3 * Server support for kernels:
4 2.6.18-164.11.1.el5 (RHEL 5)
5 2.6.18-164.11.1.0.1.el5 (OEL 5)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-164.11.1.el5 (RHEL 5),
9 2.6.18-164.11.1.0.1.el5 (OEL 5)
10 2.6.16.60-0.42.8 (SLES 10),
12 2.6.29.4-167.fc11 (FC11)
13 * Recommended e2fsprogs version: 1.41.10-sun2
14 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
15 Lustre filesystem with 4K stack may cause a stack overflow. For more
16 information, please refer to bugzilla 17630.
18 --------------------------------------------------------------------------------
20 2010-07-15 Oracle, Inc.
22 * Server support for kernels:
23 2.6.18-164.11.1.el5 (RHEL 5)
24 2.6.18-164.11.1.0.1.el5 (OEL 5)
25 * Client support for unpatched kernels:
26 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
27 2.6.18-164.11.1.el5 (RHEL 5),
28 2.6.18-164.11.1.0.1.el5 (OEL 5)
29 2.6.16.60-0.42.8 (SLES 10),
31 2.6.29.4-167.fc11 (FC11)
32 * Recommended e2fsprogs version: 1.41.10-sun2
34 Severity : enhancement
36 Description: Distribute lfs_migrate script with lustre
37 Details : The lfs_migrate mentioned in the manual is now included in the
38 lustre tree in lustre/scripts and will distributed in lustre RPMs.
41 Severity : enhancement
43 Description: Failover nid list added to import proc file.
45 Severity : enhancement
47 Description: "som_preview" mount option required to use SOM
48 Details : The clients will need to use som_preview mount option
52 Description: Lustre does not do 1MB IOs to HW RAID
53 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
55 Severity : enhancement
57 Description: lfs quota output is not very convenient for awk/sed-parsing
58 Details : Some positions in lfs quota output table could be empty or
59 non-empty which made it hard to parse it with scripts, now
60 a dash is put instead of space where there is not supposed
65 Description: Update to OFED 1.5.1
66 Details : OFED 1.5.1 will be used in environments that don't use
71 Description: Remove lnet. prefix from lctl params display
72 Details : Also renamed lnet.memused to lnet_memused to
77 Description: Use conf_param -d to remove permanant settings.
81 Description: Add list_param -R
82 Details : lctl list_param now supports recursive mode.
85 Bugzilla : 21829, 22312
86 Description: Fix llobdstat.
87 Details : 1. Now updating the write_bytes counter
88 2. Port script fixes from 1.8.
92 Description: decrease the usage of memory on clients.
93 Details : 1. On clients, recycle dentries and inodes unused.
94 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
99 Description: llobdstat fix and enhancement
100 Details : add a counter to set a limit to how many samples will be returned
101 fix a wildcard in the path to limit to obdfilter stats only
105 Description: lfs check only allowed for root
106 Details : Code cleanups and added a regresion test of non-root use of
111 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
112 Details : Fix hr->hr_index overload.
116 Description: Fix FIEMAP functionality.
120 Description: FC11 patchless client support added
124 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
125 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
129 Description: New acc-sm script and various test-framework enhancments.
133 Description: Previously landed patch was reverted.
137 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
141 Description: Add ldiskfs maxdirsize mount option.
145 Description: Fix mkfs.lustre support for 16TB LUNs.
149 Description: Print mmp_check_interval
150 Details : print mmp_check_interval and make it possible to abort
151 mount operation in case it takes too long.
155 Description: Add cascading_rw to lustre mpi tests.
159 Description: Store SOM data in EA.
160 Details : Also several SOM related cleanups.
164 Description: Add refcount to md_open_data
168 Description: Use refcount on lov_request_set to prevent memory corruption.
172 Description: Statahead: use parent dir's i_mutex for synchronization.
176 Description: loadgen improvements.
180 Description: Enable peer health detection in ksocklnd and ko2iblnd.
184 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
188 Description: Fix race on obd_devcie:md_stats
192 Description: Fix SLES 11 client builds.
193 Details : Resolved uninitalized pointer gcc warnings.
197 Description: Allow non-root access for "lfs check".
198 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
202 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
204 Severity : enhancement
206 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
207 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
211 Description: Sanity test_27p fails.
212 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
213 Sanity tests 27 were cleaned up and improved.
217 Description: Unify req format on client/servers
218 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
219 buffers and swabbers approach..
224 Description: drop join file stuff.
226 Severity : enhancement
228 Description: Add 2.6.27 and SLES11 (patchless client) support
232 Description: Fix in ptlrpc_expire_one_request() to print the signed time
235 Severity : enhancement
237 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
240 Severity : enhancement
242 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
244 Severity : enhancement
246 Description: correctly handle big reply message.
247 Details : send LNet event if reply is bigger then buffer and adjust this buffer
252 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
254 Severity : enhancement
256 Description: Add support for OEL5.
260 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
264 Description: client could not reconnect to OST because of an active request.
265 Details : The client now retries to reconnect to the same server, if a
266 connect request failed with EBUSY or -EAGAIN. Test case 35b
267 conf-sanity.sh has been created.
271 Description: Don't allow make backward step on assignin osc next id.
272 Details : race between allocation next id and ll_sync thread can be cause
273 of set wrong osc next id and can be kill valid ost objects.
277 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
281 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
285 Description: Error handling in osc_statfs_interpret() has been improved.
286 Details : Check in osc_statfs_interpret() for EBADR.
290 Description: Do not send statfs() requests to OSTs disabled by administrator.
291 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
293 Severity : enhancement
295 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
297 Severity : enhancement
299 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
301 Severity : enhancement
303 Description: Enhancement for lfs(1) command to use numeric uid/gid.
307 Description: Improved cleanup procedure for conf-sanity test_46a.
311 Description: Do not start bulk transfer, if client tried to reconnect.
312 Details : Do not start bulk transfer, if client tried to reconnect,
313 but failed due to active rpcs. Check for exp_abort_active_req
314 has been implemented.
318 Description: Prevent inconsistences between linux and lustre mount structures.
319 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
320 Make the sleep interruptible.
322 Severity : enhancement
324 Description: provide server to client comms path
325 Details : server to client communications path via new LDLM_SET_INFO rpc
329 Description: truncate starts GFP_FS allocation under transaction causing deadlock
330 Details : ldiskfs_truncate calls grab_cache_page which may start page
331 allocation under an open transaction. This may lead to
332 calling prune_icache with consequent lustre reentrance.
