3 * See http://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 * Server support for kernels:
6 * Client support for unpatched kernels:
9 2.6.18-238.19.1.el5 (RHEL5)
10 2.6.16.60-0.69.1 (SLES10)
11 * Recommended e2fsprogs version: 1.41.90.wc3
12 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
13 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
14 stack overflow. For more information, please refer to bugzilla 17630.
15 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
16 the re-exporting nodes to be mounted with "32bitapi" mount option
18 Severity : enhancement
20 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
22 Severity : enhancement
24 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
26 --------------------------------------------------------------------------------
28 09-30-2011 Whamcloud, Inc.
30 * Server support for kernels:
31 2.6.18-238.19.1.el5 (RHEL5)
32 2.6.32-131.6.1.el6 (RHEL6)
33 * Client support for unpatched kernels:
34 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
35 2.6.18-238.19.1.el5 (RHEL5)
36 2.6.32-131.6.1.el6 (RHEL6)
37 2.6.29.4-167.fc11 (FC11)
38 2.6.16.60-0.69.1 (SLES10)
40 * Recommended e2fsprogs version: 1.41.90.wc3
41 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
42 Lustre filesystem with 4K stack may cause a stack overflow. For more
43 information, please refer to bugzilla 17630.
44 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
45 mounted with "32bitapi" mount option
47 Severity : enhancement
48 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
50 Severity : enhancement
51 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
53 Severity : enhancement
54 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
57 Frequency : only with SLES10
59 Description: Use OFED "KMP" provided by Novell
60 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
61 Lustre is now built against this package. That means you need to
62 install the ofed-kmp package from Novell for the patchless client.
63 Note that the ofed-kmp that Novell ships may not exactly match the
64 kernel version but should still be compatible.
66 Severity : enhancement
68 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
69 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
71 --------------------------------------------------------------------------------
73 2010-07-15 Oracle, Inc.
75 * Server support for kernels:
76 2.6.18-164.11.1.el5 (RHEL 5)
77 2.6.18-164.11.1.0.1.el5 (OEL 5)
78 * Client support for unpatched kernels:
79 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
80 2.6.18-164.11.1.el5 (RHEL 5),
81 2.6.18-164.11.1.0.1.el5 (OEL 5)
82 2.6.16.60-0.42.8 (SLES 10),
84 2.6.29.4-167.fc11 (FC11)
85 * Recommended e2fsprogs version: 1.41.10-sun2
87 Severity : enhancement
89 Description: Distribute lfs_migrate script with lustre
90 Details : The lfs_migrate mentioned in the manual is now included in the
91 lustre tree in lustre/scripts and will distributed in lustre RPMs.
93 Severity : enhancement
95 Description: Failover nid list added to import proc file.
97 Severity : enhancement
99 Description: "som_preview" mount option required to use SOM
100 Details : The clients will need to use som_preview mount option
104 Description: Lustre does not do 1MB IOs to HW RAID
105 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
107 Severity : enhancement
109 Description: lfs quota output is not very convenient for awk/sed-parsing
110 Details : Some positions in lfs quota output table could be empty or
111 non-empty which made it hard to parse it with scripts, now
112 a dash is put instead of space where there is not supposed
117 Description: Update to OFED 1.5.1
118 Details : OFED 1.5.1 will be used in environments that don't use
119 vendor supplied OFED.
123 Description: Remove lnet. prefix from lctl params display
124 Details : Also renamed lnet.memused to lnet_memused to
129 Description: Use conf_param -d to remove permanant settings.
133 Description: Add list_param -R
134 Details : lctl list_param now supports recursive mode.
137 Bugzilla : 21829, 22312
138 Description: Fix llobdstat.
139 Details : 1. Now updating the write_bytes counter
140 2. Port script fixes from 1.8.
144 Description: decrease the usage of memory on clients.
145 Details : 1. On clients, recycle dentries and inodes unused.
146 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
151 Description: llobdstat fix and enhancement
152 Details : add a counter to set a limit to how many samples will be returned
153 fix a wildcard in the path to limit to obdfilter stats only
157 Description: lfs check only allowed for root
158 Details : Code cleanups and added a regresion test of non-root use of
163 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
164 Details : Fix hr->hr_index overload.
168 Description: Fix FIEMAP functionality.
172 Description: FC11 patchless client support added
176 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
177 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
181 Description: New acc-sm script and various test-framework enhancments.
185 Description: Previously landed patch was reverted.
189 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
193 Description: Add ldiskfs maxdirsize mount option.
197 Description: Fix mkfs.lustre support for 16TB LUNs.
201 Description: Print mmp_check_interval
202 Details : print mmp_check_interval and make it possible to abort
203 mount operation in case it takes too long.
207 Description: Add cascading_rw to lustre mpi tests.
211 Description: Store SOM data in EA.
212 Details : Also several SOM related cleanups.
216 Description: Add refcount to md_open_data
220 Description: Use refcount on lov_request_set to prevent memory corruption.
224 Description: Statahead: use parent dir's i_mutex for synchronization.
228 Description: loadgen improvements.
232 Description: Enable peer health detection in ksocklnd and ko2iblnd.
236 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
240 Description: Fix race on obd_devcie:md_stats
244 Description: Fix SLES 11 client builds.
245 Details : Resolved uninitalized pointer gcc warnings.
249 Description: Allow non-root access for "lfs check".
250 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
254 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
256 Severity : enhancement
258 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
259 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
263 Description: Sanity test_27p fails.
264 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
265 Sanity tests 27 were cleaned up and improved.
269 Description: Unify req format on client/servers
270 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
271 buffers and swabbers approach..
276 Description: drop join file stuff.
278 Severity : enhancement
280 Description: Add 2.6.27 and SLES11 (patchless client) support
284 Description: Fix in ptlrpc_expire_one_request() to print the signed time
287 Severity : enhancement
289 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
292 Severity : enhancement
294 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
296 Severity : enhancement
298 Description: correctly handle big reply message.
299 Details : send LNet event if reply is bigger then buffer and adjust this buffer
304 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
306 Severity : enhancement
308 Description: Add support for OEL5.
312 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
314 Severity : enhancement
316 Description: "weak-modules" support
317 Details : Implement "weak-modules" support which enables kernel modules
318 to be used with any kernel that implements the same kABI. In
319 order to achieve this modules are now installed in
320 /lib/modules/$(uname -r)/updates/kernel on all distributions.
324 Description: client could not reconnect to OST because of an active request.
325 Details : The client now retries to reconnect to the same server, if a
326 connect request failed with EBUSY or -EAGAIN. Test case 35b
327 conf-sanity.sh has been created.
331 Description: Don't allow make backward step on assignin osc next id.
332 Details : race between allocation next id and ll_sync thread can be cause
333 of set wrong osc next id and can be kill valid ost objects.
337 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
341 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
345 Description: Error handling in osc_statfs_interpret() has been improved.
346 Details : Check in osc_statfs_interpret() for EBADR.
350 Description: Do not send statfs() requests to OSTs disabled by administrator.
351 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
353 Severity : enhancement
355 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
357 Severity : enhancement
359 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
361 Severity : enhancement
363 Description: Enhancement for lfs(1) command to use numeric uid/gid.
367 Description: Improved cleanup procedure for conf-sanity test_46a.
371 Description: Do not start bulk transfer, if client tried to reconnect.
372 Details : Do not start bulk transfer, if client tried to reconnect,
373 but failed due to active rpcs. Check for exp_abort_active_req
374 has been implemented.
378 Description: Prevent inconsistences between linux and lustre mount structures.
379 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
380 Make the sleep interruptible.
382 Severity : enhancement
384 Description: provide server to client comms path
385 Details : server to client communications path via new LDLM_SET_INFO rpc
389 Description: truncate starts GFP_FS allocation under transaction causing deadlock
390 Details : ldiskfs_truncate calls grab_cache_page which may start page
391 allocation under an open transaction. This may lead to
392 calling prune_icache with consequent lustre reentrance.
396 Description: send correctly lsm on open replay
397 Details : MDS is trust to LSM size on replay open, but client can set wrong size
402 Description: enable client interface failover
403 Details : When a child reconnects from another NID, properly update export
404 nid hash position and ldlm reverse import.
406 Severity : enhancement
408 Description: Communicate OST degraded/readonly state via statfs to MDS
409 Details : Flags in the statfs returned from OSTs indicate whether the
410 OST is in a degraded RAID state, or if the filesystem has
411 turned read-only after a filesystem error is detected.
416 Description: don't panic if EPROTO was hit when reading symlink
417 Details : correctly handling request reference in error cases.
419 Severity : enhancement
421 Description: Add LustreNetLink, a kernel-userspace communcation path.
423 Severity : enhancement
425 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
430 Description: access to llog context before init.
431 Details : move handling CATALOGS file at osc layer and forbid access to llog
435 Frequency : rare, only when client is writing to OST on local node
437 Description: Avoid deadlock for local client writes
438 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
439 memory freeing context. This allows OST threads to set the
440 PF_MEMALLOC flag on task structures in order to allocate memory
441 from reserved pools and complete IO.
442 Use GFP_HIGHUSER for OST allocations for non-local client writes,
443 so that the OST threads generate memory pressure and allow
444 inactive pages to be reclaimed.
446 Severity : enhancement
448 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
453 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
454 Details : move ldlm namespace creation in setup phase to avoid grab
455 _lprocfs_lock with cli_sem held.
459 Description: Temporarily disable grant shrink.
460 Details : Disable the feature for debugging.
463 Frequency : only when formatting test filesystems
465 Description: Unable to run several mkfs.lustre on loop devices at the same time
466 Details : mkfs.lustre returns error 256 on the concurrent loop devices
467 formatting. The solution is to proper handle the error.
469 Severity : enhancement
471 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
473 Severity : enhancement
475 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
480 Description: resolve race between obd_disconnect and class_disconnect_exports
481 Details : if obd_disconnect will be called to already disconnected export he
482 forget release one reference and osc module can't unloaded.
484 Severity : enhancement
486 Description: Allow tuning service thread via /proc
487 Details : For each service a new
488 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
489 created that can be used to set min/max thread counts, and get the
490 current number of running threads.
495 Descriptoin: don't return error if have partially created objects for file.
496 Details : lov_update_create_set uses set->set_success as index for created
497 objects, so if some requests will be failed, they will have hole
498 at end of array and qos_shrink_lsm for allocate correct lsm.
500 Severity : enhancement
502 Description: Update OFED support to 1.4.1
507 Description: Reduce small size read RPC
508 Details : Set read-ahead limit for every file and only do read-ahead when
509 available read-ahead pages are bigger than 1M to avoid small size
512 Severity : enhancement
514 Description: move AT tunable parameters for more consistent usage
515 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
517 Severity : enhancement
519 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
520 Details : allow skip disconnected ost for send statfs request and hide error
526 Description: fix racy locking of mballoc block bitmaps causing BUG
527 Details : The locking of the mballoc buddy bitmap and the in-memory
528 block bitmap was using two different spin locks in some
529 cases. This made it possible to incorrectly access the
530 mballoc bitmap while another process was modifying it,
531 causing a sanity assertion to fail. While no on-disk corruption
532 was reported, there was some risk of this happening.
535 Frequency : rare, on llog test 6
537 Descriptoin: don't allow connect to already connected import
538 Details : allowing connect to already connected import is hide connecting problem.
543 Descriptoin: allow kill process which wait statahead result
544 Details : for some reasons 'ls' can stick in waiting result from statahead,
545 in this case need way for kill this process.
547 Severity : enhancement
549 Description: Add state history info file, enhance import info file
550 Details : Track import connection state changes in a new osc/mdc proc file;
551 add overview-type data to the osc/mdc import proc file.
553 Severity : enhancement
555 Description: MDS create should not wait for statfs RPC while holding DLM lock.
557 Severity : enhancement
559 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
564 Descriptoin: OST grant too much space to client even there are not enough space.
565 Details : Client will shrink its grant cache to OST if there are no write
566 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
567 this grant cache if there are already not enough avaible space
568 (left_space < total_clients * 32M).
571 Frequency : start MDS on uncleanly shutdowned MDS device
573 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
574 Details : stay in waiting mds<>ost recovery finished produce random bugs
575 due race between two ll_sync thread for one lov target. send
576 ACTIVATE event only if connect realy finished and import have
580 Frequency : rare, connect and disconnect target at same time
582 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
583 Details : don't call obd_disconnect under lov_lock. this long time
584 operation and can block ptlrpcd which answer to connect request.
589 Descriptoin: don't lose wakeup for imp_recovery_waitq
590 Details : recover_import_no_retry or invalidate_import and import_close can
591 both sleep on imp_recovery_waitq, but we was send only one wakeup
595 Frequency : always with long access acl
597 Descriptoin: mds can't pack reply with long acl.
598 Details : mds don't control size of acl but they limited by reint/getattr
601 Severity : enhancement
603 Description: Update to SLES10 kernel-2.6.16.60-0.33.
605 Severity : enhancement
607 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
610 Frequency : start MDS on uncleanly shutdowned MDS device
612 Descriptoin: aborting recovery hang on MDS
613 Details : don't throttle destroy RPCs for the MDT.
616 Frequency : on remount
618 Description: external journal device not working after the remount
619 Details : clear dev_rdonly flag for external journal devices in
625 Description: shutdown vs evict race
626 Details : client_disconnect_export vs connect request race.
627 if client will evicted at this time - we start invalidate
628 thread without referece to import and import can be freed
632 Frequency : rare, need acl's on inode.
634 Description: client can't handle ost additional correctly
635 Details : if ost was added after client connected to mds client can have
636 hit lnet_try_match_md ... to big messages to wide striped files.
637 in this case need teach client to handle config events about add
638 lov target and update client max ea size at that event.
640 Severity : enhancement
642 Description: Changelogs
643 Details : Changelogs are a lightweight mechanism to track filesystem
644 metadata and namespace changes. The changelog is recorded
645 permanently on the MDTs, and is periodically "consumed" / purged
646 when records are no longer needed.
648 Severity : enhancement
650 Description: compact fld format with extents
651 Details : Store range of seq rather than every seq in FLD. Seq
652 controller update FLD rather than clients. In Case of CMD, mdt0
653 has FLD, all other metadata server act as non persistent proxy
654 for FLD queries and cache fld entries in fld cache.
659 Description: don't skip ost target if they assigned to file
660 Details : Drop slow OSCs if we can, but not for requested start idx.
661 This means "if OSC is slow and it is not the requested
662 start OST, then it can be skipped, otherwise skip it only
663 if it is inactive/recovering/out-of-space.
667 Description: more cleanup in mds_lov
668 Details : not send LOV EA under replay, we can't know about they size at this
669 time. Don't allow client connect to mds before any ost connected,
670 for avoid problems with LOV EA size and returning EIO to client.
672 Severity : enhancement
674 Description: Interoperability at server side (Disk interoperability)
676 Severity : enhancement
678 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
680 Severity : enhancement
682 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
684 Severity : enhancement
686 Description: New client IO stack (CLIO).
688 Severity : enhancement
690 Description: Commit on sharing. Eliminate inter-client dependencies between
691 uncommitted transactions by doing transaction commits.
692 Thereby clients may recovery independently.
695 Frequency : Create a symlink file with a very long name
697 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
698 Details : If there is no extra space in the request for early cancels,
699 ldlm_req_handles_avail() returns 0 instead of a negative value.
701 Severity : enhancement
703 Description: Add /proc entry for import status
704 Details : The mdc, osc, and mgc import directories now have
705 an import directory that contains useful import data for debugging
708 Severity : enhancement
710 Description: Re-disable certain /proc logging
711 Details : Enable and disable client's offset_stats, extents_stats and
712 extents_stats_per_process stats logging on the fly.
715 Frequency : Only on FC kernels 2.6.22+
717 Description: oops in statahead
718 Details : Do not drop reference count for the dentry from VFS when lookup,
719 VFS will do that by itself.
721 Severity : enhancement
723 Description: Generic /proc file permissions
724 Details : Set /Proc file permissions in a more generic way to enable non-
725 root users operate on some /proc files.
729 Description: Hitting mdc_commit_close() ASSERTION
730 Details : Properly handle request reference release in
731 ll_release_openhandle().
735 Description: quota recovery deadlock during mds failover
736 Details : This patch includes att18982, att18236, att18237 in bz14840.
738 1. fix osts hang when mds does failover with quotaon
739 2. prevent watchdog storm when osts threads wait for the
744 Frequency : only patchless client
745 Description: add workaround for race between add/remove dentry from hash
747 Severity : enhancement
749 Description: Allow OST glimpses to return PW locks
753 Description: LBUG when llog conf file is full
754 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
758 Description: Prevent import from entering FULL state when server in recovery
762 Description: service mount cannot take device name with ":"
763 Details : Only when device name contains ":/" will mount treat it as
769 Description: replace ptlrpcd with the statahead thread to interpret the async
770 statahead RPC callback
774 Frequency : on recovery
775 Description: I/O failures after umount during fail back
776 Details : if client reconnected to restarted server we need join to recovery
777 instead of find server handler is changed and process self
778 eviction with cancel all locks.
780 Severity : enhancement
782 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
784 Severity : enhancement
786 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
788 Severity : enhancement
790 Description: Upcall on Lustre log has been dumped
791 Details : Allow for a user mode script to be called once a Lustre log has
792 been dumped. It passes the filename of the dumped log to the
793 script, the location of the script can be specified via
794 /proc/sys/lnet/debug_log_upcall.
799 Description: avoid idr_remove called for id which is not allocated.
800 Details : Move assigment s_dev for clustered nfs to end of initialization,
801 to avoid problem with error handling.
806 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
807 Details : When connection is reused this not moved from CONN_UNUSED_HASH
808 into CONN_USED_HASH and this prodice warning when put connection
809 again in unused hash.
812 Severity : enhancement
814 Description: File striping can now be set to use an arbitrary pool of OSTs.
816 Severity : enhancement
818 Description: Export bytes_read/bytes_write count on OSC/OST.
822 Description: Early reply size mismatch, MGC loses connection
823 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
824 the connect flags are properly negotiated.
829 Description: quotas are not honored with O_DIRECT
830 Details : all writes with the flag O_DIRECT will use grants which leads to
831 this problem. Now using OBD_BRW_SYNC to guard this.
835 Description: add quota statistics
836 Details : 1. sort out quota proc entries and proc code.
837 2. add quota statistics
839 Severity : enhancement
841 Description: enable quota support for HEAD.
845 Description: Properly propagate oinfo flags from lov to osc for statfs
846 Details : restore missing copy oi_flags to lov requests.
848 Severity : enhancement
850 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
851 : l_getgroups(8), lst(8), routerstat(8)
852 Details : included man pages for llobdstat(8), llstat(8),
853 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
855 Severity : enhancement
857 Description: configure's --enable-quota should check the
858 : kernel .config for CONFIG_QUOTA
859 Details : configure is terminated if --enable-quota is passed but
860 : no quota support is in kernel
864 Description: Remove portals compatibility
865 Details : Remove portals compatibility, not interoperable with releases
870 Description: Resolve device initialization race
871 Details : Prevent proc handler from accessing devices added to the
872 obd_devs array but yet be intialized.
874 Severity : enhancement
876 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
878 Severity : enhancement
880 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
885 Description: Using wrong pointer in osc_brw_prep_request
886 Details : Access to array[-1] can produce panic if kernel compiled with
887 CONFIG_PAGE_ALLOC enabled
891 Description: Client runs out of low memory
892 Details : Consider only lowmem when counting initial number of llap pages
896 Description: *optional* service tags registration
897 Details : if the "service tags" package is installed on a Lustre node
898 When the filesystem is mounted, a local-node service tag will
899 be created. See http://inventory.sun.com/ for more information
900 about the Service Tags asset management system.
904 Description: Kernel BUG tries to release flock
905 Details : Lustre does not destroy flock lock before last reference goes
906 away. So always drop flock locks when client is evicted and
907 perform unlock regardless of successfulness of speaking to MDS.
911 Description: add refcount protection for osc callbacks, avoid panic on shutdown
915 Description: sanity test 65a fails if stripecount of -1 is set
916 Details : handle -1 striping on filesystem in ll_dirstripe_verify
921 Description: ASSERTION(CheckWriteback(page,cmd)) failed
922 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
925 Severity : enhancement
927 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
931 Description: do not process already freed flock
932 Details : flock can possibly be freed by another thread before it reaches
933 to ldlm_flock_completion_ast.
937 Description: LBUG during stress test
938 Details : Need properly lock accesses the flock deadlock detection list.
942 Description: oops in page fault handler
943 Details : kernel page fault handler can return two special 'pages' in error
944 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
948 Description: timeout with invalidate import.
949 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
950 should be handled by ptlrpcd. This produce long age waiting and
951 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
953 Severity : enhancement
955 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
959 Description: enable MGS and MDT services start separately
960 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
961 a co-located MGS without starting the MGS, which is a complement
962 to 'nosvc' mount option.
966 Frequency : after recovery
967 Description: precreate to many object's after del orphan.
968 Details : del orphan st in oscc last_id == next_id and this triger growing
969 count of precreated objects. Set flag LOW to skip increase count
970 of precreated objects.
974 Frequency : rare, on clear nid stats
975 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
976 Details : when clean nid stats sometimes try destroy live entry,
977 and this produce panic in free.
981 Description: Stack overflow during MDS log replay
982 ease stack pressure by using a thread dealing llog_process.
986 Description: wait until IO finished before start new when do lock cancel.
987 Details : VM protocol want old IO finished before start new, in this case
988 need wait until PG_writeback is cleared until check dirty flag and
989 call writepages in lock cancel callback.
991 Severity : enhancement
993 Description: using special macro for print time and cleanup in includes.
997 Description: mds_mfd_close() ASSERTION(rc == 0)
998 Details : In mds_mfd_close(), we need protect inode's writecount change
999 within its orphan write semaphore to prevent possible races.
1003 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1007 Frequency : rare, on shutdown ost
1008 Description: don't hit live lock with umount ost.
1009 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1010 use shrink_dcache_sb instead.
1014 Description: don't panic with use echo client
1015 Details : echo client pass NULL as client nid pointer and this produce null
1016 pointer dereference.
1020 Description: fix build on ppc32
1021 Details : compile code with -m64 flag produce wrong object file for ppc32.
1025 Description: add message levels for liblustreapi
1029 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1030 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1031 will result in an -EAGAIN and in the exit call path we call
1032 alloc_rr() which will with increasing aggressiveness attempt to
1033 aquire precreated objects on the minimum number of required OSCs.
1037 Description: Use old size assignment to avoid deadlock
1038 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1039 the scheduling while holding a spinlock. We do not need locking
1040 for size in ll_update_inode() because size is only updated from
1041 the MDS for directories or files without objects, so there is no
1042 other place to do the update, and concurrent access to such inodes
1043 are protected by the inode lock.
