link and rename operations during
6153 MDS failover; recovery should be much more robust now
6154 * remove failed OSCs from LOVs (only lov_create uses this so far)
6155 * the lustre-HOWTO was brought (more) up to date (582544)
6157 2002-10-23 Phil Schwan <phil@clusterfs.com>
6160 - in-use dentries weren't being reused properly (617851)
6161 - prevent multiple LDLM setup (599178)
6162 - fix LOV size calculations for truncate (617853)
6163 - fix client handling of MDS intent errors (POSIX)
6164 - fix permission bug in lovstripe.c test (624321)
6165 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6166 - truncate past end of file could corrupt data
6167 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6168 - a race in recovery could return ETIMEDOUT to apps (623947)
6169 - building outside the source directory was fixed
6170 * the lustre-HOWTO was brought (more) up to date (582544)
6171 * major progress was made on recovery functionality
6173 2002-10-10 Phil Schwan <phil@clusterfs.com>
6176 - recovery deadlock fix
6177 - rm -rf causes LBUG fix (617817)
6178 - file open by multiple tasks fix (618962)
6179 - directory permissions bugs (602707 and 620007)
6180 - journal_stop fixed with locking (611313)
6181 - O_APPEND failures resolved (618273, perhaps 614459)
6182 - lconf PATH fix (619770)
6183 - IA64 build fix (621450)
6184 - RPC buffer sizes scale with amount of memory
6186 2002-10-01 Phil Schwan <phil@clusterfs.com>
6189 - locks would be cancelled without throwing away data pages,
6190 resulting in inconsistent data (605627)
6191 - inode attributes were not always being refreshed (605627, 612449)
6192 - lconf now continues to cleanup after lctl reports an error
6193 - MDS now enforces user permissions (602707)
6194 - lprocfs cleanup fixed, but not yet enabled (614157)
6195 - fixed infinite server hang, should a client not respond to an AST
6196 - avoid going into recovery if user calls readlink() with a buffer
6197 that's too small (613941)
6198 - AST RPCs no longer require replies (614867) -- this may be changed
6199 - don't crash server if client sends an IOV that's too big (611336)
6200 - fixed lock conversion deadlock (611892)
6201 - fixed the following of symlinks (614622)
6202 * recovery: the server can remove locks from a client that dies, other
6203 clients can make progress
6204 * more extN patch fixes
6205 * compile-time configurable ptlrpc buffer allocations
6207 - collaborative read cache document
6208 - Lustre Lite Performance CDR document-in-progress
6210 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6213 - fix typo in patch-2.4.18
6215 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6218 - clear ptlrpc request each time in handle_incoming_request()
6219 - unlink of files now destroys the object on the OST
6221 2002-09-19 Peter Braam <braam@clusterfs.com>
6223 * add hard link support
6224 * change obdfile creation method
6225 * kernel patch changed
6227 2002-09-19 Peter Braam <braam@clusterfs.com>
6230 - stack overflow bug in extN fixed
6232 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6234 * documentation updates
6235 - add man pages for config tools
6236 - update tests/README to describe testing with new config tools
6237 - finish metadata API descriptions
6238 * bug fixes and cleanups
6239 - statfs workaround for 16TB limit
6240 - LOV stripe allocation improved, can stripe on subset of OSTs
6241 - LOV file size/IO offset was wrong for files > 4GB in size
6242 - object EA data was being dropped, caused files to be unreadable
6243 - memory overflow with non-LOV OST caused memory corruption
6244 - fixed regression tests to work with new config tools, obdfilter
6245 - fixed bug when directory size became larger than 1 block
6246 - fixed bug (for single client case) when PWD was deleted
6247 - invalidate local directory pages when doing intent-based ops
6248 - avoid LDLM oops when lock callback contained bad data
6250 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6252 * documentation updates
6253 * bug fixes and cleanups
6254 - configuration tools
6257 - 64-bit compile warnings
6258 - 64-bit internal statfs data
6260 * test_brw on persistent OST devices
6262 * lprocfs (disabled)
6264 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6266 * documentation updates
6267 * bug fixes and cleanups
6268 * configuration tools
6270 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6273 * many small fixes to 0_5_4
6274 * io/network handling
6275 * thinkos in MDS operations
6277 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6280 * crucial basic fixes to 0.5.3
6281 * IOR, Iozone work over Elan
6284 2002-08-07 Phil Schwan <phil@clusterfs.com>
6285 * version 0_5_3, our first alpha
6286 * we use the new Portals iovs
6287 * documentation updates
6288 * bug fixes and cleanups
6289 * small changes in the DLM wire protocol
6291 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6292 * version 0_5_1 with some initial stability,
6293 * locking on MD and file I/O.
6294 * documentation updates
6295 * several bug fixes since 0.5.0
6296 * small changes in wire protocol
6298 2002-07-18 Phil Schwan <phil@clusterfs.com>
6300 * delivered as Lustre Light Alpha
6301 * fixed a crash after handling invalid MDS requests
6302 * fixed directory pages for architectures with non-4k pages sizes
6304 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6305 * release version v0_4_4
6306 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6308 * Fixes a number of interruption problems with OST operations.
6309 * Update documentation for portals header changes
6310 * Move all wire protocol structs/defines to lustre_idl.h
6311 * Fixes symlink length bug.
6312 * Add tcpdump to repository.
6314 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6315 * release version v0_4_3
6316 * Fixes statfs for inodes on extN.
6317 * Fixes bug in runtests which would delete /etc/hosts.
6318 * Use 64-bit object IDs wherever possible (not into VFS though)
6319 Remove ost_get_info, which is unused by lustre, and out of date.
6321 2002-07-03 Peter Braam <braam@clusterfs.com>
6322 * release version v0_4_2 Fixes a lookup error (type not passed)
6323 * move forward to head of Portals
6324 * move forward to latest Lustre kernel
6326 2002-06-25 Peter Braam <braam@clusterfs.com>
6327 * release version v0_4_1. Hopefully stable on single node use.