336 Description: send correctly lsm on open replay
337 Details : MDS is trust to LSM size on replay open, but client can set wrong size
342 Description: enable client interface failover
343 Details : When a child reconnects from another NID, properly update export
344 nid hash position and ldlm reverse import.
346 Severity : enhancement
348 Description: Communicate OST degraded/readonly state via statfs to MDS
349 Details : Flags in the statfs returned from OSTs indicate whether the
350 OST is in a degraded RAID state, or if the filesystem has
351 turned read-only after a filesystem error is detected.
356 Description: don't panic if EPROTO was hit when reading symlink
357 Details : correctly handling request reference in error cases.
359 Severity : enhancement
361 Description: Add LustreNetLink, a kernel-userspace communcation path.
363 Severity : enhancement
365 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
370 Description: access to llog context before init.
371 Details : move handling CATALOGS file at osc layer and forbid access to llog
375 Frequency : rare, only when client is writing to OST on local node
377 Description: Avoid deadlock for local client writes
378 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
379 memory freeing context. This allows OST threads to set the
380 PF_MEMALLOC flag on task structures in order to allocate memory
381 from reserved pools and complete IO.
382 Use GFP_HIGHUSER for OST allocations for non-local client writes,
383 so that the OST threads generate memory pressure and allow
384 inactive pages to be reclaimed.
386 Severity : enhancement
388 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
393 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
394 Details : move ldlm namespace creation in setup phase to avoid grab
395 _lprocfs_lock with cli_sem held.
399 Description: Temporarily disable grant shrink.
400 Details : Disable the feature for debugging.
403 Frequency : only when formatting test filesystems
405 Description: Unable to run several mkfs.lustre on loop devices at the same time
406 Details : mkfs.lustre returns error 256 on the concurrent loop devices
407 formatting. The solution is to proper handle the error.
409 Severity : enhancement
411 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
413 Severity : enhancement
415 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
420 Description: resolve race between obd_disconnect and class_disconnect_exports
421 Details : if obd_disconnect will be called to already disconnected export he
422 forget release one reference and osc module can't unloaded.
424 Severity : enhancement
426 Description: Allow tuning service thread via /proc
427 Details : For each service a new
428 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
429 created that can be used to set min/max thread counts, and get the
430 current number of running threads.
435 Descriptoin: don't return error if have partially created objects for file.
436 Details : lov_update_create_set uses set->set_success as index for created
437 objects, so if some requests will be failed, they will have hole
438 at end of array and qos_shrink_lsm for allocate correct lsm.
440 Severity : enhancement
442 Description: Update OFED support to 1.4.1
447 Description: Reduce small size read RPC
448 Details : Set read-ahead limit for every file and only do read-ahead when
449 available read-ahead pages are bigger than 1M to avoid small size
452 Severity : enhancement
454 Description: move AT tunable parameters for more consistent usage
455 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
457 Severity : enhancement
459 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
460 Details : allow skip disconnected ost for send statfs request and hide error
466 Description: fix racy locking of mballoc block bitmaps causing BUG
467 Details : The locking of the mballoc buddy bitmap and the in-memory
468 block bitmap was using two different spin locks in some
469 cases. This made it possible to incorrectly access the
470 mballoc bitmap while another process was modifying it,
471 causing a sanity assertion to fail. While no on-disk corruption
472 was reported, there was some risk of this happening.
475 Frequency : rare, on llog test 6
477 Descriptoin: don't allow connect to already connected import
478 Details : allowing connect to already connected import is hide connecting problem.
483 Descriptoin: allow kill process which wait statahead result
484 Details : for some reasons 'ls' can stick in waiting result from statahead,
485 in this case need way for kill this process.
487 Severity : enhancement
489 Description: Add state history info file, enhance import info file
490 Details : Track import connection state changes in a new osc/mdc proc file;
491 add overview-type data to the osc/mdc import proc file.
493 Severity : enhancement
495 Description: MDS create should not wait for statfs RPC while holding DLM lock.
497 Severity : enhancement
499 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
504 Descriptoin: OST grant too much space to client even there are not enough space.
505 Details : Client will shrink its grant cache to OST if there are no write
506 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
507 this grant cache if there are already not enough avaible space
508 (left_space < total_clients * 32M).
511 Frequency : start MDS on uncleanly shutdowned MDS device
513 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
514 Details : stay in waiting mds<>ost recovery finished produce random bugs
515 due race between two ll_sync thread for one lov target. send
516 ACTIVATE event only if connect realy finished and import have
520 Frequency : rare, connect and disconnect target at same time
522 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
523 Details : don't call obd_disconnect under lov_lock. this long time
524 operation and can block ptlrpcd which answer to connect request.
529 Descriptoin: don't lose wakeup for imp_recovery_waitq
530 Details : recover_import_no_retry or invalidate_import and import_close can
531 both sleep on imp_recovery_waitq, but we was send only one wakeup
535 Frequency : always with long access acl
537 Descriptoin: mds can't pack reply with long acl.
538 Details : mds don't control size of acl but they limited by reint/getattr
541 Severity : enhancement
543 Description: Update to SLES10 kernel-2.6.16.60-0.33.
545 Severity : enhancement
547 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
550 Frequency : start MDS on uncleanly shutdowned MDS device
552 Descriptoin: aborting recovery hang on MDS
553 Details : don't throttle destroy RPCs for the MDT.
556 Frequency : on remount
558 Description: external journal device not working after the remount
559 Details : clear dev_rdonly flag for external journal devices in
565 Description: shutdown vs evict race
566 Details : client_disconnect_export vs connect request race.
567 if client will evicted at this time - we start invalidate
568 thread without referece to import and import can be freed
572 Frequency : rare, need acl's on inode.
574 Description: client can't handle ost additional correctly
575 Details : if ost was added after client connected to mds client can have
576 hit lnet_try_match_md ... to big messages to wide striped files.
577 in this case need teach client to handle config events about add
578 lov target and update client max ea size at that event.
580 Severity : enhancement
582 Description: Changelogs
583 Details : Changelogs are a lightweight mechanism to track filesystem
584 metadata and namespace changes. The changelog is recorded
585 permanently on the MDTs, and is periodically "consumed" / purged
586 when records are no longer needed.
588 Severity : enhancement
590 Description: compact fld format with extents
591 Details : Store range of seq rather than every seq in FLD. Seq
592 controller update FLD rather than clients. In Case of CMD, mdt0
593 has FLD, all other metadata server act as non persistent proxy
594 for FLD queries and cache fld entries in fld cache.