1047 Description: resolve "_IOWR redefined" build error on SLES10
1051 Description: dump the memory debugging after all modules are unloaded to
1052 suppress false negative in conf_sanity test 39
1054 Severity : enhancement
1056 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1061 Description: fix SLES kernel versioning
1062 Details : the kernel version for our SLES 10 kernel did not include a "-"
1063 before the "smp" at the end. while this was not a problem in
1064 general, it did mean that software trying to use the kernel
1065 version to try to detect a vendor specific kernel would fail.
1066 this was most evident by the OFED build scripts.
1070 Description: Don't update lov_desc members until making sure they are valid
1071 Details : When updating lov_desc members via proc fs, need fix their
1072 validities before doing the real update.
1076 Description: don't put request into delay list while invalidate in flight.
1077 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1078 import in flight. this produce timeout for invalidate and sometimes
1079 can cause stale data.
1083 Frequency : on ppc only
1084 Description: not convert ost objects for directory because it's not exist.
1085 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1089 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1091 Severity : enhancement
1093 Description: Update to support 2.6.22.14 vanilla kernel.
1097 Frequency : rare, on recovery
1098 Description: read procfs can produce deadlock in some situation
1099 Details : Holding lprocfs lock which send rpc can produce block for destroy
1100 obd objects and this also block reconnect with -EALREADY. This
1101 isn't fix all lprocfs bugs - but make it rare.
1103 Severity : enhancement
1105 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1108 Frequency : frequent on X2 node
1110 Description: mdc_set_open_replay_data LBUG
1111 Details : Set replay data for requests that are eligible for replay.
1115 Description: lustre_mgs: operation 101 on unconnected MGS
1116 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1117 MGC, and late on MGC cannot successfully connect to MGS and a lot
1118 of the error messages complaining that MGS is not connected.
1122 Frequency : on network error
1123 Description: panic with double free request if network error
1124 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1125 but it's true only for synchronus enqueue, for async enqueue
1126 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1129 Severity : enhancement
1131 Description: client-side metadata stat-ahead during readdir(directory readahead)
1132 Details : perform client-side metadata stat-ahead when the client detects
1133 readdir and sequential stat of dir entries therein
1136 Frequency : on start mds
1138 Description: Implement get_info(last_id) in obdfilter.
1141 Frequency : occasional
1143 Description: Correctly check stale fid, not start epoch if ost not support SOM
1144 Details : open with flag O_CREATE need set old fid in op_fid3 because
1145 op_fid2 was overwritten with new generated fid, but mds can answer
1146 with one of these two fids and both is not stale. Setattr
1147 incorrectly started an epoch and assume will be called
1148 done_writing, but without SOM done_writing ever being called.
1151 Frequency : rare, depends on device drivers and load
1153 Description: MDS or OSS nodes crash due to stack overflow
1154 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1155 In some cases, in conjunction with device drivers that use a lot
1156 of stack the MDS (or possibly OSS) service threads could overflow
1157 the stack. One change which was identified to consume additional
1158 stack has been reworked to avoid the extra stack usage.
1161 Frequency : occasional
1163 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1164 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1165 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1166 oscc_has_objects and see OSC has no objests and call
1167 oscc_internal_create to resend create request.
1169 Severity : enhancement
1171 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1173 Severity : enhancement
1175 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1178 Frequency : very rare
1180 Description: Fix replay if there is an un-replied request and open
1181 Details : In some cases, older replay request will revert the
1182 mcd->mcd_last_xid on MDS which is used to record the client's
1183 latest sent request.
1185 Severity : enhancement
1187 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1189 Severity : enhancement
1191 Description: Add rhel5 support to HEAD.
1193 Serverity : enhancement
1195 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1200 Description: Don't allow skipping OSTs if index has been specified.
1201 Details : Don't allow skipping OSTs if index has been specified, make
1202 locking in internal create lots better.
1206 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1207 Details : access to bitfield in structure is always rounded to long
1208 and this produce problem with not atomic change any bit.
1212 Description: Lustre make rpms failed.
1213 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1214 builds Lustre rpms from tarball.
1217 Frequency : rare on shutdown ost
1219 Description: If llog cancel was not send before clean_exports phase, this can
1220 produce deadlock in llog code.
1221 Details : If llog thread has last reference to obd and call class_import_put
1222 this produce deadlock because llog_cleanup_commit_master wait when
1223 last llog_commit_thread exited, but this never success because was
1224 called from llog_commit_thread.
1228 Description: allow userland application know is lost one of stripes.
1229 Details : fill lvb_blocks with error code on ost and return it to
1230 application if error flag found.
1234 Description: NULL lov_tgts causing MDS oops
1235 Details : more safe checks for NULL lov_tgts for avoid oops.
1237 Severity : enhancement
1239 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1243 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1244 Details : If an OST is down the MDS will hang indefinitely in
1245 obd_statfs_rqset() waiting for the statfs data. While for
1246 MDS QOS usage of statfs, it should not stuck in waiting.
1248 Severity : enhancement
1250 Description: remote_acl support
1251 Details : Support ACL-based permission check for remote user.
1252 Support setfacl/getfacl for remote user with the utils
1253 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1254 the system "{s,g}etfacl" utils.
1256 Severity : enhancement
1258 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1260 Severity : enhancement
1262 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1266 Description: make mgs_setparam() handle fsname containing dash
1267 Details : fsname containing a dash does not work with lctl conf_param
1269 Severity : enhancement
1271 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1273 Severity : enhancement
1275 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1279 Frequency : rare, at shutdown
1280 Description: access already free / zero obd_namespace.
1281 Details : if client_disconnect_export was called without force flag set,
1282 and exist connect request in flight, this can produce access to
1283 NULL pointer (or already free pointer) when connect_interpret
1284 store ocd flags in obd_namespace.
1288 Frequency : only at startup
1289 Description: not alloc memory with spinlock held.
1290 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1291 if any spinlock held.
1293 Severity : enhancement
1295 Description: make lustre randomly fail allocating memory
1296 Details : Make lustre randomly failed allocating memory for testing purpose.
1298 Severity : enhancement
1300 Description: lost problems with lov objid file
1301 Details : Fixes some scability and access to not inited memory problems
1302 in work with lov objdid file.
1307 Description: lfs find does not continue on file error
1308 Details : Continue other files processing when a file/dir is absent.
1312 Description: Inconsistent usage of lustre_pack_reply()
1313 Details : Standardize the usage of lustre_pack_reply() such that it
1314 always generate a CERROR on failure.
1317 Frequency : occasional
1319 Description: MDS hang or stay in waiting lock
1320 Details : If client receive lock with CBPENDING flag ldlm need send lock
1321 cancel as separate rpc, to avoid situation when cancel request
1322 can't processed due all i/o threads stay in wait lock.
1326 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1330 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1334 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1338 Description: Update to RHEL4 latest kernel.
1340 Severity : enhancement
1342 Description: Build SLES10 patchless client fails
1343 Details : The configure was broken by run ./configure with
1344 --with-linux-obj=.... argument for patchless client. When the
1345 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1346 can't search header adequately. Use absolute path such as
1347 -I($LINUX)/include instead.
1351 Description: interrupt oig_wait produce painc on resend.
1352 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1353 private set, and this produce situation when rq_ptlrpcd_data not
1354 copyed to new allocated request and triggered LBUG on assert
1355 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1356 ptlrpcd set if request is changed and can be safety changed to use
1361 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1362 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1363 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1364 these macros and replace them with their difinitions instead.
1368 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1369 Details : Update which_patch & target file for SLES10 latest kernel.
1372 Description: Remove client patches from SLES 10 kernel.
1373 Details : This causes SLES 10 clients to behave as patchless clients
1374 even on a Lustre-patched (server) kernel.
1376 Severity : enhancement
1378 Description: self-adjustable client's lru lists
1379 Details : use adaptive algorithm for managing client cached locks lru
1380 lists according to current server load, other client's work
1381 pattern, memory activities, etc. Both, server and client
1382 side namespaces provide number of proc tunables for controlling
1385 Severity : enhancement
1387 Description: light-weight GSS support
1388 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1389 and reduce performance overhead.
1391 Severity : enhancement
1393 Description: Linux keyring support
1394 Details : Support using service of Linux keyring for Lustre GSS internal
1395 context refresh/cache mechanism.
1399 Description: Fix errors in lfs documentation
1400 Details : Fixes man pages
1404 Description: don't use GFP_* in generic Lustre code.
1405 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1409 Description: obdclass is limited by single OBD_ALLOC(idarray)
1410 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1414 Description: warnings with build patchless client with vanila 2.6.19 and up
1415 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1416 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1420 Frequency : Only for SLES
1421 Description: sanity_quota fail test_1
1422 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1423 file, which makes TSTID[2] inunique.
1427 Frequency : Only for SLES9
1428 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1429 Details : Update target/ChangeLog/which_patch .
1433 Frequency : Only for test_10 in sanity-quota.sh
1434 Description: a bug in quota test code
1435 Details : checking a test flag in a wrong place
1439 Frequency : Only for test_13 in sanity-quota.sh
1440 Description: a bug in quota test script
1441 Details : shouldn't check overfull allocated quota
1445 Description: jbd statistics
1446 Details : Port older jbd statistics patch for sles10
1448 Severity : enhancement
1450 Description: Add printing inode info into message about error in writepage.
1455 Description: reply_lock_interpret crash due to race with it and lock cancel.
1456 Details : Do not replay locks that are being cancelled. Do not reference
1457 locks by their address during replay, just by their handle.
1461 Frequency : When flocks are used.
1462 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1463 Details : Flock locks might destroy just granted lock if it could be merged
1464 with another existing flock, this is done in completion handler,
1465 so teach ldlm_cli_enquque_fini that this is a valid case for
1471 Description: Oops in read and write path when failing to allocate lock.
1472 Details : Check if lock allocation failed and return error back.
1476 Description: lstripe command fails for valid OST index
1477 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1478 instead of lov->desc.ld_active_tgt_count.
1482 Description: sanity.sh failed test 103
1483 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1488 Description: df doesn't work properly if diskfs blocksize != 4K
1489 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1493 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1494 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1496 Severity : enhancement
1498 Description: Adaptive timeouts
1499 Details : RPC timeouts adapt to changing server load and network
1500 conditions to reduce resend attempts and improve recovery time.
1504 Description: llapi_file_create() does not allow some changes
1505 Details : add llapi_file_open() that allows specifying the mode and
1506 open flags, and also returns an open file handle.
1510 Description: merge and cleanup kernel patches.
1511 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1515 Description: Add journal checksum support.(Kernel part)
1516 Details : The journal checksum feature adds two new flags i.e
1517 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1518 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1519 indicates that the commit block contains the checksum for
1520 the blocks described by the descriptor blocks. Now commit
1521 record can be sent to disk without waiting for descriptor
1522 blocks to be written to disk. This behavior is controlled
1523 using JBD2_FEATURE_ASYNC_COMMIT flag.
1527 Description: OSS needs mutliple precreate threads
1528 Details : Add ability to start more than one create thread per OSS.
1532 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1533 Details : execute lfs setstripe on client
1537 Description: mds_obd_create error creating tmp object
1538 Details : When the user sets quota on root, llog will be affected and can't
1539 create files and write files.
1542 Frequency : Always on ia64 patchless client, and possibly others.
1544 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1545 Details : This allows the patchless client to be loaded on architectures
1546 without this export.
1550 Description: RedHat Update kernel for RHEL5
1551 Details : Add the kernel config file for RHEL5.
1555 Description: RedHat Update kernel for RHEL5
1556 Details : Modify the kernel config file more closer RHEL5.
1560 Description: Build failure against Centos5 (RHEL5)
1561 Details : Define PAGE_SIZE when it isn't present.
1565 Description: test_7 on sanity-quota.sh can't be used on separate machines
1566 Details : add facet to handle it
1570 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1571 Details : with NFS, the anon dentry's parent was set to itself in
1572 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1573 corresponding dentry other than use rec->ur_name.
1575 Severity : enhancement
1577 Description: lfs setstripe enhancement
1578 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1582 Description: enable data checksumming by default
1583 Details : enable checksum by default, allow --disable-checksum
1584 configure option and "-o nochecksum" mount option. Checksums
1585 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1589 Description: lustre support for RHEL5
1590 Details : Add support for RHEL5.
1594 Description: Client eviction due to failover config
1595 Details : after a connection loss, the lustre client should attempt to
1596 reconnect to the last active server first before trying the
1597 other potential connections.
1601 Description: when mds and osts use different quota unit(32bit and 64bit),
1602 quota will be released repeatly.
1603 Details : void sending multiple quota reqs to mds, which will keep the
1604 status between the reqs.
1608 Description: osts not allocated evenly to files
1609 Details : change the condition to increase offset_idx
1612 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1614 Description: Data corruption for OSTs that are formatted larger than 2TB
1616 Details : When generating the bio request for lustre file writes the
1617 sector number would overflow a temporary variable before being
1618 used for the IO. The data reads correctly from Lustre (which
1619 will overflow in a similar manner) but other file data or
1620 filesystem metadata may be corrupted in some cases.
1624 Description: Tune the kernel for good SCSI performance.
1625 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1626 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1631 Description: rewrite ext2-derived code in obdclass/uuid.c
1632 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1633 from scratch preserving functionality.
1637 Description: "lfs find -obd UUID" prints directories
1638 Details : "lfs find -obd UUID" will return all directory names instead
1639 of just file names. It is incorrect because the directories
1640 do not reside on the OSTs.
1644 Description: use i_size_read and i_size_write in 2.6 port
1645 Details : replace inode->i_size access with i_size_read/write()
1648 Frequency : when removing large files
1650 Description: scheduling issue during removal of large Lustre files
1651 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1652 It causes scheduling issues when removing large files (17TB in the
1656 Frequency : only with liblustre clients on XT3
1658 Description: evictions taking too long
1659 Details : allow llrd to evict clients directly on OSTs
1662 Frequency : only on ppc
1664 Description: /proc/fs/lustre/devices broken on ppc
1665 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1666 We should make sure the type of 'index' is loff_t and then cast
1667 explicitly as needed below. Do not assign an explicitly cast
1671 Frequency : only for rhel5
1673 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1674 Details : Modify the target file & which_kernel.
1676 Severity : enhancement
1678 Description: omit set fsid for export NFS
1679 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1680 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1684 Frequency : Always, for kernels after 2.6.16
1685 Description: Fix warning idr_remove called for id=.. which is not allocated.
1686 Details : Last kernels save old s_dev before kill super and not allow
1687 to restore from callback - restore it before call kill_anon_super.
1691 Description: buffer overruns could theoretically occur
1692 Details : llapi_semantic_traverse() modifies the "path" argument by
1693 appending values to the end of the origin string, and a
1694 overrun may occur. Adding buffer overrun check in liblustreapi.
1698 Description: Fix error on 'ls .' at the top of the Lustre mount.
1699 Details : Don't revalidate dentry if it is a root dentry.
1703 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1704 Details : Modify vm-tunables-rhel4.patch.
1708 Description: Kernel config for 2.6.18-vanilla.
1709 Details : Modify targets/2.6-vanilla.target.in.
1710 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1711 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1712 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1713 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1717 Description: improve handling recoverable errors
1718 Details : if request processig with error which can be recoverable on server
1719 request should be resend, otherwise page released from cache and
1724 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1725 Details : Modify target file & which_patch.
1726 A flaw was found in the IA32 system call emulation provided
1727 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1728 value could be stored in the %RAX register, which could trigger an
1729 out-of-bounds system call table access. An untrusted local user
1730 could exploit this flaw to run code in the kernel
1731 (ie a root privilege escalation). (CVE-2007-4573).
1735 Description: O_DIRECT bypasses client statistics.
1736 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1737 read_ahead_stats were updated. copied the stats section from
1738 osc_send_oap_rpc() into async_internal().
1742 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1746 Description: change order of libsysio includes
1747 Details : '#include sysio.h' should always come before '#include xtio.h'
1751 Description: handle "serious error: objid * already exists" more gracefully
1752 Details : If LAST_ID value on disk is smaller than the objects existing in
1753 the O/0/d* directories, it indicates disk corruption and causes an
1754 LBUG(). If the object is 0-length, then we should use the existing
1755 object. This will help to avoid a full fsck in most cases.
1757 Severity : enhancement
1759 Description: adapt the lustre_config script to support the upgrade case
1760 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1761 server targets to 1.6 in parallel.
1765 Description: To avoid grant space > avaible space when the disk is almost
1766 full. Without this patch you might see the error "grant XXXX >
1767 available" or some LBUG about grant, when the disk is almost
1769 Details : In filter_check_grant, for non_grant cache write, we should
1770 check the left space by if (*left > ungranted + bytes), instead
1771 of (*left > ungranted), because only we are sure the left space
1772 is enough for another "bytes", then the ungrant space should be
1773 increase. In client, we should update cl_avail_grant only there
1774 is OBD_MD_FLGRANT in the reply.
1778 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1779 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1780 described in CVE-2007-4573. This problem could allow untrusted
1781 local users to gain root access.
1784 Frequency : when using O_DIRECT and quotas
1786 Description: Incorrect file ownership on O_DIRECT output files
1787 Details : block usage reported by 'lfs quota' does not take into account
1788 files that have been written with O_DIRECT.
1793 Description: touch file failed when fs is not full
1794 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1795 otherwise we can get ENOSP while fs is not full.
1799 Description: parallel lock callbacks
1800 Details : Instead of sending blocking and completion callbacks as separated
1801 requests, adding them to a set and sending in parallel.
1804 Frequency : only for Cray XT3
1805 Bugzilla : 12829/13455
1806 Description: Changing primary group doesn't change the group lustre assigns to
1808 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1809 overridden with the primary group provided by the group upcall,
1810 whereas the supplied fsgid can be trusted if it is in the list of
1811 supplementary groups returned by the group upcall.
1813 Severity : enhancement
1815 Description: Allow masking D_WARNING, D_ERROR messages from console
1816 Details : Console messages can now be disabled via lnet.printk.
1820 Description: User code with malformed file open parameter crashes client node
1821 Details : Before packing join_file req, all the related reference should be
1822 checked carefully in case some malformed flags cause fake
1823 join_file req on client.
1827 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1828 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1829 mode is equal to the granted mode. However, it should also check
1830 that LDLM_FL_CP_REQD is not set.
1834 Description: Heavy nfs access might result in deadlocks
1835 Details : After ELC code landed, it is now improper to enqueue any mds
1836 locks under och_sem, because enqueue might want to decide to
1837 cancel open locks for same inode we are holding och_sem for.
1841 Description: Client eviction while running blogbench
1842 Details : A lot of unlink operations with concurrent I/O can lead to a
1843 deadlock causing evictions. To address the problem, the number of
1844 oustanding OST_DESTROY requests is now throttled to
1845 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1846 callbacks are processed in priority.
1850 Description: enable ACLs on MDS by default
1851 Details : ACLs must be enabled on MDS by default.
1854 Frequency : PPC/PPC64 only
1856 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1857 Details : fix duplicated definitions between asm-ppc64/types.h and
1858 lustre_types.h on PPC.
1861 Frequency : PPC/PPC64 only
1863 Description: asm-ppc/segment.h does not exist
1864 Details : fix compile issue on PPC.
1868 Description: better handle error messages in extents code
1871 Frequency : RHEL4 only
1873 Description: mkfs is very slow on IA64/RHEL4
1874 Details : A performance regression has been discovered in the MPT Fusion
1875 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1876 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1877 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1879 Severity : enhancement
1881 Description: SNMP support enhancement
1882 Details : Adding total number of sampled request for an MDS node in snmp
1885 Severity : enhancement
1887 Description: Optimize ldlm waiting list processing for PR extent locks
1888 Details : When processing waiting list for read extent lock and meeting read
1889 lock that is same or wider to it that is not contended, skip
1890 processing rest of the list and immediatelly return current
1891 status of conflictness, since we are guaranteed there are no
1892 conflicting locks in the rest of the list.
1896 Description: Time out and refuse to reconnect
1897 Details : When the failover node is the primary node, it is possible
1898 to have two identical connections in imp_conn_list. We must
1899 compare not conn's pointers but NIDs, otherwise we can defeat
1900 connection throttling.
1904 Description: port llog fixes from b1_6 into HEAD
1905 Details : Port llog reference couting and some llog cleanups from b1_6
1906 (bug 10800) into HEAD, for protect from panic and access to already
1907 free llog structures.
1911 Description: Detect stride IO mode in read-ahead
1912 Details : When a client does stride read, read-ahead should detect that and
1913 read-ahead pages according to the detected stride pattern.
1917 Description: data checksumming impacts single node performance
1918 Details : add support for several checksum algorithm. Currently, only CRC32
1919 and Adler-32 are supported. The checksum type can be changed on
1920 the fly via /proc/fs/lustre/osc/*/checksum_type.
1924 Description: use adler32 for page checksums
1925 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1930 Description: build for x2 fails
1931 Details : fix compile issue on Cray systems.
1935 Description: Properly match for duplicate locks
1936 Details : Due to different lock order from skiplists code, we need to
1937 traverse entire list for now
1940 Frequency : only on PPC/SLES10
1942 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1943 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1944 be defined. Add a hack in mkfs_lustre.c to work around this compile
1949 Description: LASSERT on MDS when client holding flock lock dies
1950 Details : ldlm pool logic depends on number of granted locks equal to
1951 number of released locks which is not true for flock locks, so
1952 just exclude such locks from consideration.
1956 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1957 Details : Use fsfilt_sync() for both the whole filesystem sync and
1958 individual file sync to eliminate dangerous inode locking
1959 with I_LOCK that can lead to a deadlock.
1963 Description: performance in 1.6.3
1964 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1965 sure that 1MB requests are not fragmented by the block layer.
1967 Severity : enhancement
1969 Description: organize the server-side client stats on per-nid basis
1970 Details : Change the structure of stats under obdfilter and mds to
1978 The "uuid"s file would list the uuids of _active_ exports.
1979 And the clear entry is to clear all stats and stale nids.
1981 Severity : enhancement
1983 Description: eliminate client locks in face of contention
1984 Details : file contention detection and lockless i/o implementation
1985 for contended files.