599 Description: don't skip ost target if they assigned to file
600 Details : Drop slow OSCs if we can, but not for requested start idx.
601 This means "if OSC is slow and it is not the requested
602 start OST, then it can be skipped, otherwise skip it only
603 if it is inactive/recovering/out-of-space.
607 Description: more cleanup in mds_lov
608 Details : not send LOV EA under replay, we can't know about they size at this
609 time. Don't allow client connect to mds before any ost connected,
610 for avoid problems with LOV EA size and returning EIO to client.
612 Severity : enhancement
614 Description: Interoperability at server side (Disk interoperability)
616 Severity : enhancement
618 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
620 Severity : enhancement
622 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
624 Severity : enhancement
626 Description: New client IO stack (CLIO).
628 Severity : enhancement
630 Description: Commit on sharing. Eliminate inter-client dependencies between
631 uncommitted transactions by doing transaction commits.
632 Thereby clients may recovery independently.
635 Frequency : Create a symlink file with a very long name
637 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
638 Details : If there is no extra space in the request for early cancels,
639 ldlm_req_handles_avail() returns 0 instead of a negative value.
641 Severity : enhancement
643 Description: Add /proc entry for import status
644 Details : The mdc, osc, and mgc import directories now have
645 an import directory that contains useful import data for debugging
648 Severity : enhancement
650 Description: Re-disable certain /proc logging
651 Details : Enable and disable client's offset_stats, extents_stats and
652 extents_stats_per_process stats logging on the fly.
655 Frequency : Only on FC kernels 2.6.22+
657 Description: oops in statahead
658 Details : Do not drop reference count for the dentry from VFS when lookup,
659 VFS will do that by itself.
661 Severity : enhancement
663 Description: Generic /proc file permissions
664 Details : Set /Proc file permissions in a more generic way to enable non-
665 root users operate on some /proc files.
669 Description: Hitting mdc_commit_close() ASSERTION
670 Details : Properly handle request reference release in
671 ll_release_openhandle().
675 Description: quota recovery deadlock during mds failover
676 Details : This patch includes att18982, att18236, att18237 in bz14840.
678 1. fix osts hang when mds does failover with quotaon
679 2. prevent watchdog storm when osts threads wait for the
684 Frequency : only patchless client
685 Description: add workaround for race between add/remove dentry from hash
687 Severity : enhancement
689 Description: Allow OST glimpses to return PW locks
693 Description: LBUG when llog conf file is full
694 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
698 Description: Prevent import from entering FULL state when server in recovery
702 Description: service mount cannot take device name with ":"
703 Details : Only when device name contains ":/" will mount treat it as
709 Description: replace ptlrpcd with the statahead thread to interpret the async
710 statahead RPC callback
714 Frequency : on recovery
715 Description: I/O failures after umount during fail back
716 Details : if client reconnected to restarted server we need join to recovery
717 instead of find server handler is changed and process self
718 eviction with cancel all locks.
720 Severity : enhancement
722 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
724 Severity : enhancement
726 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
728 Severity : enhancement
730 Description: Upcall on Lustre log has been dumped
731 Details : Allow for a user mode script to be called once a Lustre log has
732 been dumped. It passes the filename of the dumped log to the
733 script, the location of the script can be specified via
734 /proc/sys/lnet/debug_log_upcall.
739 Description: avoid idr_remove called for id which is not allocated.
740 Details : Move assigment s_dev for clustered nfs to end of initialization,
741 to avoid problem with error handling.
746 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
747 Details : When connection is reused this not moved from CONN_UNUSED_HASH
748 into CONN_USED_HASH and this prodice warning when put connection
749 again in unused hash.
752 Severity : enhancement
754 Description: File striping can now be set to use an arbitrary pool of OSTs.
756 Severity : enhancement
758 Description: Export bytes_read/bytes_write count on OSC/OST.
762 Description: Early reply size mismatch, MGC loses connection
763 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
764 the connect flags are properly negotiated.
769 Description: quotas are not honored with O_DIRECT
770 Details : all writes with the flag O_DIRECT will use grants which leads to
771 this problem. Now using OBD_BRW_SYNC to guard this.
775 Description: add quota statistics
776 Details : 1. sort out quota proc entries and proc code.
777 2. add quota statistics
779 Severity : enhancement
781 Description: enable quota support for HEAD.
785 Description: Properly propagate oinfo flags from lov to osc for statfs
786 Details : restore missing copy oi_flags to lov requests.
788 Severity : enhancement
790 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
791 : l_getgroups(8), lst(8), routerstat(8)
792 Details : included man pages for llobdstat(8), llstat(8),
793 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
795 Severity : enhancement
797 Description: configure's --enable-quota should check the
798 : kernel .config for CONFIG_QUOTA
799 Details : configure is terminated if --enable-quota is passed but
800 : no quota support is in kernel
804 Description: Remove portals compatibility
805 Details : Remove portals compatibility, not interoperable with releases
810 Description: Resolve device initialization race
811 Details : Prevent proc handler from accessing devices added to the
812 obd_devs array but yet be intialized.
814 Severity : enhancement
816 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
818 Severity : enhancement
820 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
825 Description: Using wrong pointer in osc_brw_prep_request
826 Details : Access to array[-1] can produce panic if kernel compiled with
827 CONFIG_PAGE_ALLOC enabled
831 Description: Client runs out of low memory
832 Details : Consider only lowmem when counting initial number of llap pages
836 Description: *optional* service tags registration
837 Details : if the "service tags" package is installed on a Lustre node
838 When the filesystem is mounted, a local-node service tag will
839 be created. See http://inventory.sun.com/ for more information
840 about the Service Tags asset management system.
844 Description: Kernel BUG tries to release flock
845 Details : Lustre does not destroy flock lock before last reference goes
846 away. So always drop flock locks when client is evicted and
847 perform unlock regardless of successfulness of speaking to MDS.
851 Description: add refcount protection for osc callbacks, avoid panic on shutdown
855 Description: sanity test 65a fails if stripecount of -1 is set
856 Details : handle -1 striping on filesystem in ll_dirstripe_verify
861 Description: ASSERTION(CheckWriteback(page,cmd)) failed
862 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
865 Severity : enhancement
867 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
871 Description: do not process already freed flock
872 Details : flock can possibly be freed by another thread before it reaches
873 to ldlm_flock_completion_ast.
877 Description: LBUG during stress test
878 Details : Need properly lock accesses the flock deadlock detection list.
882 Description: oops in page fault handler
883 Details : kernel page fault handler can return two special 'pages' in error
884 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
888 Description: timeout with invalidate import.
889 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
890 should be handled by ptlrpcd. This produce long age waiting and
891 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
893 Severity : enhancement
895 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
899 Description: enable MGS and MDT services start separately
900 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
901 a co-located MGS without starting the MGS, which is a complement
902 to 'nosvc' mount option.
906 Frequency : after recovery
907 Description: precreate to many object's after del orphan.
908 Details : del orphan st in oscc last_id == next_id and this triger growing
909 count of precreated objects. Set flag LOW to skip increase count
910 of precreated objects.
914 Frequency : rare, on clear nid stats
915 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
916 Details : when clean nid stats sometimes try destroy live entry,
917 and this produce panic in free.
921 Description: Stack overflow during MDS log replay
922 ease stack pressure by using a thread dealing llog_process.
926 Description: wait until IO finished before start new when do lock cancel.
927 Details : VM protocol want old IO finished before start new, in this case
928 need wait until PG_writeback is cleared until check dirty flag and
929 call writepages in lock cancel callback.
931 Severity : enhancement
933 Description: using special macro for print time and cleanup in includes.
937 Description: mds_mfd_close() ASSERTION(rc == 0)
938 Details : In mds_mfd_close(), we need protect inode's writecount change
939 within its orphan write semaphore to prevent possible races.
943 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
947 Frequency : rare, on shutdown ost
948 Description: don't hit live lock with umount ost.
949 Details : shrink_dcache_parent can be in long loop with destroy dentries,
950 use shrink_dcache_sb instead.
954 Description: don't panic with use echo client
955 Details : echo client pass NULL as client nid pointer and this produce null
960 Description: fix build on ppc32
961 Details : compile code with -m64 flag produce wrong object file for ppc32.
965 Description: add message levels for liblustreapi
969 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
970 Details : Move the 'good_osts' check before the 'total_bavail' check. This
971 will result in an -EAGAIN and in the exit call path we call
972 alloc_rr() which will with increasing aggressiveness attempt to
973 aquire precreated objects on the minimum number of required OSCs.
977 Description: Use old size assignment to avoid deadlock
978 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
979 the scheduling while holding a spinlock. We do not need locking
980 for size in ll_update_inode() because size is only updated from
981 the MDS for directories or files without objects, so there is no
982 other place to do the update, and concurrent access to such inodes
983 are protected by the inode lock.
987 Description: resolve "_IOWR redefined" build error on SLES10
991 Description: dump the memory debugging after all modules are unloaded to
992 suppress false negative in conf_sanity test 39
994 Severity : enhancement
996 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1001 Description: fix SLES kernel versioning
1002 Details : the kernel version for our SLES 10 kernel did not include a "-"
1003 before the "smp" at the end. while this was not a problem in
1004 general, it did mean that software trying to use the kernel
1005 version to try to detect a vendor specific kernel would fail.
1006 this was most evident by the OFED build scripts.
1010 Description: Don't update lov_desc members until making sure they are valid
1011 Details : When updating lov_desc members via proc fs, need fix their
1012 validities before doing the real update.
1016 Description: don't put request into delay list while invalidate in flight.
1017 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1018 import in flight. this produce timeout for invalidate and sometimes
1019 can cause stale data.
1023 Frequency : on ppc only
1024 Description: not convert ost objects for directory because it's not exist.
1025 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1029 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1031 Severity : enhancement
1033 Description: Update to support 2.6.22.14 vanilla kernel.
1037 Frequency : rare, on recovery
1038 Description: read procfs can produce deadlock in some situation
1039 Details : Holding lprocfs lock which send rpc can produce block for destroy
1040 obd objects and this also block reconnect with -EALREADY. This
1041 isn't fix all lprocfs bugs - but make it rare.
1043 Severity : enhancement
1045 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1048 Frequency : frequent on X2 node
1050 Description: mdc_set_open_replay_data LBUG
1051 Details : Set replay data for requests that are eligible for replay.
1055 Description: lustre_mgs: operation 101 on unconnected MGS
1056 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1057 MGC, and late on MGC cannot successfully connect to MGS and a lot
1058 of the error messages complaining that MGS is not connected.
1062 Frequency : on network error
1063 Description: panic with double free request if network error
1064 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1065 but it's true only for synchronus enqueue, for async enqueue
1066 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1069 Severity : enhancement
1071 Description: client-side metadata stat-ahead during readdir(directory readahead)
1072 Details : perform client-side metadata stat-ahead when the client detects
1073 readdir and sequential stat of dir entries therein
1076 Frequency : on start mds
1078 Description: Implement get_info(last_id) in obdfilter.
1081 Frequency : occasional
1083 Description: Correctly check stale fid, not start epoch if ost not support SOM
1084 Details : open with flag O_CREATE need set old fid in op_fid3 because
1085 op_fid2 was overwritten with new generated fid, but mds can answer
1086 with one of these two fids and both is not stale. Setattr
1087 incorrectly started an epoch and assume will be called
1088 done_writing, but without SOM done_writing ever being called.
1091 Frequency : rare, depends on device drivers and load
1093 Description: MDS or OSS nodes crash due to stack overflow
1094 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1095 In some cases, in conjunction with device drivers that use a lot
1096 of stack the MDS (or possibly OSS) service threads could overflow
1097 the stack. One change which was identified to consume additional
1098 stack has been reworked to avoid the extra stack usage.
1101 Frequency : occasional
1103 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1104 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1105 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1106 oscc_has_objects and see OSC has no objests and call
1107 oscc_internal_create to resend create request.
1109 Severity : enhancement
1111 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1113 Severity : enhancement
1115 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1118 Frequency : very rare
1120 Description: Fix replay if there is an un-replied request and open
1121 Details : In some cases, older replay request will revert the
1122 mcd->mcd_last_xid on MDS which is used to record the client's
1123 latest sent request.
1125 Severity : enhancement
1127 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1129 Severity : enhancement
1131 Description: Add rhel5 support to HEAD.
1133 Serverity : enhancement
1135 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1140 Description: Don't allow skipping OSTs if index has been specified.
1141 Details : Don't allow skipping OSTs if index has been specified, make
1142 locking in internal create lots better.
1146 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1147 Details : access to bitfield in structure is always rounded to long
1148 and this produce problem with not atomic change any bit.
1152 Description: Lustre make rpms failed.
1153 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1154 builds Lustre rpms from tarball.
1157 Frequency : rare on shutdown ost
1159 Description: If llog cancel was not send before clean_exports phase, this can
1160 produce deadlock in llog code.