1989 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1992 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1994 Description: MDS may lose file striping (and hence file data) in some cases
1995 Details : If there are additional extended attributes stored on the MDS,
1996 in particular ACLs, SELinux, or user attributes (if user_xattr
1997 is specified for the client mount options) then there is a risk
1998 of attribute loss. Additionally, the Lustre file striping
1999 needs to be larger than default (e.g. striped over all OSTs),
2000 and an additional attribute must be stored initially in the
2001 inode and then increase in size enough to be moved to the
2002 external attribute block (e.g. ACL growing in size) for file
2007 Description: skiplist implementation simplification
2008 Details : skiplists are used to group compatible locks on granted list
2009 that was implemented as tracking first and last lock of each
2010 lock group the patch changes that to using doubly linked lists
2014 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2015 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2016 because LASSERT on both the data supplied by a client, and the
2017 data on disk is dangerous and incorrect.
2019 Severity : enhancement
2021 Description: Slow truncate/writes to huge files at high offsets.
2022 Details : Directly associate cached pages to lock that protect those pages,
2023 this allows us to quickly find what pages to write and remove
2024 once lock callback is received.
2028 Description: more ldlm soft lockups
2029 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2030 starve other threads from the resource lock for a long time in
2031 case of long waiting queue, so change the debug level from
2032 D_OTHER to the less frequently used D_INFO.
2034 Severity : enhancement
2036 Description: add -gid, -group, -uid, -user options to lfs find
2040 Description: Hung threads in invalidate_inode_pages2_range
2041 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2042 once one is completed. As a result, some RPCs are stuck in the
2043 queue and are never sent.
2047 Description: filter threads hungs on waiting journal commit
2048 Details : Cleanup filter group llog code, then only filter group llog will
2049 be only created in the MDS/OST syncing process.
2053 Description: Procfs and llog threads access destoryed import sometimes.
2054 Details : Sync the import destoryed process with procfs and llog threads by
2055 the import refcount and semaphore.
2057 Severity : enhancement
2059 Description: openlock cache of b1_6 port to HEAD
2064 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2065 Details : If insertion of an extent fails, then discard the inode
2066 preallocation and free data blocks else it can lead to duplicate
2071 Description: don't always update ctime in ext3_xattr_set_handle()
2072 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2073 In some cases the ctime should not be updated, for example for
2074 2.0->1.8 compatibility it is necessary to delete an xattr and it
2075 should not update the ctime.
2079 Bugzilla : 15713/16362
2080 Description: Assertion in iopen_connect_dentry in 1.6.3
2081 Details : looking up an inode via iopen with the wrong generation number can
2082 populate the dcache with a disconneced dentry while the inode
2083 number is in the process of being reallocated. This causes an
2084 assertion failure in iopen since the inode's dentry list contains
2085 both a connected and disconnected dentry.
2089 Description: assertion failure in ldlm_handle2lock()
2090 Details : fix a race between class_handle_unhash() and class_handle2object()
2091 introduced in lustre 1.6.5 by bug 13622.
2096 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2097 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2102 Description: oops in sd_iostats_seq_show()
2103 Details : unloading/reloading the scsi low level driver triggers a kernel
2104 bug when trying to access the sd iostat file.
2109 Description: Kernel panics during QLogic driver reload
2110 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2111 patch, causing memory corruption.
2116 Description: journal_dev option does not work in b1_6
2117 Details : pass mount option during pre-mount.
2119 Severity : enhancement
2121 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2122 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2123 extent information of a file. It can be used to map logical blocks
2124 in a file to physical blocks in the block device.
2128 Description: LDLM soft lockups - improvement
2129 Details : It is be possible to send the lock handle along with each read
2130 or write request because the client is already doing a lock match
2131 itself so there isn't any reason the OST should have to re-do that
2135 Frequency : only on Cray X2
2137 Description: X2 build failures
2138 Details : fix build failures on Cray X2.
2142 Description: xid & resent requests
2143 Details : Initialize RPC XID from clock at startup (randomly if clock is
2146 Severity : enhancement
2148 Description: Add lustre_start utility to start or stop multiple Lustre servers
2153 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2154 Details : In case of memory pressure, list_del() can be called twice on
2155 req->rq_history_list, causing a kernel oops.
2159 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2160 Details : fix stack overflow in the distributed lock manager by defering
2161 export eviction after a failed AST to the elt thread instead of
2162 handling it in the dlm interpret routine.
2166 Description: Convert some comments to new format.
2167 Details : Update documenting comments to match doxygen conventions.
2171 Description: Grammar fixes.
2172 Details : A couple of trivial spelling fixes.
2176 Description: OSD_COUNTERS-mandatory
2177 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2178 to simplify the code.
2182 Description: simplify cmm_device freeing logic.
2183 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2187 Description: Add lockdep support to dt_object_operations locking interface.
2188 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2189 parameter indicating lock ordering. Update mdd code to use new
2194 Description: Introduce failloc constants for lockless IO tests.
2195 Details : Add two new failloc constants to test lockless IO. Only one of
2196 them in implemented---another is checked in yet to be landed
2201 Description: Add lockdep support for inode mutex.
2202 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2203 in the situations where Lustre has to lock more than one inode
2208 Description: Add optional invariants checking support.
2209 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2210 configure switch. This macro is to be used for consistency and
2211 sanity checks that are too expensive to be left in `production'
2216 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2217 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2222 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2223 Details : Use spin_lock_nested() in (the only) situation where more than
2224 one ldlm_lock is locked simultaneously. Also, fix possible
2225 dead-lock in ldlm_lock_change_resource() by enforcing particular
2230 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2231 Details : Instead of specifying each ldlm_lock call-back through separate
2232 parameter, wrap them into struct ldlm_callback_suite.
2236 Description: Kill join_lru obd method and its callers.
2237 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2238 in memory---a requirement implemented through ->o_join_lru() obd
2239 method in HEAD. Remove this method and its users.
2243 Description: Add asynchronous ldlm ENQUEUE completion handler.
2244 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2245 case that stock ldlm_completion_ast() cannot handle as it waits
2246 until lock is granted. Introduce new ldlm_completion_ast_async()
2247 for this. Also comment ldlm_completion_ast().
2251 Description: ldlm_error <-> errno conversion.
2252 Details : Add functions to map (rather arbitrary) between LDLM error codes
2253 and standard errno values. CLIO needs this to prevent LDLM specific
2254 constants from escaping ldlm and osc.
2258 Description: Kill unused ldlm_handle2lock_ns() function.
2259 Details : Kill unused ldlm_handle2lock_ns() function.
2263 Description: Add lu_ref support to ldlm_lock
2264 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2265 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2266 and ldlm_resource. LDLM interface has to be changed, because code that
2267 releases a reference on a lock, has to "know" what reference this is.
2268 In the most frequent case
2270 lock = ldlm_handle2lock(handle);
2272 LDLM_LOCK_PUT(lock);
2274 no changes are required. When any other reference (received _not_ from
2275 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2276 instead of LDLM_LOCK_PUT().
2278 Arguably, changes are pervasive, and interface requires some discipline
2279 for proper use. On the other hand, it was very instrumental in finding
2280 a few leaked lock references.
2284 Description: Add ldlm_lock_addref_try().
2285 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2286 attempts to addref a lock that might be being canceled
2291 Description: Add ldlm_weigh_callback().
2292 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2293 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2294 hard-coded `number of pages' logic.
2298 Description: Add lockdep annotations to llog code.
2299 Details : Use appropriately tagged _nested() locking calls in the places
2300 where llog takes more than one ->lgh_lock lock.
2304 Description: Add loi_kms_set().
2305 Details : Wrap kms updates into a helper function.
2309 Description: Constify instances of struct lsm_operations.
2310 Details : Constify instances of struct lsm_operations.
2314 Description: lu_conf support.
2315 Details : On a server, a file system object is uniquely identified
2316 by a fid, which is sufficient to locate and load all object
2317 state (inode). On a client, on the other hand, more data are
2318 necessary instantiate an object. Change lu_object_find() and
2319 friends to take additional `lu_conf' argument describing object.
2320 Typically this includes layout information.
2324 Description: lu_context fixes.
2325 Details : Introduce new lu_context functions that are needed on the client
2326 side, where some system threads (ptlrpcd) are shared by multiple
2327 modules, and so cannot be stopped during module shutdown.
2331 Description: Add start and stop methods to lu_device_type_operations.
2332 Details : Introduce two new methods in lu_device_type_operations, that are
2333 invoked when first instance of a given type is created and last one
2334 is destroyed respectively. This is need by CLIO.
2338 Description: Add lu_ref support to struct lu_device.
2339 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2340 track leaked references.
2344 Description: Introduce lu_kmem_descr.
2345 Details : lu_kmem_descr and its companion interface allow to create
2346 and destroy a number of kmem caches at once.
2350 Description: Fix lu_object finalization race.
2351 Details : Fix a race between lu_object_find() finding an object and its
2352 concurrent finalization. This race is (most likely) not possible
2353 on the server, but might happen on the client.
2357 Description: Introduce lu_ref interface.
2358 Details : lu_ref is a debugging module allowing to track references to
2359 a given object. It is quite cpu expensive, and has to be
2360 explicitly enabled with --enable-lu_ref. See usage description
2365 Description: Factor lu_site procfs stats into a separate function.
2366 Details : Separate lu_site stats printing code into a separate function
2367 that can be reused on a client.
2371 Description: Constify instances of struct {lu,dt,md}_device_operations.
2372 Details : Constify instances of struct {lu,dt,md}_device_operations.
2376 Description: Introduce struct md_site and move meta-data specific parts of
2377 struct lu_site here.
2378 Details : Move md-specific fields out of struct lu_site into special struct
2379 md_site, so that lu_site can be used on a client.
2383 Description: Kill mdd_lov_destroy().
2384 Details : Remove unused mdd code.
2388 Description: Add st_block checking to multistat.c.
2389 Details : Add st_block checking to multistat.c.
2393 Description: Add lu_ref support to struct obd_device.
2394 Details : Add lu_ref tracking to obd_device.
2398 Description: Kill obd_set_fail_loc().
2399 Details : Remove unused code.
2403 Description: Add special type for ptlrpc_request interpret functions.
2404 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2405 there. Actual usage will be in CLIO.
2409 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2410 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2411 doesn't work with lockdep.
2415 Description: Add rwv.c test program.
2416 Details : New testing program exercising readv(2) and writev(2) (Qian).
2420 Description: Add sendfile.c test program.
2421 Details : New testing program exercising sendfile(2) (Jay).
2425 Description: Ratelimit a message that can be very frequent.
2426 Details : Ratelimit a memory allocation failure message that can
2431 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2432 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2436 Description: Liblustre build fixes.
2437 Details : Liblustre build fixes.
2441 Description: libcfs: add cfs_{need,cond}_resched() interface.
2442 Details : libcfs: add cfs_{need,cond}_resched() definition and
2443 implementations for Linux, NT, and liblustre.
2445 Severity : enhancement
2447 Description: More exported tunables for mballoc
2448 Details : Add support for tunable preallocation window and new tunables for
2449 large/small requests
2453 Description: Detect corruption of block bitmap and checking for preallocations
2454 Details : Checks validity of on-disk block bitmap. Also it does better
2455 checking of number of applied preallocations. When corruption is
2456 found, it turns filesystem readonly to prevent further corruptions.
2460 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2461 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2462 window to 0. If it does hit cache miss, and read-pattern is still
2463 stride-io mode, does not reset the stride window, but also does not
2464 increase the stride window length in this case.
2468 Frequency : only for big-endian servers
2469 Description: Check if system is big-endian while mounting fs with extents feature
2470 Details : Mounting a filesystem with extents feature will fail on big-endian
2471 systems since ext3-based ldiskfs is not supported on big-endian
2472 systems. This can be over-riden with "bigendian_extents" mount option.
2474 Severity : enhancement
2476 Description: The root squash functionality
2477 Details : A security feature, which is to prevent users from being able
2478 to mount lustre on their desktop, run as root, and delete
2479 all of the files in the filesystem. The goal is accomplished by
2480 remapping user id (UID) and group id (GID) of the root user to
2481 a UID and GID specified by the system administartor via Lustre
2482 configuration management server (MGS). The functionality also
2483 allows to specify sets of clients for which the remapping does
2488 Description: Excessive recovery window
2489 Details : With AT enabled, the recovery window can be excessively long (6000+
2490 seconds). To address this problem, we no longer use
2491 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2492 timeout no longer depends on the service time, it is set to
2493 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2494 time via pb_service_time.
2498 Description: Watchdog triggered on MDS failover
2499 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2500 the OSTs know that the MDS "UUID" can be reused for the same export
2501 from a different NID, so we do not need to wait for the export to be
2505 Frequency : rare, only if using MMP with Linux RAID
2507 Description: MMP doesn't work with Linux RAID
2508 Details : While using HA for Lustre servers with Linux RAID, it is possible
2509 that MMP will not detect multiple mounts. To make this work we
2510 need to unplug the device queue in RAID when the MMP block is being
2511 written. Also while reading the MMP block, we should read it from
2512 disk and not the cached one.
2514 Severity : enhancement
2516 Description: open file using fid
2517 Details : A file can be opened using just its fid, like
2518 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2521 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2524 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2525 Details : add proper locking for creation of jbd_4k slab cache
2529 Description: MMP check in ext3_remount() fails without displaying any error
2530 Details : When multiple mount protection fails during remount, proper error
2533 Severity : enhancement
2535 Description: Allow stripe size to be up to 4G-64k
2536 Details : Fix math logic to allow large stripe sizes.
2540 Description: add check for >8TB ldiskfs filesystems
2541 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2542 Don't allow >8TB ldiskfs filesystems to be mounted without
2543 force_over_8tb mount option
2548 Descriptoin: OSS complains "too many blocks" and crashes
2549 Details : In filter_commitrw_write():
2550 1. lquota_chkquota();
2551 2. obdo_from_inode();
2552 3. lquota_pending_commit();
2553 "2." has changed uid and gid in this case, then caused
2554 lquota_chkquota() and lquota_pending_commit() to use
2555 different uid and gid. That is the root of the bug.
2559 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2560 Details : Detect and print out the 'stripe count = -1' case properly.
2564 Description: Out or order replies might be lost on replay
2565 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2566 request with tid smaller then one currently being added, add it
2567 to the start, not end of the list.
2569 --------------------------------------------------------------------------------
2571 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2573 * Support for kernels:
2574 2.6.5-7.283 (SLES 9),
2575 2.6.9-55.EL (RHEL 4),
2576 2.6.16.46-0.14 (SLES 10),
2577 2.6.18.8 vanilla (kernel.org)
2578 * Client support for unpatched kernels:
2579 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2580 2.6.16 - 2.6.22 vanilla (kernel.org)
2581 * Due to recently discovered recovery problems, we do not recommend
2582 using patchless RHEL 4 clients with this or any earlier release.
2583 * Recommended e2fsprogs version: 1.39.cfs8
2584 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2585 * Starting with this release, the ldiskfs backing filesystem required
2586 by Lustre is now in its own package, lustre-ldiskfs. This package
2587 should be installed. It is versioned separately from Lustre and
2588 may be released separately in future.
2592 Description: block reactivating mgc import until all deactivates complete
2593 Details : Fix race when failing back MDT/MGS to itself (testing)
2596 Frequency : at statup only
2598 Description: mds_lov_synchronize race leads to various problems
2599 Details : simultaneous MDT->OST connections at startup can cause the
2600 sync to abort, leaving the OSC in a bad state.
2602 Severity : enhancement
2604 Description: add optional extra BUILD_VERSION info
2605 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2606 to override the lustre version.
2609 Frequency : 2.6.18 servers only
2611 Description: ll_kern_mount() doesn't release the module reference
2612 Details : The ldiskfs module reference count never drops down to 0
2613 because ll_kern_mount() doesn't release the module reference.
2618 Description: server LBUG when using old ost_num_threads parameter
2619 Details : Accept the old ost_num_threads parameter but warn that it
2620 is deprecated, and fix an off-by-one error that caused an LBUG.
2625 Description: Transient SCSI error results in persistent IO issue
2626 Details : iobuf->dr_error is not reinitialized to 0 between two
2630 Frequency : sometimes when underlying device returns I/O errors
2632 Description: OSTs not going read-only during write failures
2633 Details : OSTs are not remounted read-only when the journal commit threads
2634 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2635 instead of the ext3 wrappers.
2639 Description: poor connect scaling with increasing client count
2640 Details : Don't run filter_grant_sanity_check for more than 100 exports
2641 to improve scaling for large numbers of clients.
2644 Frequency : SLES10 only
2646 Description: sanity-quota.sh quotacheck failed: rc = -22
2647 Details : Quotas cannot be enabled on SLES10.
2650 Frequency : liblustre clients only
2652 Description: getdirentries does not give error when run on compute nodes
2653 Details : getdirentries does not fail when the size specified as an argument
2654 is too small to contain at least one entry
2656 Severity : enhancement
2658 Description: Add LNET router traceability for debug purposes
2659 Details : If a checksum failure occurs with a router as part of the
2660 IO path, the NID of the last router that forwarded the bulk data
2661 is printed so it can be identified.
2666 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2667 Details : Due to a race condition, liblustre clients were occasionally
2668 evicted incorrectly.
2670 Severity : enhancement
2672 Description: lfs setstripe use optional parameters instead of postional
2675 Severity : enhancement
2677 Description: Nanosecond timestamp support for ldiskfs
2678 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2679 resolution timestamps. There is not yet support for this at
2680 the Lustre filesystem level.
2683 Frequency : during server recovery
2685 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2686 Details : request with rq_no_resend flag not awake l_wait_event if they get
2690 Frequency : nfs export on patchless client
2692 Description: connectathon hang when test nfs export over patchless client
2693 Details : Disconnected dentry cannot be found with lookup, so we do not need
2694 to unhash it or make it invalid
2697 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2698 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2699 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2704 Description: open req refcounting wrong on reconnect
2705 Details : If reconnect happened between getting open reply from server and
2706 call to mdc_set_replay_data in ll_file_open, we will schedule
2707 replay for unreferenced request that we are about to free.
2708 Subsequent close will crash in variety of ways.
2709 Check that request is still eligible for replay in
2710 mdc_set_replay_data().
2715 Description: disable writes to filesystem when reading health_check file
2716 Details : the default for reading the health_check proc file has changed
2717 to NOT do a journal transaction and write to disk, because this
2718 can cause reads of the /proc file to hang and block HA state
2719 checking on a healthy but otherwise heavily loaded system. It
2720 is possible to return to the previous behaviour during configure
2721 with --enable-health-write.
2723 Severity : enhancement
2725 Description: 64-bit inode version
2726 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2727 to the inode. This will be required for version-based recovery.
2732 Description: MDS fails to start if a duplicate client export is detected
2733 Details : in some rare cases it was possible for a client to connect to
2734 an MDS multiple times. Upon recovery the MDS would detect this
2735 and fail during startup. Handle this more gracefully.
2737 Severity : enhancement
2739 Description: Add -o localflock option to simulate old noflock
2741 Details : This will achieve local-only flock/fcntl locks
2747 Description: log_commit_thread vs filter_destroy race leads to crash
2748 Details : Take import reference before releasing llog record semaphore
2753 Description: Wrong request locking in request set processing
2754 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2755 imp_delayed_list, in this place should be used imp_lock.
2758 Frequency : when reconnection
2760 Description: Grant Leak when osc reconnect to OST
2761 Details : When osc reconnect ost, OST(filter) should check whether it
2762 should grant more space to client by comparing fed_grant and
2763 cl_avail_grant, and return the granted space to client instead
2764 of "new granted" space, because client will call osc_init_grant
2765 to update the client grant space info.
2768 Frequency : when client reconnect to OST
2770 Description: Grant Leak when osc do resend and replay bulk write
2771 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2772 bulk write request, because the grant info while be sync between
2773 OSC and OST when reconnect, and we should ignore the grant info
2774 these of resend/replay write req.
2779 Description: Grant space more than avaiable left space sometimes.
2780 Details : When then OST is about to be full, if two bulk writing from
2781 different clients came to OST. Accord the avaliable space of the
2782 OST, the first req should be permitted, and the second one
2783 should be denied by ENOSPC. But if the seconde arrived before
2784 the first one is commited. The OST might wrongly permit second
2785 writing, which will cause grant space > avaiable space.
2788 Frequency : when client is evicted
2790 Description: Grant might be wrongly erased when osc is evicted by OST
2791 Details : when the import is evicted by server, it will fork another
2792 thread ptlrpc_invalidate_import_thread to invalidate the
2793 import, where the grant will be set to 0. While the original
2794 thread will update the grant it got when connecting. So if
2795 the former happened latter, the grant will be wrongly errased
2796 because of this race.
2801 Description: Checking Stale with correct fid
2802 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2803 to check whether it is stale, because sometimes, we want the
2804 enqueue happened anyway, and op_data.fid2 will not be initialized.
2806 Severity : enhancement
2808 Description: update patchless client
2809 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2812 Frequency : only with 2.4 kernel
2814 Description: random memory corruption
2815 Details : size of struct ll_inode_info is to big for union inode.u and this
2816 can be cause of random memory corruption.
2821 Description: Memory leak in recovery
2822 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2823 It should also check obd_fail before fsfilt_start, otherwise if
2824 fsfilt_start return -EROFS,(failover mds during mds recovery).
2825 then the req will return with repmsg->transno = 0 and rc = EROFS.
2826 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2827 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2828 be freed no matter whether fsfilt_commit success or not.
2831 Frequency : only with huge count clients
2833 Description: Prevents from taking the superblock lock in llap_from_page for
2835 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2836 for llap_from_page call in ll_removepage prevents from taking the
2837 superblock lock for a soon died page.
2842 Description: Not check open intent error before release open handle
2843 Details : in some rare cases, the open intent error is not checked before
2844 release open handle, which may cause
2845 ASSERTION(open_req->rq_transno != 0), because it tries to release
2846 the failed open handle.
2851 Description: Set cat log bitmap only after create log success.
2852 Details : in some rare cases, the cat log bitmap is set too early. and it
2853 should be set only after create log success.
2857 Description: Accessing a block bevice can re-enable I/O when Lustre is
2858 tearing down a device.
2859 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2864 Description: service threads may hog cpus when there are a lot of requests
2865 Details : Insert cond_resched to give other threads a chance to use some CPU
2870 Description: the cat log was not initialized in recovery
2871 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2872 unlink log on mds will not be initialized until mds post
2873 recovery. And also in mds post recovery, the unlink log will
2874 initialization will be done asynchronausly, so there will be race
2875 between add unlink log and unlink log initialization.
2879 Description: brw_stats were being printed incorrectly
2880 Details : brw_stats were being printed as log2 but all of them were not
2881 recorded as log2. Also remove some code duplication arising from
2882 filter_tally_{read,write}.
2886 Frequency : rare, only in recovery.
2887 Description: ASSERTION(req->rq_type != LI_POISON) failed
2888 Details : imp_lock should be held while iterating over imp_sending_list for
2889 prevent destroy request after get timeout in ptlrpc_queue_wait.