1161 Details : If llog thread has last reference to obd and call class_import_put
1162 this produce deadlock because llog_cleanup_commit_master wait when
1163 last llog_commit_thread exited, but this never success because was
1164 called from llog_commit_thread.
1168 Description: allow userland application know is lost one of stripes.
1169 Details : fill lvb_blocks with error code on ost and return it to
1170 application if error flag found.
1174 Description: NULL lov_tgts causing MDS oops
1175 Details : more safe checks for NULL lov_tgts for avoid oops.
1177 Severity : enhancement
1179 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1183 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1184 Details : If an OST is down the MDS will hang indefinitely in
1185 obd_statfs_rqset() waiting for the statfs data. While for
1186 MDS QOS usage of statfs, it should not stuck in waiting.
1188 Severity : enhancement
1190 Description: remote_acl support
1191 Details : Support ACL-based permission check for remote user.
1192 Support setfacl/getfacl for remote user with the utils
1193 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1194 the system "{s,g}etfacl" utils.
1196 Severity : enhancement
1198 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1200 Severity : enhancement
1202 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1206 Description: make mgs_setparam() handle fsname containing dash
1207 Details : fsname containing a dash does not work with lctl conf_param
1209 Severity : enhancement
1211 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1213 Severity : enhancement
1215 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1219 Frequency : rare, at shutdown
1220 Description: access already free / zero obd_namespace.
1221 Details : if client_disconnect_export was called without force flag set,
1222 and exist connect request in flight, this can produce access to
1223 NULL pointer (or already free pointer) when connect_interpret
1224 store ocd flags in obd_namespace.
1228 Frequency : only at startup
1229 Description: not alloc memory with spinlock held.
1230 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1231 if any spinlock held.
1233 Severity : enhancement
1235 Description: make lustre randomly fail allocating memory
1236 Details : Make lustre randomly failed allocating memory for testing purpose.
1238 Severity : enhancement
1240 Description: lost problems with lov objid file
1241 Details : Fixes some scability and access to not inited memory problems
1242 in work with lov objdid file.
1247 Description: lfs find does not continue on file error
1248 Details : Continue other files processing when a file/dir is absent.
1252 Description: Inconsistent usage of lustre_pack_reply()
1253 Details : Standardize the usage of lustre_pack_reply() such that it
1254 always generate a CERROR on failure.
1257 Frequency : occasional
1259 Description: MDS hang or stay in waiting lock
1260 Details : If client receive lock with CBPENDING flag ldlm need send lock
1261 cancel as separate rpc, to avoid situation when cancel request
1262 can't processed due all i/o threads stay in wait lock.
1266 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1270 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1274 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1278 Description: Update to RHEL4 latest kernel.
1280 Severity : enhancement
1282 Description: Build SLES10 patchless client fails
1283 Details : The configure was broken by run ./configure with
1284 --with-linux-obj=.... argument for patchless client. When the
1285 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1286 can't search header adequately. Use absolute path such as
1287 -I($LINUX)/include instead.
1291 Description: interrupt oig_wait produce painc on resend.
1292 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1293 private set, and this produce situation when rq_ptlrpcd_data not
1294 copyed to new allocated request and triggered LBUG on assert
1295 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1296 ptlrpcd set if request is changed and can be safety changed to use
1301 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1302 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1303 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1304 these macros and replace them with their difinitions instead.
1308 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1309 Details : Update which_patch & target file for SLES10 latest kernel.
1312 Description: Remove client patches from SLES 10 kernel.
1313 Details : This causes SLES 10 clients to behave as patchless clients
1314 even on a Lustre-patched (server) kernel.
1316 Severity : enhancement
1318 Description: self-adjustable client's lru lists
1319 Details : use adaptive algorithm for managing client cached locks lru
1320 lists according to current server load, other client's work
1321 pattern, memory activities, etc. Both, server and client
1322 side namespaces provide number of proc tunables for controlling
1325 Severity : enhancement
1327 Description: light-weight GSS support
1328 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1329 and reduce performance overhead.
1331 Severity : enhancement
1333 Description: Linux keyring support
1334 Details : Support using service of Linux keyring for Lustre GSS internal
1335 context refresh/cache mechanism.
1339 Description: Fix errors in lfs documentation
1340 Details : Fixes man pages
1344 Description: don't use GFP_* in generic Lustre code.
1345 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1349 Description: obdclass is limited by single OBD_ALLOC(idarray)
1350 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1354 Description: warnings with build patchless client with vanila 2.6.19 and up
1355 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1356 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1360 Frequency : Only for SLES
1361 Description: sanity_quota fail test_1
1362 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1363 file, which makes TSTID[2] inunique.
1367 Frequency : Only for SLES9
1368 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1369 Details : Update target/ChangeLog/which_patch .
1373 Frequency : Only for test_10 in sanity-quota.sh
1374 Description: a bug in quota test code
1375 Details : checking a test flag in a wrong place
1379 Frequency : Only for test_13 in sanity-quota.sh
1380 Description: a bug in quota test script
1381 Details : shouldn't check overfull allocated quota
1385 Description: jbd statistics
1386 Details : Port older jbd statistics patch for sles10
1388 Severity : enhancement
1390 Description: Add printing inode info into message about error in writepage.
1395 Description: reply_lock_interpret crash due to race with it and lock cancel.
1396 Details : Do not replay locks that are being cancelled. Do not reference
1397 locks by their address during replay, just by their handle.
1401 Frequency : When flocks are used.
1402 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1403 Details : Flock locks might destroy just granted lock if it could be merged
1404 with another existing flock, this is done in completion handler,
1405 so teach ldlm_cli_enquque_fini that this is a valid case for
1411 Description: Oops in read and write path when failing to allocate lock.
1412 Details : Check if lock allocation failed and return error back.
1416 Description: lstripe command fails for valid OST index
1417 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1418 instead of lov->desc.ld_active_tgt_count.
1422 Description: sanity.sh failed test 103
1423 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1428 Description: df doesn't work properly if diskfs blocksize != 4K
1429 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1433 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1434 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1436 Severity : enhancement
1438 Description: Adaptive timeouts
1439 Details : RPC timeouts adapt to changing server load and network
1440 conditions to reduce resend attempts and improve recovery time.
1444 Description: llapi_file_create() does not allow some changes
1445 Details : add llapi_file_open() that allows specifying the mode and
1446 open flags, and also returns an open file handle.
1450 Description: merge and cleanup kernel patches.
1451 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1455 Description: Add journal checksum support.(Kernel part)
1456 Details : The journal checksum feature adds two new flags i.e
1457 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1458 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1459 indicates that the commit block contains the checksum for
1460 the blocks described by the descriptor blocks. Now commit
1461 record can be sent to disk without waiting for descriptor
1462 blocks to be written to disk. This behavior is controlled
1463 using JBD2_FEATURE_ASYNC_COMMIT flag.
1467 Description: OSS needs mutliple precreate threads
1468 Details : Add ability to start more than one create thread per OSS.
1472 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1473 Details : execute lfs setstripe on client
1477 Description: mds_obd_create error creating tmp object
1478 Details : When the user sets quota on root, llog will be affected and can't
1479 create files and write files.
1482 Frequency : Always on ia64 patchless client, and possibly others.
1484 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1485 Details : This allows the patchless client to be loaded on architectures
1486 without this export.
1490 Description: RedHat Update kernel for RHEL5
1491 Details : Add the kernel config file for RHEL5.
1495 Description: RedHat Update kernel for RHEL5
1496 Details : Modify the kernel config file more closer RHEL5.
1500 Description: Build failure against Centos5 (RHEL5)
1501 Details : Define PAGE_SIZE when it isn't present.
1505 Description: test_7 on sanity-quota.sh can't be used on separate machines
1506 Details : add facet to handle it
1510 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1511 Details : with NFS, the anon dentry's parent was set to itself in
1512 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1513 corresponding dentry other than use rec->ur_name.
1515 Severity : enhancement
1517 Description: lfs setstripe enhancement
1518 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1522 Description: enable data checksumming by default
1523 Details : enable checksum by default, allow --disable-checksum
1524 configure option and "-o nochecksum" mount option. Checksums
1525 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1529 Description: lustre support for RHEL5
1530 Details : Add support for RHEL5.
1534 Description: Client eviction due to failover config
1535 Details : after a connection loss, the lustre client should attempt to
1536 reconnect to the last active server first before trying the
1537 other potential connections.
1541 Description: when mds and osts use different quota unit(32bit and 64bit),
1542 quota will be released repeatly.
1543 Details : void sending multiple quota reqs to mds, which will keep the
1544 status between the reqs.
1548 Description: osts not allocated evenly to files
1549 Details : change the condition to increase offset_idx
1552 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1554 Description: Data corruption for OSTs that are formatted larger than 2TB
1556 Details : When generating the bio request for lustre file writes the
1557 sector number would overflow a temporary variable before being
1558 used for the IO. The data reads correctly from Lustre (which
1559 will overflow in a similar manner) but other file data or
1560 filesystem metadata may be corrupted in some cases.
1564 Description: Tune the kernel for good SCSI performance.
1565 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1566 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1571 Description: rewrite ext2-derived code in obdclass/uuid.c
1572 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1573 from scratch preserving functionality.
1577 Description: "lfs find -obd UUID" prints directories
1578 Details : "lfs find -obd UUID" will return all directory names instead
1579 of just file names. It is incorrect because the directories
1580 do not reside on the OSTs.
1584 Description: use i_size_read and i_size_write in 2.6 port
1585 Details : replace inode->i_size access with i_size_read/write()
1588 Frequency : when removing large files
1590 Description: scheduling issue during removal of large Lustre files
1591 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1592 It causes scheduling issues when removing large files (17TB in the
1596 Frequency : only with liblustre clients on XT3
1598 Description: evictions taking too long
1599 Details : allow llrd to evict clients directly on OSTs
1602 Frequency : only on ppc
1604 Description: /proc/fs/lustre/devices broken on ppc
1605 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1606 We should make sure the type of 'index' is loff_t and then cast
1607 explicitly as needed below. Do not assign an explicitly cast
1611 Frequency : only for rhel5
1613 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1614 Details : Modify the target file & which_kernel.
1616 Severity : enhancement
1618 Description: omit set fsid for export NFS
1619 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1620 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1624 Frequency : Always, for kernels after 2.6.16
1625 Description: Fix warning idr_remove called for id=.. which is not allocated.
1626 Details : Last kernels save old s_dev before kill super and not allow
1627 to restore from callback - restore it before call kill_anon_super.
1631 Description: buffer overruns could theoretically occur
1632 Details : llapi_semantic_traverse() modifies the "path" argument by
1633 appending values to the end of the origin string, and a
1634 overrun may occur. Adding buffer overrun check in liblustreapi.
1638 Description: Fix error on 'ls .' at the top of the Lustre mount.
1639 Details : Don't revalidate dentry if it is a root dentry.
1643 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1644 Details : Modify vm-tunables-rhel4.patch.
1648 Description: Kernel config for 2.6.18-vanilla.
1649 Details : Modify targets/2.6-vanilla.target.in.
1650 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1651 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1652 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1653 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1657 Description: improve handling recoverable errors
1658 Details : if request processig with error which can be recoverable on server
1659 request should be resend, otherwise page released from cache and
1664 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1665 Details : Modify target file & which_patch.
1666 A flaw was found in the IA32 system call emulation provided
1667 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1668 value could be stored in the %RAX register, which could trigger an
1669 out-of-bounds system call table access. An untrusted local user
1670 could exploit this flaw to run code in the kernel
1671 (ie a root privilege escalation). (CVE-2007-4573).
1675 Description: O_DIRECT bypasses client statistics.
1676 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1677 read_ahead_stats were updated. copied the stats section from
1678 osc_send_oap_rpc() into async_internal().
1682 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1686 Description: change order of libsysio includes
1687 Details : '#include sysio.h' should always come before '#include xtio.h'
1691 Description: handle "serious error: objid * already exists" more gracefully
1692 Details : If LAST_ID value on disk is smaller than the objects existing in
1693 the O/0/d* directories, it indicates disk corruption and causes an
1694 LBUG(). If the object is 0-length, then we should use the existing
1695 object. This will help to avoid a full fsck in most cases.
1697 Severity : enhancement
1699 Description: adapt the lustre_config script to support the upgrade case
1700 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1701 server targets to 1.6 in parallel.
1705 Description: To avoid grant space > avaible space when the disk is almost
1706 full. Without this patch you might see the error "grant XXXX >
1707 available" or some LBUG about grant, when the disk is almost
1709 Details : In filter_check_grant, for non_grant cache write, we should
1710 check the left space by if (*left > ungranted + bytes), instead
1711 of (*left > ungranted), because only we are sure the left space
1712 is enough for another "bytes", then the ungrant space should be
1713 increase. In client, we should update cl_avail_grant only there
1714 is OBD_MD_FLGRANT in the reply.