2893 Description: replay-single.sh test 52 fails
2894 Details : A lock's skiplist need to be cleanup when it being unlinked
2895 from its resource list.
2899 Description: Short directio read returns full requested size rather than
2901 Details : Direct I/O operations should return actual amount of bytes
2902 transferred rather than requested size.
2904 Severity : enhancement
2906 Description: metadata RPC reduction (e.g. for rm performance)
2907 Details : decrease the amount of synchronous RPC between clients and servers
2908 by canceling conflicing lock before the operation on the client
2909 and packing thier handles into the main operation RPC to server.
2911 Severity : enhancement
2913 Description: Async OSC create to avoid the blocking unnecessarily.
2914 Details : If a OST has no remain object, system will block on the creating
2915 when need to create a new object on this OST. Now, ways use
2916 pre-created objects when available, instead of blocking on an
2917 empty osc while others are not empty. If we must block, we block
2918 for the shortest possible period of time.
2922 Description: improve handling recoverable errors
2923 Details : if request processig with error which can be recoverable on server
2924 request should be resend, otherwise page released from cache and
2927 Severity : enhancement
2929 Description: refine locking for avoid write wrong info into lov_objid file
2930 Details : fix possible races with add new target and write/update data in
2933 --------------------------------------------------------------------------------
2935 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2940 Frequency : on some architectures
2942 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2943 Details : Uninitialized flags sometimes cause configuration commands to
2947 Frequency : patchless clients only
2949 Description: missing __iget() symbol export
2950 Details : The __iget() symbol export is missing. To avoid the need for
2951 this on patchless clients the deathrow inode reaper is turned
2952 off, and we depend on the VM to clean up old inodes. This
2953 dependency was during via the fix for bug 12181.
2955 --------------------------------------------------------------------------------
2957 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2959 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2960 INTEROPERATE with older versions automatically. In many cases a
2961 special upgrade step is needed. Please read the
2962 user documentation before upgrading any part of a live system.
2963 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2964 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2965 * WARNING: Lustre configuration and startup changes are required with
2966 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2970 Severity : enhancement
2972 Description: MountConf
2973 Details : Lustre configuration is now managed via mkfs and mount
2974 commands instead of lmc and lconf. New obd types (MGS, MGC)
2975 are added for dynamic configuration management. See
2976 https://mail.clusterfs.com/wikis/lustre/MountConf for
2979 Severity : enhancement
2981 Description: dynamic OST addition
2982 Details : OSTs can now be added to a live filesystem
2984 Severity : enhancement
2986 Description: startup order invariance
2987 Details : MDTs and OSTs can be started in any order. Clients only
2988 require the MDT to complete startup.
2990 Severity : enhancement
2992 Description: parallel, asynchronous orphan cleanup
2993 Details : orphan cleanup is now performed in separate threads for each
2994 OST, allowing parallel non-blocking operation.
2996 Severity : enhancement
2998 Description: optimized stripe assignment
2999 Details : stripe assignments are now made based on ost space available,
3000 ost previous usage, and OSS previous usage, in order to try
3001 to optimize storage space and networking resources.
3003 Severity : enhancement
3005 Description: Permanently set tunables
3006 Details : All writable /proc/fs/lustre tunables can now be permanently
3007 set on a per-server basis, at mkfs time or on a live system.
3009 Severity : enhancement
3011 Description: Lustre message v2
3012 Details : Add lustre message format v2.
3014 Severity : enhancement
3016 Description: client OST exclusion list
3017 Details : Clients can be started with a list of OSTs that should be
3018 declared "inactive" for known non-responsive OSTs.
3022 Description: SPEC SFS validation failure on NFS v2 over lustre.
3023 Details : Changes the blocksize for regular files to be 2x RPC size,
3024 and not depend on stripe size.
3026 Severity : enhancement
3028 Description: Multiple MD RPCs in flight.
3029 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3030 To avoid overly-overloading MDS, introduce a limit on number of
3031 MDS RPCs in flight for a single client and add /proc controls
3032 to adjust this limit.
3034 Severity : enhancement
3036 Description: client read/write statistics
3037 Details : Add client read/write call usage stats for performance
3038 analysis of user processes.
3039 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3040 file access. extents_stats shows chunk size distribution.
3041 extents_stats_per_process show chunk size distribution per
3044 Severity : enhancement
3046 Description: mds statistics
3047 Details : Add detailed mds operations statistics in
3048 /proc/fs/lustre/mds/*/stats.
3052 Description: Failure of copying files with lustre special EAs.
3053 Details : Client side always return success for setxattr call for lustre
3054 special xattr (currently only "trusted.lov").
3059 Description: Refcount LNET uuids
3060 Details : The global LNET uuid list grew linearly with every startup;
3061 refcount repeated list entries instead of always adding to
3064 Severity : enhancement
3066 Description: Dynamic service threads
3067 Details : Within a small range, start extra service threads
3068 automatically when the request queue builds up.
3071 Frequency : mixed-endian client/server environments
3073 Description: mixed-endian crashes
3074 Details : The new msg_v2 system had some failures in mixed-endian
3077 Severity : enhancement
3079 Description: Easy OST removal
3080 Details : OSTs can be permanently deactivated with e.g. 'lctl
3081 conf_param lustre-OST0001.osc.active=0'
3083 Severity : enhancement
3085 Description: MGS proc entries
3086 Details : Added basic proc entries for the MGS showing what filesystems
3089 Severity : enhancement
3091 Description: provide MGS failover
3092 Details : Added config lock reacquisition after MGS server failover.
3094 Severity : enhancement
3096 Description: add Linux 2.4 support
3097 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3101 Description: a large application tries to do I/O to the same resource and dies
3102 in the middle of it.
3103 Details : Check the req->rq_arrival time after the call to
3104 ost_brw_lock_get(), but before we do anything about
3105 processing it & sending the BULK transfer request. This
3106 should help move old stale pending locks off the queue as
3107 quickly as obd_timeout.
3110 Frequency : when an incorrect nid is specified during startup
3112 Description: ptlrpc connect to non-existant node causes kernel crash
3113 Details : LNET can't be re-entered from an event callback, which
3114 happened when we expire a message after the export has been
3115 cleaned up. Instead, hand the zombie cleanup off to another
3118 Severity : enhancement
3120 Description: plain/inodebits lock performance improvement
3121 Details : Grouping plain/inodebits in granted list by their request modes
3122 and bits policy, thus improving the performance of search through
3126 Frequency : only if OST filesystem is corrupted
3128 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3129 Details : for a short time RPCs with bulk IO are in the replay list,
3130 but replay of bulk IOs is unimplemented. If the OST filesystem
3131 is corrupted due to disk cache incoherency and then replay is
3132 started it is possible to trip an assertion. Avoid putting
3133 committed RPCs into the replay list at all to avoid this issue.
3136 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3138 Description: System hang on startup
3139 Details : This bug allowed the liblustre (e.g. catamount) client to
3140 return to the app before handling all startup RPCs. This
3141 could leave the node unresponsive to lustre network traffic
3142 and manifested as a server ptllnd timeout.
3144 Severity : enhancement
3146 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3147 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3148 boolean to control whether to print peer debug info when a
3149 client's RPC times out.
3152 Frequency : only for kernels with patches from Lustre below 1.4.3
3154 Description: Remove old rdonly API
3155 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3158 Frequency : only for devices with external journals
3160 Description: Set external device read-only also
3161 Details : During a commanded failover stop, we set the disk device
3162 read-only while the server shuts down. We now also set any
3163 external journal device read-only at the same time.
3166 Frequency : when upgrading from 1.4 while trying to change parameters
3168 Description: The wrong (new) MDC name was used when setting parameters for
3169 upgraded MDT's. Also allows changing of OSC (and MDC)
3170 parameters if --writeconf is specified at tunefs upgrade time.
3173 Frequency : when setting specific ost indicies
3175 Description: QOS code breaks on skipped indicies
3176 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3177 created with --index need not be sequential.
3179 Severity : enhancement
3181 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3182 Details : The uninit_groups feature works in conjunction with the kernel
3183 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3184 up the pass1 processing of e2fsck. This is a read-only feature
3185 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3186 mount filesystems that have had this feature enabled.
3188 Severity : enhancement
3190 Description: Improve multi-block allocation algorithm to avoid fragmentation
3191 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3192 allocation locality and avoid filesystem fragmentation.
3194 ------------------------------------------------------------------------------
3196 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3198 * Support for kernels:
3199 2.4.21-47.0.1.EL (RHEL 3)
3200 2.6.5-7.283 (SLES 9)
3201 2.6.9-42.0.10.EL (RHEL 4)
3202 2.6.12.6 vanilla (kernel.org)
3203 2.6.16.27-0.9 (SLES 10)
3204 * Recommended e2fsprogs version: 1.39.cfs5
3206 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3210 Frequency : occasional, depends on client load and configuration
3211 Bugzilla : 12181, 12203
3212 Description: data loss for recently-modified files
3214 Details : In some cases it is possible that recently written or created
3215 files may not be written to disk in a timely manner (this should
3216 normally be within 30s unless client IO load is very high).
3217 The problem appears as zero-length files or files that are a
3218 multiple of 1MB in size after a client crash or client eviction
3219 that are missing data at the end of the file.
3221 This problem is more likely to be hit on clients where files are
3222 repeatedly created and unlinked in the same directory, clients
3223 have a large amount of RAM, have many CPUs, the filesystem has
3224 many OSTs, the clients are rebooted frequently, and/or the files
3225 are not accessed by other nodes after being written.
3227 The presence of the problem can be detected by looking at
3228 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3229 smaller than the second (nr_unused) then dirty files will not
3230 be flushed automatically to disk. "sync; sleep 10" should be
3231 run several times on the node before unmounting it to update
3232 Lustre (this is also safe to run on nodes without this problem).
3234 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3235 that can cause this same problem, so customers using that kernel
3236 also need to update the kernel in addition to Lustre. In order
3237 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3239 It is normal that files written just before a client crash (less
3240 than 30s) may not yet have been flushed to disk, even for local
3244 Frequency : frequent on thin XT3 nodes
3246 Description: UUID collision on thin XT3 Linux nodes
3247 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3248 randomly, since we relied on an insufficiently-seeded PRNG.
3253 Description: OSS hangs after "All ost request buffers busy"
3254 Details : A deadlock between quota and journal operations caused OSS
3255 hangs after printing "All ost request buffers busy."
3258 Frequency : always on liblustre builds
3260 Description: Cleanup compiler warnings on liblustre
3263 Frequency : always on liblustre builds on XT3
3265 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3269 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3270 > 32000 subdirectories
3271 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3272 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3273 subdirectory count crosses 32000. This will aid e2fsck to
3274 correctly handle more than 32000 subdirectories.
3277 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3279 Description: System hang on startup
3280 Details : This bug allowed the liblustre (e.g. catamount) client to
3281 return to the app before handling all startup RPCs. This
3282 could leave the node unresponsive to lustre network traffic
3283 and manifested as a server ptllnd timeout.
3285 Severity : enhancement
3287 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3288 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3289 boolean to control whether to print peer debug info when a
3290 client's RPC times out.
3295 Description: make O_SYNC working on 2.6 kernels
3296 Details : 2.6 kernels use different method for mark pages for write,
3297 so need add a code to lustre for O_SYNC work.
3302 Description: Failure to close file and release space on NFS
3303 Details : Put inode details into lock acquired in ll_intent_file_open.
3304 Use mdc_intent_lock in ll_intent_open to properly
3305 detect all kind of errors unhandled by mdc_enqueue.
3310 Description: proc file read during shutdown sometimes raced obd removal,
3312 Details : Add lock to prevent obd access after proc file removal.
3315 Frequency : Only for files larger than 4GB on 32-bit clients.
3317 Description: improperly doing page alignment of locks
3318 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3319 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3324 Description: RPCs being resent when they shouldn't be
3325 Details : Some RPCs that should not be resent are being resent. This
3326 can cause inconsistencies in the RPC state machine. Do not
3327 resend such requests.
3330 Frequency : rare, only with NFS export
3332 Description: Crash on NFS re-export node
3333 Details : under very unusual load conditions an assertion is hit in
3334 ll_intent_file_open()
3337 Frequency : only if OST filesystem is corrupted
3339 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3340 Details : for a short time RPCs with bulk IO are in the replay list,
3341 but replay of bulk IOs is unimplemented. If the OST filesystem
3342 is corrupted due to disk cache incoherency and then replay is
3343 started it is possible to trip an assertion. Avoid putting
3344 committed RPCs into the replay list at all to avoid this issue.
3349 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3350 Details : Large single O_DIRECT read and write calls can fail to allocate
3351 a sufficiently large buffer to process the request. In case of
3352 allocation failure the allocation is retried with a smaller
3353 buffer and broken into smaller requests.
3355 Severity : enhancement
3357 Description: Add -o localflock option to simulate old noflock behaviour.
3358 Details : This will achieve local-only flock/fcntl locks coherentness.
3363 Description: versioning check is incomplete
3364 Details : Checking the version difference of client vs. server, report
3365 error if the gap is too big.
3370 Description: add support PG_writeback bit
3371 Details : add support for PG_writeback bit for Lustre, for more carefull
3372 work with page cache in 2.6 kernel. This also fix some deadlocks
3373 and remove hack for work O_SYNC with 2.6 kernel.
3375 Severity : enhancement
3377 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3378 Details : The uninit_groups feature works in conjunction with the kernel
3379 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3380 up the pass1 processing of e2fsck. This is a read-only feature
3381 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3382 mount filesystems that have had this feature enabled.
3384 Severity : enhancement
3386 Description: Improve multi-block allocation algorithm to avoid fragmentation
3387 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3388 allocation locality and avoid filesystem fragmentation.
3390 ------------------------------------------------------------------------------
3392 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3394 * Support for kernels:
3395 2.6.9-42.0.3.EL (RHEL 4)
3396 2.6.5-7.276 (SLES 9)
3397 2.4.21-47.0.1.EL (RHEL 3)
3398 2.6.12.6 vanilla (kernel.org)
3399 2.6.16.21-0.8 (SLES10)
3400 * Recommended e2fsprogs version: 1.39.cfs2-0
3402 * The backwards-compatible /proc/sys/portals symlink has been removed
3403 in this release. Before upgrading, please ensure that you change
3404 any configuration scripts or /etc/sysctl.conf files that access
3405 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3406 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3407 in advance of the upgrade on any system running Lustre 1.4.6 or
3408 newer, since /proc/sys/lnet was added in that version.
3409 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3413 Frequency : only when quota is used
3415 Description: avoid scanning export list for quota master
3416 Details : Change the algorithms to avoid scanning export list in order
3417 to improve the efficiency.
3420 Frequency : MDS failover only, very rarely
3422 Description: "went back in time" messages on mds failover
3423 Details : The greatest transno may be lost when the current operation
3424 finishes with an error (transno==0) and the client's last_rcvd
3425 record is over-written. Save the greatest transno in the
3426 mds_last_transno for this case.
3429 Frequency : always for specific kernels and striping counts
3431 Description: client may get "Matching packet too big" without ACL support
3432 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3433 when trying to access files in certain configurations. The
3434 clients should in fact be denied when mounting because they do
3435 not understand ACLs.
3438 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3440 Description: many clients connecting with IO in progress causes connect timeouts
3441 Details : Avoid synchronous journal commits to avoid delays caused by many
3442 clients connecting/disconnecting when bulk IO is in progress.
3443 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3444 OST_IO_PORTAL to avoid delays behind potentially many pending
3448 Frequency : occasionally with multiple writers to a single file
3450 Description: shared writes to file may result in wrong size reported by stat()
3451 Details : Allow growing of kms when extent lock is cancelled
3454 Frequency : always with random mmap IO to multi-striped file
3456 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3457 Details : Only if the hole is at the end of OST object so that kms is too
3458 small. Fix is to increase kms accordingly in ll_nopage.
3461 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3463 Description: writes to a missing object would leak memory on the OST
3464 Details : If there is an inconsistency between the MDS and OST filesystems,
3465 such that the MDS references an object that doesn't exist, writes
3466 to that object will leak memory due to incorrect cleanup in the
3467 error handling path, eventually running out of memory on the OST.
3472 Description: Creating too long symlink causes lustre errors
3473 Details : Check symlink and name lengths before sending requests to MDS.
3476 Frequency : only if flock is enabled (not on by default)
3478 Description: posix locks not released on fd closure on 2.6.9+
3479 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3480 kernels, this caused such locks not to be released on fd close and
3481 then assertions on fs unmount about still used locks.
3484 Frequency : MDS failover only, very rarely
3486 Description: clients may get ASSERTION(granted_lock != NULL)
3487 Details : When request was taking a long time, and a client was resending
3488 a getattr by name lock request. The were multiple lock requests
3489 with the same client lock handle and
3490 mds_getattr_name->fixup_handle_for_resent_request found one of the
3491 lock handles but later failed with ASSERTION(granted_lock != NULL).
3496 Description: handle->h_buffer_credits > 0, assertion failure
3497 Details : h_buffer_credits is zero after truncate, causing assertion
3498 failure. This patch extends the transaction or creates a new
3502 Frequency : NFS re-export or patchless client
3503 Bugzilla : 11179, 10796
3504 Description: Crash on NFS re-export node (__d_move)
3505 Details : We do not want to hash the dentry if we don't have a lock.
3506 But if this dentry is later used in d_move, we'd hit uninitialised
3507 list head d_hash, so we just do this to init d_hash field but
3508 leave dentry unhashed.
3511 Frequency : NFS re-export or patchless client
3513 Description: NFS exports has problem with symbolic link
3514 Details : lustre client didn't properly install dentry when re-exported
3515 to NFS or running patchless client.
3518 Frequency : NFS re-export or patchless client
3520 Description: Various nfs/patchless fixes.
3521 Details : fixes reuse disconected alias for lookup process - this fixes
3522 warning "find_exported_dentry: npd != pd",
3523 fix permission error with open files at nfs.
3524 fix apply umask when do revalidate.
3527 Frequency : occasional
3529 Description: Crash on NFS re-export node
3530 Details : calling clear_page() on the wrong pointer triggered oops in
3531 generic_mapping_read().
3534 Frequency : rarely, using O_DIRECT IO
3536 Description: unaligned directio crashes client with LASSERT
3537 Details : check for unaligned buffers before trying any requests.
3540 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3542 Description: stale data returned from RAID cache
3543 Details : If only a small amount of IO is done to the RAID device before
3544 reading it again it is possible to get stale data from the RAID
3545 cache instead of reading it from disk.
3548 Frequency : always for sles10 kernel
3550 Description: sles10 support
3551 Details : ll_follow_link: compile fixes and using of nd_set_link
3552 under newer kernels.
3555 Frequency : depends on arch, kernel and compiler version, always on sles10
3558 Description: recursive or deep enough symlinks cause stack overflow
3559 Details : getting rid of large stack-allocated variable in
3563 Frequency : depends on hardware
3565 Description: lustre write performance loss in the SLES10 kernel
3566 Details : the performance loss is caused by using of write barriers in the
3567 ext3 code. The SLES10 kernel turns barrier support on by
3568 default. The fix is to undo that change for ldiskfs.
3570 ------------------------------------------------------------------------------
3572 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3574 * Support for kernels:
3575 2.6.9-42.0.3EL (RHEL 4)
3576 2.6.5-7.276 (SLES 9)
3577 2.4.21-47.0.1.EL (RHEL 3)
3578 2.6.12.6 vanilla (kernel.org)
3582 Frequency : quota enabled and large files being deleted
3584 Description: releasing more than 4GB of quota at once hangs OST
3585 Details : If a user deletes more than 4GB of files on a single OST it
3586 will cause the OST to spin in an infinite loop. Release
3587 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3592 Description: statfs data retrieved from /proc may be stale or zero
3593 Details : When reading per-device statfs data from /proc, in the
3594 {kbytes,files}_{total,free,avail} files, it may appear
3595 as zero or be out of date.
3598 Frequency : systems with MD RAID1 external journal devices
3600 Description: lconf's call to blkid is confused by RAID1 journal devices
3601 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3602 of returning all block devices that match the journal UUID.
3605 Frequency : always, for aggregate stripe size over 4GB
3607 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3608 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3609 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3612 Frequency : always if "lfs find" used on a local file/directory
3614 Description: "lfs find" segfaults if used on a local file/directory
3615 Details : The case where a directory component was not specified wasn't
3616 handled correctly. Handle this properly.
3619 Frequency : always on ppc64
3621 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3622 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3623 perform endianness conversion.
3626 Frequency : rarely (truncate to non-zero file size after write under load)
3627 Bugzilla : 10730, 10687
3628 Description: Files padded with zeros to next 4K multiple
3629 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3630 it is possible that files that are truncated to a non-zero size
3631 immediately after being written are filled with zero bytes beyond
3632 the truncated size. No file data is lost.
3634 Severity : enhancement
3636 Description: Allow recovery/failover for liblustre clients.
3637 Details : liblustre clients were unaware of failover configurations until
3640 Severity : enhancement
3642 Description: user file locks should fail when not mounting with flock option
3643 Details : Set up an error-returning stub in ll_file_operations.lock field
3644 to prevent incorrect behaviour when client is mounted without
3645 flock option. Also, set up properly f_op->flock field for
3649 Frequency : always on ia64
3651 Description: "lfs df" loops on printing out MDS statfs information
3652 Details : The obd_ioctl_data was not initialized and in some systems
3653 this caused a failure during the ioctl that did not return
3654 an error. Initialize the struct and return an error on failure.
3657 Frequency : SLES 9 only
3659 Description: Error of copying files with lustre special EAs as root
3660 Details : Client side always return success for setxattr call for lustre
3661 special xattr (currently only "trusted.lov").
3664 Frequency : rarely on clusters with both ia64+i386 clients
3666 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3667 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3668 clients are doing concurrent writes to the same file it is
3669 possible that the ia64 clients may LASSERT because the OST
3670 extent locks are not PAGE_SIZE aligned. Ensure that grown
3671 locks are always aligned on the request boundary.
3674 Frequency : specific use, occasional
3676 Description: Overwriting in use executable truncates on-disk binary image
3677 Details : If one node attempts to overwrite an executable in use by
3678 another node, we now correctly return ETXTBSY instead of
3679 truncating the file.
3681 Severity : enhancement
3683 Description: Async OSC create to avoid the blocking unnecessarily.
3684 Details : If a OST has no remain object, system will block on the creating
3685 when need to create a new object on this OST. Now, ways use
3686 pre-created objects when available, instead of blocking on an
3687 empty osc while others are not empty. If we must block, we block
3688 for the shortest possible period of time.
3693 Description: chmod on Lustre root is propagated to other clients
3694 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3695 invalid by the follow_mount time.
3698 Frequency : liblustre clients only
3700 Description: Race in 'instant cancel' lock handling could lead to such locks
3701 never to be granted in case of SMP MDS
3702 Details : Do not destroy not yet granted but cbpending locks in
3706 Frequency : replay/resend of open
3708 Description: non null lock assetion failure in mds_intent_policy
3709 Details : Trying to replay/resend lockless open requests resulted in
3710 mds_open() returning 0 with no lock. Now it sets a flag if
3711 a lock is going to be returned.
3713 Severity : enhancement
3715 Description: Checksum enhancements
3716 Details : New checksum enhancements allow for resending RPCs that failed
3719 Severity : enhancement
3721 Description: Tunables on number of dirty pages in cacche
3722 Details : Allow to set limit on number of dirty pages cached.
3727 Description: client crash on unmount - lock still has references
3728 Details : In some error handling cases it was possible to leak a lock
3729 reference on a client while accessing a file. This was not
3730 harmful to the client during operation, but would cause the
3731 client to crash when the filesystem is unmounted.
3734 Frequency : specific case, rare
3736 Description: ETXTBSY on mds though file not in use by client
3737 Details : ETXTBSY is no longer incorrectly returned when attempting to
3738 chmod or chown a directory that the user previously tried to
3739 execute or a currently-executing binary.
3742 Frequency : extremely rare except on liblustre-based clients
3744 Description: Lustre space not freed when files are deleted
3745 Details : Clean up open-unlinked files after client eviction. Previously
3746 the unlink was skipped and the files remained as orphans.
3751 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3752 Details : In some cases it was possible to send a blocking callback to
3753 a client doing a glimpse, even though that client didn't get
3754 a lock granted. When the glimpse lock is cancelled on the OST
3755 the freed lock is left on the waiting list and corrupted the list.
3758 Frequency : all core dumps
3760 Description: Broke core dumps to lustre
3761 Details : Negative dentry may be unhashed if parent does not have UPDATE
3762 lock, but some callers, e.g. do_coredump, expect dentry to be
3763 hashed after successful create, hash it in ll_create_it.
3765 ------------------------------------------------------------------------------
3767 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3769 * Support for kernels:
3770 2.6.9-42.0.2.EL (RHEL 4)
3771 2.6.5-7.276 (SLES 9)
3772 2.4.21-40.EL (RHEL 3)
3773 2.6.12.6 vanilla (kernel.org)
3777 Frequency : always on RHEL 3
3779 Description: Number of open files grows over time
3780 Details : The number of open files grows over time, whether or not
3781 Lustre is started. This was due to a filp leak introduced
3782 by one of our kernel patches.
3784 ------------------------------------------------------------------------------
3786 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3788 * Support for kernels:
3789 2.6.9-42.EL (RHEL 4)
3790 2.6.5-7.276 (SLES 9)
3791 2.4.21-40.EL (RHEL 3)
3792 2.6.12.6 vanilla (kernel.org)
3797 Bugzilla : 5719, 9635, 9792, 9684
3798 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3799 Details : If a server is under heavy load and cannot reply to new
3800 connection requests before the client resends the (re)connect,
3801 the connection handling code can behave badly if two service
3802 threads are concurrently handing separate (re)connections from
3803 the same client. Add better locking to the connection handling
3804 code, and ensure that only a single connection will be processed
3805 for a given client UUID, even if the lock is dropped.
3807 Severity : enhancement
3809 Description: add TCP zero-copy support to kernel
3810 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3811 sends if the hardware supports scatter-gather and checksumming.
3812 This allows socklnd to do client-write and server-read more
3813 efficiently and reduce CPU utilization from skbuf copying.
3816 Frequency : only if NFS exporting from client
3818 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3819 Details : If setting attributes on a file created under NFS that had
3820 never been opened it would be possible to oops the client
3821 if the file had no objects.
3824 Frequency : always for liblustre
3826 Description: liblustre client does MDS+OSTs setattr RPC for each write
3827 Details : When doing a write from a liblustre client, the client
3828 incorrectly issued an RPC to the MDS and each OST the file was
3829 striped over in order to update the timestamps. When writing
3830 with small chunks and many clients this could overwhelm the MDS
3831 with RPCs. In all cases it would slow down the write because
3832 these RPCs are unnecessary.
3834 Severity : enhancement
3836 Description: allow number of MDS service threads to be changed at module load
3837 Details : It is now possible to change the number of MDS service threads
3838 running. Adding "options mds mds_num_threads={N}" to the MDS's
3839 /etc/modprobe.conf will set the number of threads for the next
3840 time Lustre is restarted (assuming the "mds" module is also
3841 reloaded at that time). The default number of threads will
3842 stay the same, 32 for most systems.
3847 Description: OST crash if filesystem is unformatted or corrupt
3848 Details : If an OST is started on a device that has never been formatted
3849 or if the filesystem is corrupt and cannot even mount then the
3850 error handling cleanup routines would dereference a NULL pointer.
3855 Description: NULL pointer deref in llap_from_page.
3856 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3857 incorrect) mapping if the page was truncated/reclaimed while it was
3858 searched for. Check for this condition and skip such pages when
3859 doing readahead. Introduce extra check to llap_from_page() to
3860 verify page->mapping->host is non-NULL (so page is not anonymous).
3863 Frequency : Sometimes when using sys_sendfile
3865 Description: "page not covered by a lock" warnings from ll_readpage
3866 Details : sendfile called ll_readpage without right page locks present.
3867 Now we introduced ll_file_sendfile that does necessary locking
3868 around call to generic_file_sendfile() much like we do in
3872 Frequency : with certain MDS communication failures at client mount time
3874 Description: NULL pointer deref after failed client mount
3875 Details : a client connection request may delayed by the network layer
3876 and not be sent until after the PTLRPC layer has timed out the
3877 request. If the client fails the mount immediately it will try
3878 to clean up before the network times out the request. Add a
3879 reference from the request import to the obd device and delay
3880 the cleanup until the network drops the request.
3883 Frequency : occasionally during client (re)connect
3885 Description: assertion failure during client (re)connect
3886 Details : processing a client connection request may be delayed by the
3887 client or server longer than the client connect timeout. This
3888 causes the client to resend the connection request. If the
3889 original connection request is replied in this interval, the
3890 client may trip an assertion failure in ptlrpc_connect_interpret()
3891 which thought it would be the only running connect process.
3894 Frequency : only with obd_echo servers and clients that are rebooted
3896 Description: kernel BUG accessing uninitialized data structure
3897 Details : When running an obd_echo server it did not start the ping_evictor
3898 thread, and when a client was evicted an uninitialized data
3899 structure was accessed. Start the ping_evictor in the RPC
3900 service startup instead of the OBD startup.
3902 Severity : enhancement
3903 Bugzilla : 10193 (patchless)
3904 Description: Remove dependency on various unexported kernel interfaces.
3905 Details : No longer need reparent_to_init, exit_mm, exit_files,
3906 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3909 Frequency : rare (only users of deprecated and unsupported LDAP config)
3911 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3912 Details : LDAP apparently contains 'lustreName' attributes instead of
3913 'name'. A simple remapping of the name is sufficient.
3916 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3918 Description: waiting_locks_callback trips kernel BUG if client is evicted
3919 Details : Running with the dump_on_timeout debug flag turned on makes
3920 it possible that the waiting_locks_callback() can try to dump
3921 the Lustre kernel debug logs from an interrupt handler. Defer
3922 this log dumping to the expired_lock_main() thread.
3924 Severity : enhancement
3926 Description: Support NFS exporting on 2.6 kernels.
3927 Details : Implement non-rawops metadata methods for NFS server to use without
3928 changing NFS server code.
3931 Frequency : very rare (synthetic metadata workload only)
3933 Description: two racing renames might cause an MDS thread to deadlock
3934 Details : Running the "racer" program may cause one MDS thread to rename
3935 a file from being the source of a rename to being the target of
3936 a rename at exactly the same time that another thread is doing
3937 so, and the second thread has already enqueued these locks after
3938 doing a lookup of the target and is trying to relock them in
3939 order. Ensure that we don't try to re-lock the same resource.
3942 Frequency : only very large systems with liblustre clients
3944 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3945 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3946 all OSTs in parallel. This allows the eviction of stale liblustre
3947 clients to proceed much faster than if they were done in series,
3948 and also offers similar improvements for other set_info RPCs.
3953 Description: excessive CPU usage during initial read phase on client
3954 Details : During the initial read phase on a client, it would agressively
3955 retry readahead on the file, consuming too much CPU and impacting
3956 performance (since 1.4.5.8). Improve the readahead algorithm
3957 to avoid this, and also improve some other common cases (read
3958 of small files in particular, where "small" is files smaller than
3959 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3964 Description: MDS crash when receiving packet with unknown intent.
3965 Details : Do not LBUG in unknown intent case, just return -EFAULT
3967 Severity : enhancement
3968 Bugzilla : 9293, 9385
3969 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3970 Details : Do not serialize getattr (non-intent version) and statfs.
3973 Frequency : occasional, when OST network is overloaded/intermittent
3975 Description: client evicted by OST after bulk IO timeout
3976 Details : If a client sends a bulk IO request (read or write) the OST
3977 may evict the client if it is unresposive to its data GET/PUT
3978 request. This is incorrect if the network is overloaded (takes
3979 too long to transfer the RPC data) or dropped the OST GET/PUT
3980 request. There is no need to evict the client at all, since
3981 the pinger and/or lock callbacks will handle this, and the
3982 client can restart the bulk request.
3985 Frequency : Always when mmapping file with no objects
3987 Description: client crashes when mmapping file with no objects
3988 Details : Check that we actually have objects in a file before doing any
3989 operations on objects in ll_vm_open, ll_vm_close and
3995 Description: Request leak when working with deleted CWD
3996 Details : Introduce advanced request refcount tracking for requests
3997 referenced from lustre intent.
3999 Severity : Enhancement
4001 Description: Cache open file handles on client.
4002 Details : MDS now will return special lock along with openhandle, if
4003 requested and client is allowed to hold openhandle, even if unused,
4004 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4005 closing files for every read/write openration.
4007 Severity : Enhancement
4009 Description: Cache open negative dentries on client when possible.
4010 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4011 negative dentries on lock revocation.
4016 Description: Remounting a client read-only wasn't possible with a zconf mount
4017 Details : It wasn't possible to remount a client read-only with llmount.
4019 Severity : enhancement
4020 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4021 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4022 MPICH is included to add Lustre-specific ADIO interfaces.
4023 This is based closely on the UFS ADIO layer and only differs
4024 in file creation, in order to allow the OST striping to be set.
4025 This is user-contributed code and not supported by CFS.
4030 Description: extended inode attributes (immutable, append-only) work improperly
4031 when 2.4 and 2.6 kernels are used on client/server or vice versa
4032 Details : Introduce kernel-independent values for these flags.
4034 Severity : enhancement
4037 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4038 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4039 granularity. Now, Fractional MB granularity is be supported,
4040 this is very useful for low memory system.
4042 Severity : enhancement
4044 Description: Getattr by fid
4045 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4046 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4047 have locks and avoids getattr rpc requests for it on every stat.
4050 Frequency : Always, for filesystems larger than 2TB
4052 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4053 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4054 filesystems larger than 2TB to be mounted with mballoc enabled.
4057 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4059 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4060 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4061 filesystems larger than 2TB is mounted then the kernel will
4062 silently corrupt the start of the filesystem. CONFIG_LBD is
4063 enabled for all CFS-supported kernels, but the possibility of
4064 this happening with a modified kernel config exists.
4066 Severity : enhancement
4068 Description: add client O_DIRECT support for 2.6 kernels
4069 Details : It is now possible to do O_DIRECT reads and writes to files
4070 in the Lustre client mountpoint on 2.6 kernel clients.
4072 Severity : enhancement
4074 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4075 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4076 parallel, not waiting for response from every OST before sending
4077 a rpc to the next OST.
4082 Description: setattr vs write race when updating file timestamps
4083 Details : Client processes that update a file timestamp into the past
4084 right after writing to the file (e.g. tar) it is possible that
4085 the updated file modification time can be reset to the current
4086 time due to a race between processing the setattr and write RPC.
4088 Severity : enhancement
4090 Description: Bring 'lfs find' closer in line with regular Linux find.
4091 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4092 -print0 options and obtains all the needed info through the lustre
4095 Severity : enhancement
4097 Description: support up to 1024 configured devices on one node
4098 Details : change obd_dev array from statically allocated to dynamically
4099 allocated structs as they are first used to reduce memory usage
4104 Description: Flush dirty partially truncated pages during truncate
4105 Details : Immediatelly flush partially truncated pages in filter_setattr,
4106 this way we completely avoid having any pages in page cache on OST
4107 and can retire ugly workarounds during writes to flush such pages.
4112 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4113 Details : protect inode from truncation within vfs_unlink() context
4114 just take a reference before calling vfs_unlink() and release it
4115 when parent's i_sem is free.
4120 Description: last_id value checked outside lock on OST caused LASSERT failure
4121 Details : If there were multiple MDS->OST object precreate requests in
4122 flight, it was possible that the OST's last object id was checked
4123 outside a lock and incorrectly tripped an assertion. Move checks
4124 inside locks, and discard old precreate requests.
4127 Frequency : always, if extents are used on OSTs
4129 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4130 Details : OSTs using the extents format would not zero the high 16 bits of
4131 the index physical block number. This is not a problem for any
4132 OST filesystems smaller than 16TB, and no kernels support ext3
4133 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4134 new/modified files) and can be fixed for existing filesystems
4135 with e2fsprogs-1.39-cfs1.
4140 Description: import connection selection may be incorrect if timer wraps
4141 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4142 import connections to be ignored if the 32-bit jiffies counter
4143 wraps. Use a 64-bit jiffies counter.
4146 Frequency : very large clusters immediately after boot
4148 Description: LNET request buffers exhausted under heavy short-term load
4149 Details : If a large number of client requests are generated on a service
4150 that has previously never seen so many requests it is possible
4151 that the request buffer growth cannot keep up with the spike in
4152 demand. Instead of dropping incoming requests, they are held in
4153 the LND until the RPC service can accept more requests.
4156 Frequency : Sometimes during replay
4158 Description: Assertion failure in ll_local_open after replay.
4159 Details : If replay happened on an open request reply before we were able
4160 to set replay handler, reply will become not swabbed tripping the
4161 assertion in ll_local_open. Now we set the handler right after
4162 recognising of open request
4165 Frequency : very rare
4167 Description: kernel reports "badness in vsnprintf"
4168 Details : Reading from the "recovery_status" /proc file in small chunks
4169 may cause a negative length in lprocfs_obd_rd_recovery_status()
4170 call to vsnprintf() (which is otherwise harmless). Exit early
4171 if there is no more space in the output buffer.
4173 Severity : enhancement
4175 Description: clear OBD RPC statistics by writing to them
4176 Details : It is now possible to clear the OBD RPC statistics by writing
4177 to the "stats" file.
4182 Description: Client mtime is not the same on different clients after utimes
4183 Details : In some cases, the client was using the utimes() syscall on
4184 a file cached on another node. The clients now validate the
4185 ctime from the MDS + OSTs to determine which one is right.
4190 Description: Inability to activate failout mode
4191 Details : lconf script incorrectly assumed that in pythong string's numeric
4192 value is used in comparisons.
4195 Frequency : always with multiple stripes per file
4197 Description: Inefficient object allocation for mutli-stripe files
4198 Details : When selecting which OSTs to stripe files over, for files with
4199 a stripe count that divides evenly into the number of OSTs,
4200 the MDS is always picking the same starting OST for each file.
4201 Return the OST selection heuristic to the original design.
4206 Description: mount failures may take full timeout to return an error
4207 Details : Under some heavy load conditions it is possible that a
4208 failed mount can wait for the full obd_timeout interval,
4209 possibly several minutes, before reporting an error.
4210 Instead return an error as soon as the status is known.
4212 Frequency : quota enabled and large files being deleted
4214 Description: releasing more than 4GB of quota at once hangs OST
4215 Details : If a user deletes more than 4GB of files on a single OST it
4216 will cause the OST to spin in an infinite loop. Release
4217 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4222 Description: statfs data retrieved from /proc may be stale or zero
4223 Details : When reading per-device statfs data from /proc, in the
4224 {kbytes,files}_{total,free,avail} files, it may appear
4225 as zero or be out of date.
4228 Frequency : systems with MD RAID1 external journal devices
4230 Description: lconf's call to blkid is confused by RAID1 journal devices
4231 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4232 of returning all block devices that match the journal UUID.
4235 Frequency : always, for aggregate stripe size over 4GB
4237 Description: assertion fails when trying to use 4GB stripe size
4238 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4239 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4242 Frequency : always on ppc64
4244 Description: the first write on an ext3 filesystem with mballoc got stuck
4245 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4246 perform endianness conversion.
4248 ------------------------------------------------------------------------------
4250 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4252 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4253 INTEROPERATE with older versions automatically. Please read the
4254 user documentation before upgrading any part of a live system.
4255 * WARNING: Lustre networking configuration changes are required with
4256 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4259 * Support for kernels:
4260 2.6.9-22.0.2.EL (RHEL 4)
4261 2.6.5-7.244 (SLES 9)
4262 2.6.12.6 vanilla (kernel.org)
4265 Severity : enhancement
4266 Bugzilla : 7981/8208
4267 Description: Introduced Lustre Networking (LNET)
4268 Details : LNET is new networking infrastructure for Lustre, it includes
4269 a reorganized network configuration mode (see the user
4270 documentation for full details) as well as support for routing
4271 between different network fabrics. Lustre Networking Devices
4272 (LNDs) for the supported network fabrics have also been
4273 created for this new infrastructure.
4275 Severity : enhancement
4276 Description: Introduced Access control lists
4277 Details : clients can set ACLs on files and directories in order to have
4278 more fine-grained permissions than the standard Unix UGO+RWX.
4279 The MDS must be started with the "-o acl" mount option.
4281 Severity : enhancement
4282 Description: Introduced filesystem quotas
4283 Details : Administrators may now establish per-user quotas on the
4286 Severity : enhancement
4288 Description: Configuration change for the XT3
4289 The PTLLND is now used to run Lustre over Portals on the XT3
4290 The configure option(s) --with-cray-portals are no longer used.
4291 Rather --with-portals=<path-to-portals-includes> is used to
4292 enable building on the XT3. In addition to enable XT3 specific
4293 features the option --enable-cray-xt3 must be used.
4298 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4299 Details : A race between allocating a new llcd and re-getting the llcd_lock
4300 allowed another thread to grab newly-allocated llcd.
4302 Severity : enhancement
4304 Description: 2.6 OST async journal commit and locking fix to improve performance
4305 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4306 2.6 kernels are now async as they already were in 2.4 kernels so
4307 that they can commit concurrently with the network bulk transfer.
4308 For block-allocated files the filter allocation semaphore is held
4309 to avoid filesystem fragmentation during allocation. BKL lock
4310 removed for 2.6 xattr operations where it is no longer needed.
4315 Description: lconf incorrectly determined whether two IP networks could talk
4316 Details : In some more complicated routing and multiple-network
4317 configurations, lconf will avoid trying to make a network
4318 connection to a disjoint part of the IP space. It was doing the
4319 math incorrectly for one set of cases.
4324 Description: Fix for potential infinite loop processing records in an llog.
4325 Details : If an llog record is corrupted/zeroed, it is possible to loop
4326 forever in llog_process(). Validate the llog record length
4327 and skip the remainder of the block on error.
4330 Frequency : occasional (liblustre only)
4332 Description: liblustre could not open files whose last component is a symlink
4333 Details : sysio_path_walk() would incorrectly pass the open intent to
4334 intermediate path components.
4337 Frequency : rare (liblustre only with non-standard tuning)
4338 Bugzilla : 7201 (7350)
4339 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4340 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4341 after it has been released, only do this for referenced locks
4343 Severity : enhancement
4345 Description: specifying an (invalid) directory default stripe_size of -1
4346 would reset the directory default striping
4347 Details : stripe_size -1 was used internally to signal directory stripe
4348 removal, now use "all default" to signal dir stripe removal
4349 as a directory striping of "all default" is not useful
4352 Frequency : common for large clusters running liblustre clients
4354 Description: doing an ls when liblustre clients are running is slow
4355 Details : sending a glimpse AST to a liblustre client waits for every AST
4356 to time out, as liblustre clients will not respond. Since they
4357 cannot cache data we refresh the OST lock LVB from disk instead.
4359 Severity : enhancement
4361 Description: doing an ls at the same time as file IO can be slow
4362 Details : enqueue and other "small" requests can be blocked behind many
4363 large IO requests. Create a new OST IO portal for non-IO
4364 requests so they can be processed faster.
4367 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4369 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4370 created file with mode 000
4371 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4372 allow an ftruncate() as a non-root user to a file with mode 000.
4373 Lustre now respects this flag to disable mode checks when
4374 truncating a file owned by the user
4377 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4380 Description: Revoking locks from clients that went dead or catatonic might take
4382 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4383 cancellation of such locks instant on servers without waiting for
4384 any reply from clients. Clients drops these locks when cancel
4385 notification from server is received without replying.
4388 Frequency : liblustre-only, when liblustre client dies or becomes busy
4390 Description: Doing ls on Linux clients can take a long time with active
4392 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4393 granting such locks to it in the first place if possible. Locks
4394 are taken by proxy on the OST during the read or write and
4395 dropped immediately afterward. Add connect flags handling, do
4396 not grant locks to liblustre clients for glimpse ASTs.
4398 Severity : enhancement
4400 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4401 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4402 sequential reads and behaves badly with applications doing random
4403 reads. Improve it by reading ahead at least read region, and
4404 avoiding excessive large RPC for small reads.
4406 Severity : enhancement
4408 Description: Creating more than 1000 files for a single job may cause a load
4409 imbalance on the OSTs if there are also a large number of OSTs.
4410 Details : qos_prep_create() uses an OST index reseed value that is an
4411 even multiple of the number of available OSTs so that if the
4412 reseed happens in the middle of the object allocation it will
4413 still utilize the OSTs as uniformly as possible.
4418 Description: OST or MDS may oops in ping_evictor_main()
4419 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4420 but doesn't restart at beginning of obd_exports_timed list
4423 Severity : enhancement
4425 Description: improve by-nid export eviction on the MDS and OST
4426 Details : allow multiple exports with the same NID to be evicted at one
4427 time without re-searching the exports list.