1718 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1719 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1720 described in CVE-2007-4573. This problem could allow untrusted
1721 local users to gain root access.
1724 Frequency : when using O_DIRECT and quotas
1726 Description: Incorrect file ownership on O_DIRECT output files
1727 Details : block usage reported by 'lfs quota' does not take into account
1728 files that have been written with O_DIRECT.
1733 Description: touch file failed when fs is not full
1734 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1735 otherwise we can get ENOSP while fs is not full.
1739 Description: parallel lock callbacks
1740 Details : Instead of sending blocking and completion callbacks as separated
1741 requests, adding them to a set and sending in parallel.
1744 Frequency : only for Cray XT3
1745 Bugzilla : 12829/13455
1746 Description: Changing primary group doesn't change the group lustre assigns to
1748 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1749 overridden with the primary group provided by the group upcall,
1750 whereas the supplied fsgid can be trusted if it is in the list of
1751 supplementary groups returned by the group upcall.
1753 Severity : enhancement
1755 Description: Allow masking D_WARNING, D_ERROR messages from console
1756 Details : Console messages can now be disabled via lnet.printk.
1760 Description: User code with malformed file open parameter crashes client node
1761 Details : Before packing join_file req, all the related reference should be
1762 checked carefully in case some malformed flags cause fake
1763 join_file req on client.
1767 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1768 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1769 mode is equal to the granted mode. However, it should also check
1770 that LDLM_FL_CP_REQD is not set.
1774 Description: Heavy nfs access might result in deadlocks
1775 Details : After ELC code landed, it is now improper to enqueue any mds
1776 locks under och_sem, because enqueue might want to decide to
1777 cancel open locks for same inode we are holding och_sem for.
1781 Description: Client eviction while running blogbench
1782 Details : A lot of unlink operations with concurrent I/O can lead to a
1783 deadlock causing evictions. To address the problem, the number of
1784 oustanding OST_DESTROY requests is now throttled to
1785 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1786 callbacks are processed in priority.
1790 Description: enable ACLs on MDS by default
1791 Details : ACLs must be enabled on MDS by default.
1794 Frequency : PPC/PPC64 only
1796 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1797 Details : fix duplicated definitions between asm-ppc64/types.h and
1798 lustre_types.h on PPC.
1801 Frequency : PPC/PPC64 only
1803 Description: asm-ppc/segment.h does not exist
1804 Details : fix compile issue on PPC.
1808 Description: better handle error messages in extents code
1811 Frequency : RHEL4 only
1813 Description: mkfs is very slow on IA64/RHEL4
1814 Details : A performance regression has been discovered in the MPT Fusion
1815 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1816 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1817 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1819 Severity : enhancement
1821 Description: SNMP support enhancement
1822 Details : Adding total number of sampled request for an MDS node in snmp
1825 Severity : enhancement
1827 Description: Optimize ldlm waiting list processing for PR extent locks
1828 Details : When processing waiting list for read extent lock and meeting read
1829 lock that is same or wider to it that is not contended, skip
1830 processing rest of the list and immediatelly return current
1831 status of conflictness, since we are guaranteed there are no
1832 conflicting locks in the rest of the list.
1836 Description: Time out and refuse to reconnect
1837 Details : When the failover node is the primary node, it is possible
1838 to have two identical connections in imp_conn_list. We must
1839 compare not conn's pointers but NIDs, otherwise we can defeat
1840 connection throttling.
1844 Description: port llog fixes from b1_6 into HEAD
1845 Details : Port llog reference couting and some llog cleanups from b1_6
1846 (bug 10800) into HEAD, for protect from panic and access to already
1847 free llog structures.
1851 Description: Detect stride IO mode in read-ahead
1852 Details : When a client does stride read, read-ahead should detect that and
1853 read-ahead pages according to the detected stride pattern.
1857 Description: data checksumming impacts single node performance
1858 Details : add support for several checksum algorithm. Currently, only CRC32
1859 and Adler-32 are supported. The checksum type can be changed on
1860 the fly via /proc/fs/lustre/osc/*/checksum_type.
1864 Description: use adler32 for page checksums
1865 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1870 Description: build for x2 fails
1871 Details : fix compile issue on Cray systems.
1875 Description: Properly match for duplicate locks
1876 Details : Due to different lock order from skiplists code, we need to
1877 traverse entire list for now
1880 Frequency : only on PPC/SLES10
1882 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1883 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1884 be defined. Add a hack in mkfs_lustre.c to work around this compile
1889 Description: LASSERT on MDS when client holding flock lock dies
1890 Details : ldlm pool logic depends on number of granted locks equal to
1891 number of released locks which is not true for flock locks, so
1892 just exclude such locks from consideration.
1896 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1897 Details : Use fsfilt_sync() for both the whole filesystem sync and
1898 individual file sync to eliminate dangerous inode locking
1899 with I_LOCK that can lead to a deadlock.
1903 Description: performance in 1.6.3
1904 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1905 sure that 1MB requests are not fragmented by the block layer.
1907 Severity : enhancement
1909 Description: organize the server-side client stats on per-nid basis
1910 Details : Change the structure of stats under obdfilter and mds to
1918 The "uuid"s file would list the uuids of _active_ exports.
1919 And the clear entry is to clear all stats and stale nids.
1921 Severity : enhancement
1923 Description: eliminate client locks in face of contention
1924 Details : file contention detection and lockless i/o implementation
1925 for contended files.
1929 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1932 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1934 Description: MDS may lose file striping (and hence file data) in some cases
1935 Details : If there are additional extended attributes stored on the MDS,
1936 in particular ACLs, SELinux, or user attributes (if user_xattr
1937 is specified for the client mount options) then there is a risk
1938 of attribute loss. Additionally, the Lustre file striping
1939 needs to be larger than default (e.g. striped over all OSTs),
1940 and an additional attribute must be stored initially in the
1941 inode and then increase in size enough to be moved to the
1942 external attribute block (e.g. ACL growing in size) for file
1947 Description: skiplist implementation simplification
1948 Details : skiplists are used to group compatible locks on granted list
1949 that was implemented as tracking first and last lock of each
1950 lock group the patch changes that to using doubly linked lists
1954 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1955 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1956 because LASSERT on both the data supplied by a client, and the
1957 data on disk is dangerous and incorrect.
1959 Severity : enhancement
1961 Description: Slow truncate/writes to huge files at high offsets.
1962 Details : Directly associate cached pages to lock that protect those pages,
1963 this allows us to quickly find what pages to write and remove
1964 once lock callback is received.