4430 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4432 Description: MDS may oops in groups_free()
4433 Details : in rare race conditions a newly allocated group_info struct is
4434 freed again, and this can be NULL. The 2.4 compatibility code
4435 for groups_free() checked for a NULL pointer, but 2.6 did not.
4438 Frequency : common for liblustre clients doing little filesystem IO
4439 Bugzilla : 9352, 7313
4440 Description: server may evict liblustre clients accessing contended locks
4441 Details : if a client is granted a lock or receives a completion AST
4442 with a blocking AST already set it would not reply to the AST
4443 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4446 Frequency : lfs setstripe, only systems with more than 160 OSTs
4448 Description: unable to set striping with a starting offset beyond OST 160
4449 Details : llapi_create_file() incorrectly limited the starting stripe
4450 index to the maximum single-file stripe count.
4453 Frequency : LDAP users only
4455 Description: lconf did not handle in-kernel recovery with LDAP properly
4456 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4458 Severity : enhancement
4460 Description: bind OST threads to NUMA nodes to improve performance
4461 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4462 node and do their allocations there to localize memory access
4464 Severity : enhancement
4466 Description: llmount can determine client NID directly from Myrinet (GM)
4467 Details : the client NID code from gmnalnid was moved directly into
4468 llmount, removing the need to use this or specifying the
4469 client NID explicitly when mounting GM clients with zeroconf
4472 Frequency : if client is started with down MDS
4474 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4475 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4476 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4477 ptlrpc_import_delayed_req() because we didn't check if the
4478 request was interrupted, and we also didn't break out of the
4479 event loop if there was no timeout
4484 Description: data loss during non-page-aligned writes to a single file from
4485 both multiple nodes and multiple threads on one node at same time
4486 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4487 inconsistency led to false short-reads, that were cached and later
4488 used by ->prepare_write() to fill in partially written page,
4489 leading to data loss.
4492 Frequency : always, if lconf --abort_recovery used
4494 Description: lconf --abort_recovery fails with 'Operation not supported'
4495 Details : lconf was attempting to abort recovery on the MDT device and not
4498 Severity : enhancement
4500 Description: remove cleanup logs
4501 Details : replace lconf-generated cleanup logs with lustre internal
4502 cleanup routines. Eliminates the need for client-cleanup and
4505 Severity : enhancement
4507 Description: add support for EAs (user and system) on lustre filesystems
4508 Details : it is now possible to store extended attributes in the Lustre
4509 client filesystem, and with the user_xattr mount option it
4510 is possible to allow users to store EAs on their files also
4512 Severity : enhancement
4514 Description: Add possibility (config option) to show minimal available OST free
4516 Details : When compiled with --enable-mindf configure option, statfs(2)
4517 (and so, df) will return least minimal free space available from
4518 all OSTs as amount of free space on FS, instead of summary of
4519 free spaces of all OSTs.
4521 Severity : enhancement
4523 Description: do not expand extent locks acquired on OST-side
4524 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4525 by server: they are not cached anyway.
4528 Frequency : when mmap is used/binaries executed from Lustre
4530 Description: Unmmap pages before throwing them away from read cache.
4531 Details : llap_shrink cache now attempts to unmap pages before discarding
4532 them (if unmapping failed - do not discard). SLES9 kernel has
4533 extra checks that trigger if this unmapping is not done first.
4538 Description: lconf didn't resolve symlinks before checking to see whether a
4539 given mountpoint was already in use
4542 Frequency : when migrating failover services
4543 Bugzilla : 6395, 9514
4544 Description: When migrating a subset of services from a node (e.g. failback
4545 from a failover service node) the remaining services would
4546 time out and evict clients.
4547 Details : lconf --force (implied by --failover) sets the global obd_timeout
4548 to 5 seconds in order to quickly disconnect, but this caused
4549 other RPCs to time out too quickly. Do not change the global
4550 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4552 Severity : enhancement
4553 Frequency : if MDS is started with down OST
4554 Bugzilla : 9439,5706
4555 Description: Allow startup/shutdown of an MDS without depending on the
4556 availability of the OSTs.
4557 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4558 Add appropriate locking and lov-osc refcounts for safe
4559 cleaning. Add osc abort_inflight calls in case the
4560 synchronize never started.
4563 Frequency : occasional (Cray XT3 only)
4565 Description: root not authorized to access files in CRAY_PORTALS environment
4566 Details : The client process capabilities were not honoured on the MDS in
4567 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4568 been accessed by an authorized user then root was able to access
4569 the file on the local client also. The root user capabilities
4570 are now allowed on the MDS, as this environment has secure UID.
4573 Frequency : occasional
4575 Description: ldiskfs "too long searching" message happens too often
4576 Details : A debugging message (otherwise harmless) prints too often on
4577 the OST console. This has been reduced to only happen when
4578 there are fragmentation problems on the filesystem.
4583 Description: Division by zero in statfs when all OSCs are inactive
4584 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4585 lov_statfs divides by value returned by lov_get_stripecnt().
4589 Bugzilla : 9489, 3273
4590 Description: First write from each client to each OST was only 4kB in size,
4591 to initialize client writeback cache, which caused sub-optimal
4592 RPCs and poor layout on disk for the first writen file.
4593 Details : Clients now request an initial cache grant at (re)connect time
4594 and so that they can start streaming writes to the cache right
4595 away and always do full-sized RPCs if there is enough data.
4596 If the OST is rebooted the client also re-establishes its grant
4597 so that client cached writes will be honoured under the grant.
4602 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4603 Details : Now I/O RPCs go to different portal number and (presumably) fast
4604 lock requests (and glimses) and other RPCs get their own service
4605 threads pool that should be able to service those RPCs
4608 Severity : enhancement
4610 Description: Ability to exchange lustre version between client and servers and
4611 issue warnings at client side if client is too old. Also for
4612 liblustre clients there is ability to refuse connection of too old
4614 Details : New 'version' field is added to connect data structure that is
4615 filled with version info. That info is later checked by server and
4619 Frequency : rare, liblustre only.
4620 Bugzilla : 9296, 9581
4621 Description: Two simultaneous writes from liblustre at offset within same page
4622 might proceed at the same time overwriting eachother with stale
4624 Details : I/O lock withing llu_file_prwv was released too early, before data
4625 actually was hitting the wire. Extended lock-holding time until
4626 server acknowledges receiving data.
4629 Frequency : extremely rare. Never observed in practice.
4631 Description: avoid generating lustre_handle cookie of 0.
4632 Details : class_handle_hash() generates handle cookies by incrementing
4633 global counter, and can hit 0 occasionaly (this is unlikely, but
4634 not impossible, because initial value of cookie counter is
4635 selected randonly). Value of 0 is used as a sentinel meaning
4636 "unassigned handle" --- avoid it. Also coalesce two critical
4637 sections in this function into one.
4639 Severity : enhancement
4641 Description: allow liblustre clients to delegate truncate locking to OST
4642 Details : To avoid overhead of locking, liblustre client instructs OST to
4643 take extent lock in ost_punch() on client's behalf. New connection
4644 flag is added to handle backward compatibility.
4646 Severity : enhancement
4647 Bugzilla : 4928, 7341, 9758
4648 Description: allow number of OST service threads to be specified
4649 Details : a module parameter allows the number of OST service threads
4650 to be specified via "options ost ost_num_threads={N}" in the
4651 OSS's /etc/modules.conf or /etc/modprobe.conf.
4655 Bugzilla : 6146, 9635, 9895
4656 Description: servers crash with bad pointer in target_handle_connect()
4657 Details : In rare cases when a client is reconnecting it was possible that
4658 the connection request was the last reference for that export.
4659 We would temporarily drop the export reference and get a new
4660 one, but this may have been the last reference and the export
4661 was just destroyed. Get new reference before dropping old one.
4663 Severity : enhancement
4664 Frequency : if client is started with failover MDS
4666 Description: Allow multiple MDS hostnames in the mount command
4667 Details : Try to read the configuration from all specified MDS
4668 hostnames during a client mount in case the "primary"
4671 Severity : enhancement
4673 Description: Stop sending data to evicted clients as soon as possible.
4674 Details : Check if the client we are about to send or are sending data to
4675 was evicted already. (Check is done every second of waiting,
4676 for which l_wait_event interface was extended to allow checking
4677 of exit condition at specified intervals).
4680 Frequency : rare, normally only when NFS exporting is done from client
4682 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4684 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4687 Frequency : always, if srand() is called before liblustre initialization
4689 Description: Liblustre uses system PRNG disturbing its usage by user application
4690 Details : Introduce internal to lustre fast and high-quality PRNG for
4691 lustre usage and make liblustre and some other places in generic
4692 lustre code to use it.
4694 Severity : enhancement
4695 Bugzilla : 9477, 9557, 9870
4696 Description: Verify that the MDS configuration logs are updated when xml is
4697 Details : Check if the .xml configuration logs are newer than the config
4698 logs stored on the MDS and report an error if this is the case.
4699 Request --write-conf, or allow starting with --old_conf.
4701 Severity : enhancement
4703 Description: Handle symlinks in the path when checking if Lustre is mounted.
4704 Details : Resolve intermediate symlinks when checking if a client has
4705 mounted a filesystem to avoid duplicate client mounts.
4710 Description: lconf can hit an error exception but still return success.
4711 Details : The lconf command catches the Command error exception at the top
4712 level script context and will exit with the associated exit
4713 status, but doesn't ensure that this exit status is non-zero.
4718 Description: failure of ptlrpc thread startup can cause oops
4719 Details : Starting a ptlrpc service thread can fail if there are a large
4720 number of threads or the server memory is very fragmented.
4721 Handle this without oopsing.
4724 Frequency : always, only if liblustre and non-default acceptor port was used
4726 Description: liblustre cannot connect to servers with non-default acceptor port
4727 Details : tcpnal_set_default_params() was not called and was therefore
4728 ignoring the environment varaible TCPNAL_PORT, as well as other
4729 TCPNAL_ environment variables
4734 Description: two objects could be created on the same OST for a single file
4735 Details : If an OST is down, in some cases it was possible to create two
4736 objects on a single OST for a single file. No problems other
4737 than potential performance impact and spurious error messages.
4741 Bugzilla : 5681, 9562
4742 Description: Client may oops in ll_unhash_aliases
4743 Details : Client dcache may become inconsistent in race condition.
4744 In some cases "getcwd" can fail if the current directory is
4750 Description: Inode refcounting problems in NFS export code
4751 Details : link_raw functions used to call d_instantiate without obtaining
4752 extra inode reference first.
4756 Bugzilla : 9942, 9903
4757 Description: Referencing freed requests leading to crash, memleaks with NFS.
4758 Details : We used to require that call to ll_revalidate_it was always
4759 followed by ll_lookup_it. Also with revalidate_special() it is
4760 possible to call ll_revalidate_it() twice for the same dentry
4761 even if first occurence returned success. This fix changes semantic
4762 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4763 reference on a request referred from the intent.
4764 ll_intent_release() then releases such a request.
4767 Frequency : rare, normally benchmark loads only
4769 Description: unlinked inodes were kept in memory on the client
4770 Details : If a client is repeatedly creating and unlinking files it
4771 can accumulate a lot of stale inodes in the inode slab cache.
4772 If there is no other client load running this can cause the
4773 client node to run out of memory. Instead flush old inodes
4774 from client cache that have the same inode number as a new inode.
4777 Frequency : SLES9 2.6.5 kernel and long filenames only
4778 Bugzilla : 9969, 10379
4779 Description: utime reports stale NFS file handle
4780 Details : SLES9 uses out-of-dentry names in some cases, which confused
4781 the lustre dentry revalidation. Change it to always use the
4785 Frequency : rare, unless heavy write-truncate concurrency is continuous
4786 Bugzilla : 4180, 6984, 7171, 9963, 9331
4787 Description: OST becomes very slow and/or deadlocked during object unlink
4788 Details : filter_destroy() was holding onto the parent directory lock
4789 while truncating+unlinking objects. For very large objects this
4790 may block other threads for a long time and slow overall OST
4791 responsiveness. It may also be possible to get a lock ordering
4792 deadlock in this case, or run out of journal credits because of
4793 the combined truncate+unlink. Solution is to do object truncate
4794 first in one transaction without parent lock, and then do the
4795 final unlink in a new transaction with the parent lock. This
4796 reduces the lock hold time dramatically.
4799 Frequency : rare, 2.4 kernels only
4801 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4802 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4803 before unmounting their filesystem in order to avoid deadlock.
4804 The kernel_locked() function in 2.4 kernels only checks whether
4805 the kernel lock is held, not whether it is this process that is
4806 holding it as 2.6 kernels do.
4811 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4812 Details : The client ptlrpc code may be trying to reconnect to a down
4813 server before a previous connection attempt has timed out.
4814 Increase the reconnect interval to be longer than the connection
4815 timeout interval to avoid sending duplicate connections to
4819 Frequency : echo_client brw_test command
4821 Description: fix echo_client to work with OST preallocated code
4822 Details : OST preallocation code (5137) didn't take echo_client IO path
4823 into account: echo_client calls filter methods outside of any
4824 OST thread and, hence, there is no per-thread preallocated
4825 pages and buffers to use. Solution: hijack pga pages for IO. As
4826 a byproduct, this avoids unnecessary data copying.
4830 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4831 Description: Client can oops in mdc_commit_close() after open replay
4832 Details : It was possible for the MDS to return an open request with no
4833 transaction number in mds_finish_transno() if the client was
4834 evicted, but without actually returning an error. Clients
4835 would later try to replay that open and may trip an assertion
4836 Simplify the client close codepath, and always return an error
4837 from the MDS in case the open is not successful.
4840 Frequency : rare, 2.6 OSTs only
4842 Description: OST may deadlock under high load on fragmented files
4843 Details : If there was a heavy load and highly-fragmented OST filesystems
4844 it was possible to have all the OST threads deadlock waiting on
4845 allocation of biovecs, because the biovecs were not released
4846 until the entire RPC IO was completed. Instead, release biovecs
4847 as soon as they are complete to ensure forward IO progress.
4849 Severity : enhancement
4851 Description: Support for specifying external journal device at mount
4852 Details : If an OST or MDS device is formatted with an external journal
4853 device, this device major/minor is stored in the ext3 superblock
4854 and may not be valid for failover. Allow detecting and
4855 specifying the external journal at mount time.
4860 Description: Mounting an MDS with pending unlinked files may cause oops
4861 Details : target_finish_recovery() calls mds_postrecov() which returned
4862 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4863 considers this an error and immediately begins cleaning up the
4864 lov, just after starting the mds_lov process
4866 Severity : enhancement
4868 Description: Implement 'lfs df' to report actual free space on per-OST basis
4869 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4870 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4871 usage of objects; '-h' to report in human readable format.
4873 ------------------------------------------------------------------------------
4875 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4882 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4883 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4884 at startup during orphan unlinking. Instead, skip buddy update
4885 before setup, as it will be regenerated after recovery is complete.
4890 Description: If an OST is inactive, its locks might reference stale inodes.
4891 Details : lov_change_cbdata() must iterate over all namespaces, even if
4892 they are inactive to clear inode references from the lock.
4894 Severity : enhancement
4895 Frequency : occasional, if non-standard max_dirty_mb used
4897 Description: Client will block write RPCs if not enough grant
4898 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4899 then the client will block writes while waiting for another RPC
4900 to complete instead of consuming its dirty limit. With change
4901 we get improved performance when max_dirty_mb is small.
4903 Severity : enhancement
4904 Bugzilla : 3389, 6253
4905 Description: Add support for supplementary groups on the MDS.
4906 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4907 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4908 lookups for user supplementary groups into a cache.
4912 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4913 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4914 for liblustre O_EXCL is enforced by MDS.
4918 Description: readdir never returns NULL in liblustre.
4919 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4920 dentry in struct dirent.
4924 Frequency : liblustre only, depends on application IO pattern
4925 Description: liblustre clients evicted if not contacting servers
4926 Details : Don't put liblustre clients into the ping_evictor list, so
4927 they will not be evicted by the pinger ever.
4929 Severity : enhancement
4931 Description: Add ability to evict clients by NID from MDS.
4932 Details : By echoing "nid:$NID" string into
4933 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4934 $NID would be instantly evicted from this MDS and from all active
4935 OSTs connected to it.
4939 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4940 Details : lookup_it_finish() used to query file size from OSTs that was not
4945 Description: service threads change working directory to that of init
4946 Details : Starting lustre service threads may pin the working directory
4947 of the parent thread, making that filesystem busy. Threads
4948 now change to the working directory of init to avoid this.
4952 Frequency : during shutdown only
4953 Description: shutdown with a failed MDS or OST can cause unmount to hang
4954 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4959 Frequency : 2.6 only
4960 Description: chmod/chown may include an extra supplementary group
4961 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4962 supplementary group and if none is specified this is used.
4965 Bugzilla : 5479 (6816)
4967 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4968 Details : If lookup is in progress on a file that is unlinked we might try
4969 to revalidate the inode and fail in revalidate after lookup is
4970 complete and ll_file_open() enqueues the open again but
4971 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4974 Frequency : always, if lconf --abort_recovery used
4976 Description: lconf --abort_recovery fails with 'Operation not supported'
4977 Details : lconf was attempting to abort recovery on the MDT device and not
4980 ------------------------------------------------------------------------------
4982 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4987 Frequency : rare (only unsupported configurations with a node running as an
4989 Bugzilla : 6514, 5137
4990 Description: Mounting a Lustre file system on a node running as an OST could
4992 Details : OSTs now preallocates memory needed to write out data at
4993 startup, instead of when needed, to avoid having to
4994 allocate memory in possibly low memory situations.
4995 Specifically, if the file system is mounted on on OST,
4996 memory pressure could force it to try to write out data,
4997 which it needed to allocate memory to do. Due to the low
4998 memory, it would be unable to do so and the node would
4999 become unresponsive.
5001 Severity : enhancement
5003 Description: Addition of lconf --service command line option
5004 Details : lconf now accepts a '--service <arg>' option, which is
5005 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5007 Severity : enhancement
5009 Description: Failover mode is now the default for OSTs.
5010 Details : By default, OSTs will now run in failover mode. To return to
5011 the old behaviour, add '--failout' to the lmc line for OSTs.
5013 Severity : enhancement
5015 Description: Health checks are now provided for MDS and OSTs
5016 Details : Additional detailed health check information on MSD and OSTs
5017 is now provided through the procfs health_check value.
5020 Frequency : occasional, depends on IO load
5022 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5023 after numerous create/delete cycles
5024 Details : The ext3 inode allocation policy would not allocate new inodes
5025 very well on the OSTs because there are no new directories
5026 being created. Instead we look for groups with free space if
5027 the parent directories are nearly full.
5032 Description: Network or server problems during mount may cause partially
5033 mounted clients instead of returning an error.
5034 Details : The config llog parsing code may overwrite the error return
5035 code during mount error handling, returning success instead
5041 Description: MDS can fail to allocate large reply buffers
5042 Details : After long uptimes the MDS can fail to allocate large reply
5043 buffers (e.g. zconf client mount config records) due to memory
5044 fragmentation or consumption by the buffer cache. Preallocate
5045 some large reply buffers so that these replies can be sent even
5046 under memory pressure.
5050 Frequency : rare (liblustre)
5051 Description: fsx running with liblustre complained that using truncate() to
5052 extend the file doesn't work. This patch corrects that issue.
5053 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5054 ATTR_SIZE and lsm use in llu_setattr_raw.
5058 Frequency : rare, only 2.6 kernels
5059 Description: Unusual file access patterns on the MDS may result in inode
5060 data being lost in very rare circumstances.
5061 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5062 code in the 2.6 kernel caused the inode and/or EA data not to
5063 be read from disk, causing single-file corruption.
5067 Frequency : rare, only 2.6 filesystems using extents
5068 Description: Heavy concurrent write and delete load may cause data corruption.
5069 Details : It was possible under high-load situations to have an extent
5070 metadata block in the block device cache from a just-unlinked
5071 file overwrite a newly-allocated data block. We now unmap any
5072 metadata buffers that alias just-allocated data blocks.
5076 Frequency : filesystems with default stripe_count larger than 77
5077 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5078 Details : lconf specifies an inode size of 4096 bytes when the default
5079 stripe_count is larger than 77. This conflicts with the default
5080 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5081 this case to avoid pinning too much memory for large EAs.
5083 ------------------------------------------------------------------------------
5085 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5090 Frequency : rare (extremely heavy IO load with hundreds of clients)
5092 Description: Client is evicted, gets IO error writing to file
5093 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5094 caused clients to be evicted for AST timeouts. The fixes in
5095 bug 5192 mean we no longer need to have such short AST timeouts
5096 so ldlm_timeout has been increased.
5099 Frequency : occasional during --force or --failover shutdown under load
5100 Bugzilla : 5949, 4834
5101 Description: Server oops/LBUG if stopped with --force or --failover under load
5102 Details : a collection of import/export refcount and cleanup ordering
5103 issues fixed for safer force cleanup
5106 Frequency : only filesystems larger than 120 OSTs
5107 Bugzilla : 5990, 6223
5108 Description: lfs getstripe would oops on a very large filesystem
5109 Details : lov_getconfig used kfree on vmalloc'd memory
5112 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5113 Bugzilla : 6242, 6243
5114 Description: reading from files that had been truncated to a non-zero size
5115 but never opened returned no data
5116 Details : ll_file_read() reads zeros from no-object files to EOF
5121 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5122 Details : There's little harm in aborting MDS/OSS recovery and letting it
5123 try again, so I removed the LASSERT and return an error instead.
5125 Severity : enhancement
5127 Description: New debugging infrastructure for tracking down data corruption
5128 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5129 (b) cover more of the client-side code path, and (c) try to narrow
5130 down where problems occurred
5134 Bugzilla : 3819, 4364, 4397, 6313
5135 Description: Racing close and eviction MDS could cause assertion in mds_close
5136 Details : It was possible to get multiple mfd references during close and
5137 client eviction, leading to one thread referencing a freed mfd.
5139 Severity: : enhancement
5140 Bugzilla : 3262, 6359
5141 Description: Attempts to reconnect to servers are now more aggressive.
5142 Details : This builds on the enhanced upcall-less recovery that was added
5143 in 1.4.2. When trying to reconnect to servers, clients will
5144 now try each server in the failover group every 10 seconds. By
5145 default, clients would previously try one server every 25 seconds.
5150 Description: After recovery, certain operations trigger a failed
5151 assertion on a client.
5152 Details : Failing over an mds, using lconf -d --failover, while a
5153 client was doing a readdir() call would cause the client to
5154 LBUG after recovery completed and the readdir() was resent.
5156 Severity : enhancement
5158 Description: Default groups are now added by lconf
5159 Details : You can now run lconf --group <servicename> without having to
5160 manually add groups with lmc.
5163 Frequency : occasional
5165 Description: Nodes with an elan id of 0 trigger a failed assertion
5168 Frequency : always when accessing e.g. tty/console device nodes
5170 Description: tty and some other devices nodes cannot be used on lustre
5171 Details : file's private_data field is used by device data and lustre
5172 values in there got lost. New field was added to struct file to
5173 store fs-specific private data.
5176 Frequency : when exporting Lustre via NFS
5178 Description: NFSD failed occasionally when looking up a path component
5179 Details : NFSD is looking up ".." which was broken in ext3 directories
5180 that had grown large enough to become hashed.
5183 Frequency : Clusters with multiple interfaces not on the same subnet
5185 Description: Nodes will repeatedly try to reconnect to an interface which it
5186 cannot reach and report an error to the log.
5187 Details : Extra peer list entries will be created by lconf with some peers
5188 unreachable. lconf now validates the peer before adding it.
5191 Frequency : Only if a default stripe is set on the filesystem root.
5193 Description: Setting a default stripe on the filesystem root prevented the
5194 filesystem from being remounted.
5195 Details : The client was sending extra request flags in the root getattr
5196 request and did not allocate a reply buffer for the dir EA.
5199 Frequency : occasional, higher if lots of files are accessed by one client
5200 Bugzilla : 6159, 6097
5201 Description: Client trips assertion regarding lsm mismatch/magic
5202 Details : While revalidating inodes the VFS looks up inodes with ifind()
5203 and in rare cases can find an inode that is being freed.
5204 The ll_test_inode() code will free the lsm during ifind()
5205 when it finds an existing inode and then the VFS later attaches
5206 this free lsm to a new inode.
5210 Bugzilla : 6422, 7030
5211 Description: MDS deadlock between mkdir and client eviction
5212 Details : Creating a new file via mkdir or mknod (starting a transaction
5213 and getting the ns lock) can deadlock with client eviction
5214 (gets ns lock and trying to finish a synchronous transaction).
5217 Frequency : occasional
5218 Description: While starting a server, the fsfilt_ext3 module could not be
5220 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5221 kernels. Previously, lconf would still use the ext3 name
5222 when trying to load modules. Now, it will correctly use
5223 ext3 on 2.4 and ldiskfs on 2.6.
5225 Severity : enhancement
5226 Description: The default stripe count has been changed to 1
5227 Details : The interpretation of the default stripe count (0, to lfs
5228 or lmc) has been changed to mean striping across a single
5229 OST, rather than all available. For general usage we have
5230 found a stripe count of 1 or 2 works best.
5232 Severity : enhancement
5233 Description: Add support for compiling against Cray portals.
5234 Details : Conditional compiling for some areas that are different
5238 Frequency : occasional
5239 Bugzilla : 6409, 6834
5240 Description: Creating files with an explicit stripe count may lead to
5241 a failed assertion on the MDS
5242 Details : If some OSTs are full or unavailable, creating files may
5243 trigger a failed assertion on the MDS. Now, Lustre will
5244 try to use other servers or return an error to the
5248 Frequency : occasional
5250 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5251 caused lock timeout during readahead (since 1.4.2).
5252 Details : Processes doing readahead might match a lock that hasn't been
5253 granted yet if there are overlapping and conflicting lock
5254 requests. The readahead process waits on ungranted lock
5255 (original lock is CBPENDING), while OST waits for that process
5256 to cancel CBPENDING read lock and eventually evicts client.
5258 Severity : enhancement
5260 Description: Initial enabling of flock support for clients
5261 Details : Implements fcntl advisory locking and file status functions.
5262 This feature is provided as an optional mount flag (default
5263 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5264 locking are implemented yet, and those that are are not guaranteed
5265 to be completely correct in production environments.
5266 mount -t lustre -o [flock|noflock] ...
5269 Frequency : occasional
5271 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5272 assertion in the ext3 JBD (journaling) layer.
5273 Details : The b_committed_data struct is protected by the big kernel lock
5274 in 2.4 kernels, serializing journal_commit_transaction() and
5275 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5276 access to this struct. In 2.6 kernels there is finer grained
5277 locking to improve SMP performance of the JBD layer.
5281 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5286 Description: merge and cleanup kernel patches.
5289 -----------------------------------------------------------------------------
5291 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5293 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5294 versions of Lustre. Please update all servers and clients to
5295 version 1.4.2 or later at the same time. You must also run
5296 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5297 to update the configuration logs.
5299 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5300 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5301 - swab llog records of type '0' so we get proper header size/idx (5861)
5302 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5303 - fix rename of one directory over another leaking an inode (5953)
5304 - avoid SetPageDirty on 2.6 (5981)
5305 - don't re-add just-being-destroyed locks to the waiting list (5653)
5306 - when creating new directories, inherit the parent's custom
5307 striping settings if present parent (3048)
5308 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5309 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5310 - don't hold client locks on temporary worklist from l_lru (5666)
5311 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5312 - automatically evict dead clients (5921)
5313 - Update file size properly in create+truncate+fstat case (6196)
5314 - Do not unhash mountpoint dentries, do not allow removal of
5316 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5317 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5318 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5319 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5320 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5321 - Correctly handle reads of files with no objects (6243)
5322 - lctl recover will also mark a device active if deactivate used (5933)
5324 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5325 - allow --write-conf on an MDS with different nettype than client (5619)
5326 - don't write config llogs to MDS for mounts not from that MDS (5617)
5327 - lconf should create multiple TCP connections from a client (5201)
5328 - init scripts are now turned off by default; run chkconfig --on
5329 lustre and chkconfig --on lustrefs to use them
5330 - upcalls are no longer needed for clients to recover to failover
5332 - add --abort-recovery option to lconf to abort recovery on device
5334 - add support for an arbitrary number of OSTs (3026)
5335 - Quota support protocol changes.
5336 - forward compatibility changes to wire structs (6007)
5337 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5338 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5339 - improved "lustre status" script
5340 - initialize blocksize for non-regular files (6062)
5341 - added --disable-server and --disable-client configure options (5782)
5342 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5343 - Vanilla 2.4.29 support
5344 - increase maximum number of obd devices to 520 (6242)
5345 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5346 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5347 SLES 9 2.6.5 kernels
5348 - limit stripes per file to 160 (the maximum EA size) (6093)
5350 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5353 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5354 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5355 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5356 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5357 - small ext3 extents cleanups and fixes (5733)
5358 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5359 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5360 - avoid needless client->OST connect, fix handle mismatch (5317)
5361 - fix DLM error path that led to out-of-sync client, long delays (5779)
5362 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5363 - fix several locking issues related to i_size (5492,5624,5654,5672)
5364 - don't move pending lock onto export if it is already evicted (5683)
5365 - fix kernel oops when creating .foo in unlinked directory (5548)
5366 - fix deadlock in obdfilter statistics vs. object create (5811)
5367 - use time_{before,after} to avoid timer jiffies wrap (5882)
5368 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5369 - Do not leak request if server was not able to process it (5154)
5370 - If mds_open unable to find parent dir, make that negative lookup(5154)
5371 - don't create new directories with extent-mapping (5909, 5936)
5373 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5374 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5375 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5376 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5377 - support for RHEL4 kernel on i686 (5773)
5378 - provide error messages when incompatible logs are encountered (5898)
5380 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5381 * version 1.4.0.10 (1.4.1 release candidate 1)
5383 - don't keep a lock reference when lock is not granted (4238)
5384 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5385 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5386 - drop import inflight refcount on signal_completed_replay error (5255)
5387 - unlock page after async write error during send (3677)
5388 - handle missing objects in filter_preprw_read properly (5265)
5389 - no transno return for symlink open, don't save no-trasno open (3440)
5390 - don't try to complete elan receive that already failed (4012)
5391 - free RPC server reply state on error (5406)
5392 - clean up thread from ptlrpc_start_thread() on error (5160)
5393 - readahead could read extra page into cache that wasn't ejected (5388)
5394 - prevent races in class_attach/setup/cleanup/detach (5260)
5395 - don't dereference de->d_inode after l_dput of de (5458)
5396 - use "int" for stripe value returned from lock_to_stripe (5544)
5397 - mballoc allocation and error-checking fixes in 2.6 (5504)
5398 - block device patches to fix I/O request sizes in 2.6 (5482)
5399 - look up hostnames for IB nals (5602)
5400 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5401 - don't start multiple acceptors for the same port (5277)
5402 - fix incorrect LASSERT in mds_getattr_name (5635)
5403 - export a proc file for general "ping" checking (5628)
5404 - fix "lfs check" to not block when the MDS is down (5628)
5406 - service request history (4965)
5407 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5408 - create an "evict_client" /proc entry on OSTs, like the MDS has
5409 - fix mount usage message, return errors per mount(8) (5168)
5410 - change grep [] to grep "[]" in tests so they work in more UMLs
5411 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5412 - remove ancient LOV_MAGIC_V0 EA support (5047)
5413 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5414 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5415 - allow readahead allocations to fail when low on memory (5383)
5416 - mmap locking landed again, after considerable improvement (2828)
5417 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5419 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5422 - send OST transaction number in read/write reply to free req (4966)
5423 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5424 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5425 - the watchdog thread now runs as interruptible (5246)
5426 - flock/lockf fixes (but it's still disabled, pending 5135)
5427 - don't use EXT3 constants in llite code (5094)
5428 - memory shortage at startup could cause assertion (5176)
5430 - reorganization of lov code
5431 - single portals codebase
5433 - add extents/mballoc support (5025)
5434 - direct I/O reads in the obdfilter (4048)
5435 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5437 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5440 - send OST transaction number in read/write reply to free req (4966)
5441 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5442 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5443 - the watchdog thread now runs as interruptible (5246)
5444 - handle missing objects in filter_preprw_read properly (5265)
5445 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5446 - drop import inflight refcount on signal_completed_replay error (5255)
5447 - unlock page after async write error during send (3677)
5448 - return original error code on reconstructed replies (3761)
5449 - no transno return for symlink open, don't save no-trasno open (3440)
5451 - add pid to ldlm debugging output (4922)
5452 - bump the watchdog timeouts -- we can't handle 30sec yet
5453 - extra debugging for orphan dentry/inode bug (5259)
5455 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5458 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5459 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5460 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5461 - fix NULL dereference in /proc/sys/portals/routes (4827)
5462 - allow failed mdc_close() operations to be interrupted (4561)
5463 - stop precreate on OST before MDS would time out on it (4778)
5464 - don't send partial-page writes before EOF from client (4410)
5465 - discard client grant for sub-page writes on large-page clients (4520)
5466 - don't free dentries not owned by NFS code, check generation (4806)
5467 - fix lsm leak if mds_create_objects() fails (4801)
5468 - limit debug_daemon file size, always print CERROR messages (4789)
5469 - use transno after validating reply (3892)
5470 - process timed out requests if import state changes (3754)
5471 - update mtime on OST during writes, return in glimpse (4829)
5472 - add mkfsoptions to LDAP (4679)
5473 - use ->max_readahead method instead of zapping global ra (5039)
5474 - don't interrupt __l_wait_event() during strace
5476 - add software watchdogs to catch hung threads quickly (4941)
5477 - make lustrefs init script start after nfs is mounted
5478 - fix CWARN/ERROR duplication (4930)
5479 - return async write errors to application if possible (2248)
5480 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5481 - print NAL number in %x format (4645)
5482 - update barely-supported suse-2.4.21-171 series (4842)
5483 - support for sles 9 %post scripts
5484 - support for building 2.6 kernel-source packages
5485 - support for sles km_* packages
5487 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5490 - ignore -ENOENT errors in osc_destroy (3639)
5491 - notify osc create thread that OSC is being cleaned up (4600)
5492 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5493 - reconstruct ost_handle() like mds_handle() (4657)
5494 - create a new thread to do import eviction to avoid deadlock (3969)
5495 - let lconf resolve symlinked-to devices (4629)
5496 - don't unlink "objects" from directory with default EA (4554)
5497 - hold socknal file ref over connect in case target is down (4394)
5498 - allow more than 32000 subdirectories in a single directory (3244)
5499 - fix blocks count for O_DIRECT writes (3751)
5500 - OST returns ENOSPC from object create when no space left (4539)
5501 - don't send truncate RPC if file size isn't changing (4410)
5502 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5503 - bind to privileged port in socknal and tcpnal (3689)
5505 - rate limit CERROR/CWARN console message to avoid overload (4519)
5506 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5507 - basic mmap support (3918)
5508 - kernel patch series update from b1_4 (4711)
5510 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5513 - avoid crash during MDS cleanup with OST shut down (2775)
5514 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5515 - don't use bad inodes on the MDS (3744)
5516 - dynamic object preallocation to improve recovery speed (4236)
5517 - don't hold spinlock over lock dumping or change debug flags (4401)
5518 - don't zero obd_dev when it is force cleaned (3651)
5519 - print grants to console if they go negative (4431)
5520 - "lctl deactivate" will stop automatic recovery attempts (3406)
5521 - look for existing locks in ldlm_handle_enqueue() (3764)
5522 - don't resolve lock handle twice in recovery avoiding race (4401)
5523 - revalidate should check working dir is a directory (4134)
5525 - don't always mark "slow" obdfilter messages as errors (4418)
5527 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5530 - don't close LustreDB during write_conf until it is done (3860)
5531 - fix typo in lconf for_each_profile (3821)
5532 - allow dumping logs from multiple threads at one time (3820)
5533 - don't allow multiple threads in OSC recovery (3812)
5534 - fix debug_size parameters (3864)
5535 - fix mds_postrecov to initialize import for llog ctxt (3121)
5536 - replace config semaphore with spinlock (3306)
5537 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5538 - don't allow enqueue to complete on a destroyed export (3822)
5539 - down write_lock before checking llog header bitmap (3825)
5540 - recover from lock replay timeout (3764)
5541 - up llog sem before sending rpc (3652)
5542 - reduce ns lock hold times when setting kms (3267)
5543 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5544 - fix NULL deref and obd_dev leak on setup error (3312)
5545 - replace some LBUG about llog ops with error handling (3841)
5546 - don't match INVALID dentries from d_lookup and spin (3784)
5547 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5548 - fix invalid assertion in ptlrpc_set_wait (3880)
5550 - add libwrap support for the TCP acceptor (3996)
5551 - add /proc/sys/portals/routes for non-root route listing (3994)
5552 - allow setting MDS UUID in .xml (2580)
5553 - print the stack of a process that LBUGs (4228)
5555 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5558 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5559 - make sure to unset replay flag from failed open requests (3440)
5560 - if default stripe count is 0, use OST count for inode size (3636)
5561 - update parent mtime/ctime on client for create/unlink (2611)
5562 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5563 - free recovery state on server during a forced cleanup (3571)
5564 - unregister_reply for resent reqs (3063)
5565 - loop back devices mounting and status check on 2.6 (3563)
5566 - fix resource-creation race that can provoke i_size == 0 (3513)
5567 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5568 - more debugging for page-accounting assertion (3746)
5569 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5570 - avoid deadlock after precreation failure (3758)
5571 - fix race and lock order deadlock in orphan handling (3450, 3750)
5572 - add validity checks when grabbing inodes from l_ast_data (3599)
5574 - add /proc/.../recovery_status to obdfilter (3428)
5575 - lightweight CDEBUG infrastructure, debug daemon (3668)
5576 - change default OSC RPC parameters to be better on small clusters
5577 - turn off OST read cache for files smaller than 32MB
5578 - install man pages and include them in rpms (3100)
5579 - add new init script for (un)mounting lustre filesystems (2593)
5580 - run chkconfig in %post for init scripts (3701)
5581 - drop scimac NAL (unmaintained)
5583 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5586 - clean kiobufs before and after use (3485)
5587 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5588 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5589 - change init script to fail more clearly if not run as root (1528)
5590 - allow clients to reconnect during replay (1742)
5591 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5592 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5593 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5594 - fix cond_resched (3554)
5595 - only evict unfinished clients after recovery (3515)
5596 - allow bulk resend, prevent data loss (3570)
5597 - dynamic ptlrpc request buffer allocation (2102)
5598 - don't allow unlinking open directory if it isn't empty (2904)
5599 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5600 - remove extraneous obd dereference causing LASSERT failure (3334)
5601 - don't use get_cycles() when creating temp. files on the mds (3156)
5602 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5603 - handle EEXIST for set-stripe, set proper directory name (3336)
5605 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5606 - fix ksocknal_fmb_callback() error messages (2918)
5608 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5611 - don't copy lvb into (possibly NULL) reply on error (2983)
5612 - don't deref dentry after dput, don't free lvb on error (2922)
5613 - use the kms to determine writeback rpc length (2947)
5614 - increment oti_logcookies when osc is inactive (2948)
5615 - update client's i_blocks count via lvb messages (2543)
5616 - handle intent open/close of special files properly (1557)
5617 - mount MDS with errors=remount-ro, like obdfilter (2009)
5618 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5619 - don't use cancelling-locks' kms values (2947)
5620 - use highest lock extent for kms, not last one (2925)
5621 - don't dereference ERR_PTR() dentry in error handling path (3107)
5622 - fix thread race in portals_debug_dumplog() (3122)
5623 - create lprocfs device entries at setup instead of at attach (1519)
5624 - common AST error handler, don't evict client on completion race (3145)
5625 - zero nameidata in detach_mnt in 2.6 (3118)
5626 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5627 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5628 - initialize RPC timeout timer earlier for 2.6 (3219)
5629 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5630 - print nal/nid for unknown nid (3258)
5631 - additional checks for oscc recovery before doing precreate (3284)
5632 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5633 - don't crash in mdc_close for bad permissions on open (3285)
5634 - zero i_rdev for non-device files (3147)
5635 - clear page->private before handing to FS, better assertion (3119)
5636 - tune the read pipeline (3236)
5637 - fix incorrect decref of invalidated dentry (2350)
5638 - provide read-ahead stats and refine rpc in flight stats (3328)
5639 - don't hold journal transaction open across create RPC (3313)
5640 - update atime on MDS at close time (3265)
5641 - close LDAP connection when recovering to avoid server load (3315)
5642 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5643 - don't leak open file on MDS after open resend (3325)
5644 - serialize filter_precreate and filter_destroy_precreated (3329)
5645 - loop device shouldn't call sync_dev() for nul device (3092)
5646 - clear page cache after eviction (2766)
5647 - resynchronize MDS->OST in background (2824)
5648 - refuse to mount the same filesystem twice on same mountpoint (3394)
5649 - allow llmount to create routes for mounting behind routers (3320)
5650 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5651 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5652 - fix rare problem with rename on htree directories (3417)
5654 - allow default OST striping configuration per directory (1414)
5655 - fix compilation for qswnal for 2.6 kernels (3125)
5656 - increase maximum number of MDS request buffers for large systems
5657 - change liblustreapi to be useful for external progs like lfsck (3098)
5658 - increase local configuration timeout for slow disks (3353)
5659 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5661 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5664 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5665 - don't overwrite extent policy data in reply if lock was blocked (2901)
5666 - drop filter export grants atomically with removal from device (2663)
5667 - del obd_self_export from work_list in class_disconnect_exports (2908)
5668 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5669 - swab reply message in mdc_close, other PPC fixes (2464)
5670 - fix destroying of named logs (2325)
5671 - overwrite old logs when running lconf --write_conf (2264)
5672 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5673 - fix race in target_handle_connect (2898)
5674 - mds_reint_create() should take same inode create lock (2926)
5675 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5676 - don't close files for self_export to avoid uninitialized obd (2936)
5677 - allow MDS with the same name as client node (2939)
5678 - hold dentry reference for closed log files for unlink (2325)
5679 - reserve space for all logs during transactions (2059)
5680 - don't evict page beyond end of stripe extent (2925)
5681 - don't oops on a deleted current working directory (2399)
5682 - handle hard links to targets without a parent properly (2517)
5683 - don't dereference NULL lock when racing during eviction (2867)
5684 - don't grow lock extents when lots of conflicting locks (2919)
5686 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5689 - account for cache space usage on clients to avoid data loss (974)
5690 - lfsck support in lustre kernel code (2349)
5691 - reduce journal credits needed for BRW writes (2370)
5692 - orphan handling to avoid losing space on client/server crashes
5693 - ptlrpcd can be blocked, stopping ALL progress (2477)
5694 - use lock value blocks to assist in proper KMS, faster stat (1021)
5695 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5696 - recovery for initial connections (2355)
5697 - fixes for mds_cleanup_orphans (1934)
5698 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5699 - block all file creations until orphan recovery completes (1901)
5700 - client remove rq_connection from request struct (2423)
5701 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5702 - recovery timer race (2670)
5703 - mdc_close recovey bug (2532)
5704 - ptlrpc cleanup bug (2710)
5705 - mds timeout on local locks (2588)
5706 - namespace lock held during RPCs (2431)
5707 - handle interrupted sync write properly (2503)
5708 - don't try to handle a message that hasn't been replied to (2699)
5709 - client assert failure during cleanup after abort recovery (2701)
5710 - leak mdc device after failed mount (2712)
5711 - ptlrpc_check_set allows timedout requests to complete (2714)
5712 - wait for inflight reqs when ptlrpcd finishes (2710)
5713 - make sure unregistered services are removed from the srv_list
5714 - reset bulk XID's when resending them (caught by 1138 test)
5715 - unregister_bulk after timeout
5716 - fix lconf error (2694)
5717 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5718 - readahead locks pages, leaves pending causing memory pressure (2673)
5719 - increase OST request buffers to 4096 on large machines (2729)
5720 - fix up permission of existing directories in simple_mkdir (2661)
5721 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5722 - don't assign transno to errored transactions (2742)
5723 - don't delete objects on OST if given a bogus objid from MDS (2751)
5724 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5725 - if rq_no_resend, then timeout request after recovery (2432)
5726 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5727 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5728 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5729 - MDS recovery shouldn't delete objects that it has given out (2730)
5730 - if enqueue arrives after completion, don't clobber LVB (2819)
5731 - don't unlock pages twice when trigger_group_io returns error (2814)
5732 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5733 - don't write pages to disk if there was an error (1450)
5734 - don't ping imports that have recovery disabled (2676)
5735 - take buffered bytes into account when balancing socknal conn (2817)
5736 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5737 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5738 - remove little-endian swabbing of llog records (1987)
5739 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5740 - retry reposting request buffers if they fail (1191)
5741 - grow extent at grant time to avoid granting a revoked lock (2809)
5742 - lock revoke doesn't evict page if covered by a second lock (2765)
5743 - disable VM readahead to avoid reading outside lock extents (2805)
5745 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5746 - updated kernel patches for hp-2.4.20 kernel (2681)
5748 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5751 - fix truncated write corruption (2366)
5752 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5754 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5755 - ensure len(uuid) < 37 in lmc (1171)
5756 - fix ia64 OOPS in llog_test (2255)
5757 - zero end of page at obdfilter for partial page writes (2648)
5758 - don't leave stale dentries around after renames (bug 2428)
5759 - fix timeouts when evicting a client with a single lock held (2642)
5760 - set deadline for the initial HELLO message to drain (2634)
5761 - print out dotted-quad IP addresses in the socknal (2302)
5763 - additional debugging for MDS client eviction problem (2443)
5764 - fix mkfsoptions support for osts (2603, 2604)
5766 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5769 - add series for the vanilla 2.6.0 kernel
5770 - add series for the vanilla 2.4.24 kernel
5771 - add series for a cray x86/64 UL kernel drop
5772 - fix xattr patches for the vanilla 2.4.19 series
5774 - generate true UUIDs in lmc (1171)
5775 - have portals stack dumping break in UML (2466)
5776 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5777 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5778 - wake callback waiting threads on client eviction (2460)
5779 - Add --ptldebug and --subsystem to lmc (1719)
5780 - update assertion to allow safe interrupt allocation
5781 - set rq_no_resend for cancel requests (2432)
5782 - recalculate ptlrpcd timeout after resend (2494)
5783 - call vfs_rmdir when removing pending directories (2368)
5784 - fix renaming a file to itself (2429)
5785 - lmc creates a default one-stripe lov (2454)
5786 - expand procfs space to handle large clusters (2326)
5787 - increase UML stack to avoid overflow
5788 - update lconf's list of debug and subsystem masks
5789 - fix lfs find --obd (2510)
5790 - /proc tunable for disabling filter read caching (2591)
5791 - stop rpm packages from altering slapd.conf (2301)
5792 - disable nagle in the socknal under 0conf (2578)
5793 - choose mds inode size based on stripe count (2572)
5794 - fix kernel-source rpm problems (2516)
5796 - add --disable-doc to avoid pdf generation (2421)
5797 - update documentation, tests, type-os, comments
5798 - avoid format warnings on ia64
5799 - remove the TOE NAL
5800 - tiny code cleanups by removing unused fields
5802 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5805 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5806 - remove the most blatant lies from BUILDING file (2371)
5807 - change default debug level to reasonable production setting
5808 - reduce client side cache size to reduce cache flush time
5809 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5810 - make TCP zerocopy and pinger support enabled by default (2476)
5811 - sync writes completed after process exits caused crashes (2319)
5812 - maintain correct mount count on the MDS (2356)
5813 - backout 1557, because 2316 wasn't really fixed
5814 - better file I/O statistics gathering in /proc
5815 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5816 - another kernel patch to fix zero-copy TCP function export
5817 - don't take duplicate lock when processing re-sent getattr (2420)
5818 - lctl uses obd_self_export instead of creating new conn (2353)
5819 - MDS/OST recovery case which requires object creation asserted (2425)
5820 - move lfs from /usr/sbin to /usr/bin in packages
5821 - fix race between mds_client_add and mds_client_free (2417)
5822 - use kmalloc instead of slabs in portals (2430)
5823 - don't create duplicate records when a failover MDS is present (2442)
5824 - remove unnecessary mount age check (2332)
5825 - don't remove directory inodes from locks prematurely (2451)
5826 - don't break if MDS service name is the same as hostname (2103)
5827 - fix races in client write RPC generation when cache full (2482)
5829 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5832 - remove now-unused request->rq_obd (278)
5833 - if an allocation fails, print out how much memory we've used (1933)
5834 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5835 - add the "configurable stack size" patch to most series files (1256)
5836 - ability to write large log records, for 100+ OST configs (2306)
5837 - fix NULL deref when filter_prep fails (2314)
5838 - fix operator precedence error in filter_sync
5839 - dynamic allocation of socknal TX descriptors (2315)
5840 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5841 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5842 - ensure that CWARN messages in Portals always get to the syslog
5843 - __init/__exit are not for prototype decls (ldlm_init/exit)
5844 - x86-64 compile warning fixes
5845 - fix gateway LMC keyword conflict (2318)
5846 - fix MDS lock inversions in getattr/reint paths (1844)
5847 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5848 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5849 - DLM race condition prevented some lock evictions (2328)
5850 - ENOMEM detection and retry on socknal sends (2230)
5851 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5852 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5853 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5854 - better error messages when a client is rejected during recovery (1505)
5855 - avoid cancelling locks which were never granted, after failure (2330)
5856 - fix i_sem/journal inversion in mds_client_add (2333)
5857 - fix truncate/getattr lock cycle deadlock (2334)
5858 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5859 - fix two rare exit paths which could leak an l_lock() ref (2321)
5860 - fencepost error in MDS/OST orphan recovery (2226)
5861 - make log record alignment 8 bytes (1988)
5862 - lstripe now fails when requested offset > ost_count (2237)
5863 - ensure that all kernel series have a complete list.h (1607)
5864 - fix crashes in special-file operations (2316)
5865 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5867 - allow configurable automake binary, for testing new versions
5868 - small update to the lfs documentation
5870 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5872 * fix negative export reference count in fsfilt_sync (2312)
5874 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5875 * release candidate 0.9.1
5877 - orphans are moved into the PENDING directory for possible recovery
5878 - replayed opens now open by fid for orphan/rename safety (1042)
5879 - last close of an orphan inode generates a transno (683)
5880 - chdir() and mount() now pin the directory entry (1020)
5881 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5882 - discard very old requests without processing them (1502)
5883 - remove some common, well-understood CERRORs (1505)
5884 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5885 - clear "grant" flags in OST replies until OST grant code lands (1644)
5886 - fix read performance by not clobbering i_blksize on client (1598)
5887 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5888 - make LRU size a /proc tunable, clears locks when reduced (707)
5889 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5890 - server locks take references on exports now (1558)
5891 - build fixes for 2.4.20-rh trees (1663)
5892 - return an error from lov_create if all OSCs are inactive (1751)
5893 - fix import levels when a reconnect happens without a timeout (1597)
5894 - exit early from mds_open if we get a lookup error (1749)
5895 - partial page read at EOF wouldn't wait for disk before sending (1642)
5896 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5897 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5898 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5899 - write/truncate lock inversion (1639)
5900 - Don't auto-load obdclass, portals modules during cleanup (1495)
5901 - fix timestamps from jumping to "now" (1763)
5902 - extra journal assertions (1648)
5903 - add an extra multiunlink test (1771)
5904 - fix read_record/write_record API (1776)
5905 - fix leak of offset_extent, possible incorrect i_size later (1772)
5906 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5907 - Debugging for the kqswnal_get_idle_tx problems (1820)
5908 - Allow recovery to be attempted multiple times (1536)
5909 - Write out MDS last_rcvd file after it is first created (1600)
5910 - Fix tx_descriptor leak in failed transmit situations (1827)
5911 - ext3 journaling fixes for assertion failure after IO error (1871)
5912 - class_export_put() on freed export after completion AST error (1896)
5913 - Fix revalidate looping in VFS (1322)
5914 - Don't access a freed export during MDS_REINT timeout (1521)
5915 - Add open-unlink recovery support on the MDS (1673,1764)
5916 - Return an error if no MDS data was read from last_rcvd (1946)
5917 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5918 - Don't setup a disk device that doesn't match exported UUID (317)
5919 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5920 - avoid committing NULL handle in force close
5921 - local.sh is now a one-stripe LOV configuration
5922 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5923 - fix ext3 htree duplicate directory entry corruption (1516)
5924 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5925 - update to new LOV EA format (2097)
5926 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5928 2003-06-15 Phil Schwan <phil@clusterfs.com>
5931 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5932 - per-import/export recovery handling (958, 931, 959)
5933 - multiple last-rcvd slots, for serving multiple FSes (949)
5934 - connections are again shared between multiple imp/exports (963, 964)
5935 - "umount -f" would hang if any requests needed to be sent (393, 978)
5936 - avoid pinning large req buffer by copying for queued messages (989)
5937 - add "uuid" to "lctl device" command to help upcalls (991)
5938 - "open" RPCs with transnos would confuse recovery counters (1037)
5939 - do proper endian conversion of all wire messages (288, 340, 891)
5940 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5941 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5942 - fix ptlrpc_connection leak in target_handle_connect (1174)
5943 - fix import refcounting bug in OST and MDS cleanup (1134)
5944 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5945 - fix very unlikely obd_types race condition (501)
5946 - remove osc_open hack for echo_client (1187)
5947 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5948 - a failure in read_inode2 leads to deadlock (1139)
5949 - cancel ack-locks as soon as transaction is committed (1072)
5950 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5951 - make sure to commitrw after any preprw to avoid deadlock (1162)
5952 - failing to execute a file in a lustre FS would lock inode (1203)
5953 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5954 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5955 - obd_uuid2tgt would walk off the end of the list (1255)
5956 - on IA64 the osc would give portals incorrect bulk size (1258)
5957 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5958 - fix lock inversion caused by new llite matching code (1282)
5959 - limit the number of dirty pages on a client to 10MB (1286)
5960 - timed out locks were not being corrected cancelled (1289)
5961 - fix O_DIRECT above 4GB on IA-32 (1292)
5962 * major user-visible changes
5963 - fail out/fail over policy now controlled by the upcall (993)
5965 - add OBD_PING to check server availability and failure (954)
5966 - lustre messages are now sent in sending host order (288, 340, 891)
5967 - add eadatalen to MDS getattr reply (340)
5968 - OST read replies may contain second buffer, with per-page status (593)
5970 2003-03-11 Phil Schwan <phil@clusterfs.com>
5973 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5974 - failed open()s could cause deadlock; fixed (867, 869)
5975 - stop cancelling OST locks when files are closed (481)
5976 - overlapping XID spaces caused network corruption (851, 853)
5977 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5978 - setattr_raw added, to do single-RPC, server-side setattrs
5979 - lmc/lconf syntax change for OST UUIDs
5980 - fix crashy race condition between ptlrpc_free_req and osc_close
5981 - don't use request in mdc_enqueue if we hit a timeout (889)
5982 - don't set the inode i_size for regular files from the MDS (896)
5983 - handle out of order completion AST (842)
5984 - don't LBUG if a lock request times out after receiving AST (913)
5985 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5986 - if a bad lock AST arrives, send an error instead of dropping entirely
5987 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5988 - fix leak in bulk IO when only partially completed (899, 900, 926)
5989 - fix O_DIRECT for ia64 (55)
5990 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5991 - C-z after timeout could hang a process forever; fixed (977)
5993 - client-side I/O cache (678, 924, 929, 941, 970)
5995 - READPAGE and SETATTRs which don't take server-side locks get
5998 2003-02-11 Phil Schwan <phil@clusterfs.com>
6001 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6002 - Fix fsfilt_extN_readpage() to read a full page of directory
6003 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6004 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6005 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6006 - Most intent operations were removed, in favour of a new RPC mode
6007 that does a single RPC to the server and bypasses most of the VFS
6008 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6009 - Aggressively cancel local locks on DLM servers
6010 - mds_reint_unlink sends EA to the client if it's the last nlink.
6011 client uses that EA to unlink OST objects.
6012 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6013 - recursive symlinks were fixed (439)
6014 - fixed NULL deref in DEBUG_REQ
6015 - filter_update_lastobjid no longer calls sync, which annoyed extN
6016 - fixed multi-client small-writes to a single file problem (445)
6017 - fixed mtime updates during file writes (607)
6018 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6019 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6020 - fixed Linux OST size reporting problem (444, 656)
6021 - OST now updates object mtime with writes or setattr (607, 619)
6022 - client verifies file size before zeroing page past EOF (445)
6023 - OST now writes last allocated objid to disk with allocation (108)
6024 - LOV on echo now works (409)
6026 - mds_reint_unlink sends a new buffer, with the EA included. this
6027 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6028 set if a regular file was being unlinked, and it was the last link
6029 - use PtlGet from the target for bulk writes (315)
6030 - OST now updates object mtime with writes or setattr (607, 619)
6031 - LDLM now has a grant-time callback to revalidate locked items, if
6033 - Many MDS operations were reorganized to combat race conditions
6035 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6036 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6038 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6041 - Fully reactivate OST imports after reconnection (512, others)
6042 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6043 - More graceful error handling for truncating on dead OST (515)
6044 - Don't error out unless we're actually accessing dead stripes (474)
6045 - Fix garbage sizes when stripes are missing (410)
6046 - LRU counters were broken, causing constant lock purge (433, 432)
6047 - garbage on read from stripes with failed OSTs (441)
6048 - mark OSCs as active before reconnecting during recovery (438)
6049 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6050 - lfind did not preserve OST order in output (443)
6051 - symlinks cause hung clients, incorrect data (439)
6052 - stop dereferencing request after dropping refcount (457)
6053 - don't LASSERT(spin_is_locked) on non-SMP (455)
6054 - fixes for many rename() bugs
6055 - fstat didn't correctly synchronize attributes (399)
6056 - server must handle lock cancellation during blocking AST prep (487)
6057 - bulk descriptors were free()d too soon (511)
6058 - fix paths in lconf, which would load incorrect modules (451, 507)
6059 - fix confusing lconf 'host not found' error message (386)
6060 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6061 - fix race condition in mdc_blocking_ast() for inode access (526)
6062 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6063 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6064 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6065 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6066 - fix lstripe in several ways (kernel side) (527)
6067 - fix request leak in ldlm_cli_enqueue (262)
6068 - incorrect OSC was marked inactive after OST failure
6069 - call mds_fs_cleanup before unmounting filesystem (524)
6070 - fix races between taking ns_lock and ldlm_lock_change_resource
6071 - fix races updating LOV export open file list
6072 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6073 - fix recovery NULL deref in ldlm_cli_cancel_unused
6074 - fix some DLM races by using new hash table for lock handles (419)
6075 - permit the client to specify desired inodes, at replay
6076 - duplicate requests when we queue them for replay reintegration
6077 - fix last_rcvd offset calculation
6078 - sync after each recovered transaction, so we always make progress
6079 - never, not always, ERESTART requests without transnos
6080 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6082 - skip replay if the MDS says that the client is already connected
6083 - don't check for a recovery-enabled export to match lctl's UUID
6084 - don't INC_USE_COUNT for phantom exports
6085 - don't crash when cleaning up phantom exports (567)
6086 - don't double-finish or set replay data for errored mdc_open requests
6087 - abort requests when they time out, so we don't get old replies
6088 - send/receive replies for AST messages again
6089 - if the client says that it doesn't have the lock, cancel it on the
6091 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6092 mark it as destroyed, it will all work out when decref is called
6093 - fix module use counts (22, 581)
6095 - ASTs now expect a reply (server cancels lock on error reply)
6097 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6100 - fix many simultaneous client startup (392)
6101 - fix dentry->d_it clobbering
6102 - credentials weren't being shipped for readdir/getattr operations
6103 - remove invalid assertions triggered during some concurrent MD
6105 - proper Lustre versions added (336, 389)
6106 - fix memory leak for create error case (398)
6107 - fix LOV locking bug that would get cli/srv out of sync
6108 - fix echo client over LOV (409)
6109 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6110 - fix double-O_EXCL intent crash (424)
6111 - avoid sending multiple lock CANCELs (352)
6113 - MDS can do multi-client recovery (modulo bugs in new code)
6115 - many updates, edits, cleanups
6117 2002-11-18 Phil Schwan <phil@clusterfs.com>
6120 - fix null d_it dereference (346)
6121 - fix full OST/dbench hang (333)
6122 - fix permission problem with file removal (286)
6123 - fix removal of OSCs from LOV when they fail
6124 - fix NULL deref during bulk timeout (214)
6125 - fix problems related to multiple filesystems on one MDS (241)
6126 - fixed serious subtle metadata locking bugs
6127 - free locks on clients when inodes are removed due to memory
6129 - fix inode pointer in lock data (285)
6130 - partial support for multiple MDS on a single host (241)
6131 - data locks weren't cancelled at clear_inode time (290, 311)
6132 - intent locks could lead to unbounded lock growth (205)
6133 - added a maximum lock count, an LRU list, and a flusher
6134 - fix multiple rename (365)
6135 - properly abstracted the echo client
6136 - OSC locked 1 byte too many; fixed
6137 - rewrote brw callback code:
6138 - fixed recovery bugs related to LOVs (306)
6139 - fixed too-many-pages-in-one-write crash (191)
6140 - fixed (again) crash in sync_io_timeout (214)
6141 - probably fixed callback-related race (385)
6143 - Add capability to MDS protocol
6144 - LDLM cancellations and callbacks on different portals
6146 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6149 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6150 - defer open object destruction to close time (601981, 138)
6151 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6152 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6153 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6154 - don't call obd_setattr in ll_file_release for destroyed objects
6155 * protocol change to lustre_msg: move |version| and add |flags|
6156 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6157 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6158 * added replay of create, unlink, link and rename operations during
6159 MDS failover; recovery should be much more robust now
6160 * remove failed OSCs from LOVs (only lov_create uses this so far)
6161 * the lustre-HOWTO was brought (more) up to date (582544)
6163 2002-10-23 Phil Schwan <phil@clusterfs.com>
6166 - in-use dentries weren't being reused properly (617851)
6167 - prevent multiple LDLM setup (599178)
6168 - fix LOV size calculations for truncate (617853)
6169 - fix client handling of MDS intent errors (POSIX)
6170 - fix permission bug in lovstripe.c test (624321)
6171 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6172 - truncate past end of file could corrupt data
6173 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6174 - a race in recovery could return ETIMEDOUT to apps (623947)
6175 - building outside the source directory was fixed
6176 * the lustre-HOWTO was brought (more) up to date (582544)
6177 * major progress was made on recovery functionality
6179 2002-10-10 Phil Schwan <phil@clusterfs.com>
6182 - recovery deadlock fix
6183 - rm -rf causes LBUG fix (617817)
6184 - file open by multiple tasks fix (618962)
6185 - directory permissions bugs (602707 and 620007)
6186 - journal_stop fixed with locking (611313)
6187 - O_APPEND failures resolved (618273, perhaps 614459)
6188 - lconf PATH fix (619770)
6189 - IA64 build fix (621450)
6190 - RPC buffer sizes scale with amount of memory
6192 2002-10-01 Phil Schwan <phil@clusterfs.com>
6195 - locks would be cancelled without throwing away data pages,
6196 resulting in inconsistent data (605627)
6197 - inode attributes were not always being refreshed (605627, 612449)
6198 - lconf now continues to cleanup after lctl reports an error
6199 - MDS now enforces user permissions (602707)
6200 - lprocfs cleanup fixed, but not yet enabled (614157)
6201 - fixed infinite server hang, should a client not respond to an AST
6202 - avoid going into recovery if user calls readlink() with a buffer
6203 that's too small (613941)
6204 - AST RPCs no longer require replies (614867) -- this may be changed
6205 - don't crash server if client sends an IOV that's too big (611336)
6206 - fixed lock conversion deadlock (611892)
6207 - fixed the following of symlinks (614622)
6208 * recovery: the server can remove locks from a client that dies, other
6209 clients can make progress
6210 * more extN patch fixes
6211 * compile-time configurable ptlrpc buffer allocations
6213 - collaborative read cache document
6214 - Lustre Lite Performance CDR document-in-progress
6216 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6219 - fix typo in patch-2.4.18
6221 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6224 - clear ptlrpc request each time in handle_incoming_request()
6225 - unlink of files now destroys the object on the OST
6227 2002-09-19 Peter Braam <braam@clusterfs.com>
6229 * add hard link support
6230 * change obdfile creation method
6231 * kernel patch changed
6233 2002-09-19 Peter Braam <braam@clusterfs.com>
6236 - stack overflow bug in extN fixed
6238 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6240 * documentation updates
6241 - add man pages for config tools
6242 - update tests/README to describe testing with new config tools
6243 - finish metadata API descriptions
6244 * bug fixes and cleanups
6245 - statfs workaround for 16TB limit
6246 - LOV stripe allocation improved, can stripe on subset of OSTs
6247 - LOV file size/IO offset was wrong for files > 4GB in size
6248 - object EA data was being dropped, caused files to be unreadable
6249 - memory overflow with non-LOV OST caused memory corruption
6250 - fixed regression tests to work with new config tools, obdfilter
6251 - fixed bug when directory size became larger than 1 block
6252 - fixed bug (for single client case) when PWD was deleted
6253 - invalidate local directory pages when doing intent-based ops
6254 - avoid LDLM oops when lock callback contained bad data
6256 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6258 * documentation updates
6259 * bug fixes and cleanups
6260 - configuration tools
6263 - 64-bit compile warnings
6264 - 64-bit internal statfs data
6266 * test_brw on persistent OST devices
6268 * lprocfs (disabled)
6270 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6272 * documentation updates
6273 * bug fixes and cleanups
6274 * configuration tools
6276 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6279 * many small fixes to 0_5_4
6280 * io/network handling
6281 * thinkos in MDS operations
6283 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6286 * crucial basic fixes to 0.5.3
6287 * IOR, Iozone work over Elan
6290 2002-08-07 Phil Schwan <phil@clusterfs.com>
6291 * version 0_5_3, our first alpha
6292 * we use the new Portals iovs
6293 * documentation updates
6294 * bug fixes and cleanups
6295 * small changes in the DLM wire protocol
6297 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6298 * version 0_5_1 with some initial stability,
6299 * locking on MD and file I/O.
6300 * documentation updates
6301 * several bug fixes since 0.5.0
6302 * small changes in wire protocol
6304 2002-07-18 Phil Schwan <phil@clusterfs.com>
6306 * delivered as Lustre Light Alpha
6307 * fixed a crash after handling invalid MDS requests
6308 * fixed directory pages for architectures with non-4k pages sizes
6310 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6311 * release version v0_4_4
6312 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6314 * Fixes a number of interruption problems with OST operations.
6315 * Update documentation for portals header changes
6316 * Move all wire protocol structs/defines to lustre_idl.h
6317 * Fixes symlink length bug.
6318 * Add tcpdump to repository.
6320 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6321 * release version v0_4_3
6322 * Fixes statfs for inodes on extN.
6323 * Fixes bug in runtests which would delete /etc/hosts.
6324 * Use 64-bit object IDs wherever possible (not into VFS though)
6325 Remove ost_get_info, which is unused by lustre, and out of date.
6327 2002-07-03 Peter Braam <braam@clusterfs.com>
6328 * release version v0_4_2 Fixes a lookup error (type not passed)
6329 * move forward to head of Portals
6330 * move forward to latest Lustre kernel
6332 2002-06-25 Peter Braam <braam@clusterfs.com>
6333 * release version v0_4_1. Hopefully stable on single node use.