1968 Description: more ldlm soft lockups
1969 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1970 starve other threads from the resource lock for a long time in
1971 case of long waiting queue, so change the debug level from
1972 D_OTHER to the less frequently used D_INFO.
1974 Severity : enhancement
1976 Description: add -gid, -group, -uid, -user options to lfs find
1980 Description: Hung threads in invalidate_inode_pages2_range
1981 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1982 once one is completed. As a result, some RPCs are stuck in the
1983 queue and are never sent.
1987 Description: filter threads hungs on waiting journal commit
1988 Details : Cleanup filter group llog code, then only filter group llog will
1989 be only created in the MDS/OST syncing process.
1993 Description: Procfs and llog threads access destoryed import sometimes.
1994 Details : Sync the import destoryed process with procfs and llog threads by
1995 the import refcount and semaphore.
1997 Severity : enhancement
1999 Description: openlock cache of b1_6 port to HEAD
2004 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2005 Details : If insertion of an extent fails, then discard the inode
2006 preallocation and free data blocks else it can lead to duplicate
2011 Description: don't always update ctime in ext3_xattr_set_handle()
2012 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2013 In some cases the ctime should not be updated, for example for
2014 2.0->1.8 compatibility it is necessary to delete an xattr and it
2015 should not update the ctime.
2019 Bugzilla : 15713/16362
2020 Description: Assertion in iopen_connect_dentry in 1.6.3
2021 Details : looking up an inode via iopen with the wrong generation number can
2022 populate the dcache with a disconneced dentry while the inode
2023 number is in the process of being reallocated. This causes an
2024 assertion failure in iopen since the inode's dentry list contains
2025 both a connected and disconnected dentry.
2029 Description: assertion failure in ldlm_handle2lock()
2030 Details : fix a race between class_handle_unhash() and class_handle2object()
2031 introduced in lustre 1.6.5 by bug 13622.
2036 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2037 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2042 Description: oops in sd_iostats_seq_show()
2043 Details : unloading/reloading the scsi low level driver triggers a kernel
2044 bug when trying to access the sd iostat file.
2049 Description: Kernel panics during QLogic driver reload
2050 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2051 patch, causing memory corruption.
2056 Description: journal_dev option does not work in b1_6
2057 Details : pass mount option during pre-mount.
2059 Severity : enhancement
2061 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2062 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2063 extent information of a file. It can be used to map logical blocks
2064 in a file to physical blocks in the block device.
2068 Description: LDLM soft lockups - improvement
2069 Details : It is be possible to send the lock handle along with each read
2070 or write request because the client is already doing a lock match
2071 itself so there isn't any reason the OST should have to re-do that
2075 Frequency : only on Cray X2
2077 Description: X2 build failures
2078 Details : fix build failures on Cray X2.
2082 Description: xid & resent requests
2083 Details : Initialize RPC XID from clock at startup (randomly if clock is
2086 Severity : enhancement
2088 Description: Add lustre_start utility to start or stop multiple Lustre servers
2093 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2094 Details : In case of memory pressure, list_del() can be called twice on
2095 req->rq_history_list, causing a kernel oops.
2099 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2100 Details : fix stack overflow in the distributed lock manager by defering
2101 export eviction after a failed AST to the elt thread instead of
2102 handling it in the dlm interpret routine.
2106 Description: Convert some comments to new format.
2107 Details : Update documenting comments to match doxygen conventions.
2111 Description: Grammar fixes.
2112 Details : A couple of trivial spelling fixes.
2116 Description: OSD_COUNTERS-mandatory
2117 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2118 to simplify the code.
2122 Description: simplify cmm_device freeing logic.
2123 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2127 Description: Add lockdep support to dt_object_operations locking interface.
2128 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2129 parameter indicating lock ordering. Update mdd code to use new
2134 Description: Introduce failloc constants for lockless IO tests.
2135 Details : Add two new failloc constants to test lockless IO. Only one of
2136 them in implemented---another is checked in yet to be landed
2141 Description: Add lockdep support for inode mutex.
2142 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2143 in the situations where Lustre has to lock more than one inode
2148 Description: Add optional invariants checking support.
2149 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2150 configure switch. This macro is to be used for consistency and
2151 sanity checks that are too expensive to be left in `production'
2156 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2157 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2162 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2163 Details : Use spin_lock_nested() in (the only) situation where more than
2164 one ldlm_lock is locked simultaneously. Also, fix possible
2165 dead-lock in ldlm_lock_change_resource() by enforcing particular
2170 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2171 Details : Instead of specifying each ldlm_lock call-back through separate
2172 parameter, wrap them into struct ldlm_callback_suite.
2176 Description: Kill join_lru obd method and its callers.
2177 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2178 in memory---a requirement implemented through ->o_join_lru() obd
2179 method in HEAD. Remove this method and its users.
2183 Description: Add asynchronous ldlm ENQUEUE completion handler.
2184 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2185 case that stock ldlm_completion_ast() cannot handle as it waits
2186 until lock is granted. Introduce new ldlm_completion_ast_async()
2187 for this. Also comment ldlm_completion_ast().
2191 Description: ldlm_error <-> errno conversion.
2192 Details : Add functions to map (rather arbitrary) between LDLM error codes
2193 and standard errno values. CLIO needs this to prevent LDLM specific
2194 constants from escaping ldlm and osc.
2198 Description: Kill unused ldlm_handle2lock_ns() function.
2199 Details : Kill unused ldlm_handle2lock_ns() function.
2203 Description: Add lu_ref support to ldlm_lock
2204 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2205 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2206 and ldlm_resource. LDLM interface has to be changed, because code that
2207 releases a reference on a lock, has to "know" what reference this is.
2208 In the most frequent case
2210 lock = ldlm_handle2lock(handle);
2212 LDLM_LOCK_PUT(lock);
2214 no changes are required. When any other reference (received _not_ from
2215 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2216 instead of LDLM_LOCK_PUT().
2218 Arguably, changes are pervasive, and interface requires some discipline
2219 for proper use. On the other hand, it was very instrumental in finding
2220 a few leaked lock references.
2224 Description: Add ldlm_lock_addref_try().
2225 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2226 attempts to addref a lock that might be being canceled
2231 Description: Add ldlm_weigh_callback().
2232 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2233 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2234 hard-coded `number of pages' logic.
2238 Description: Add lockdep annotations to llog code.
2239 Details : Use appropriately tagged _nested() locking calls in the places
2240 where llog takes more than one ->lgh_lock lock.
2244 Description: Add loi_kms_set().
2245 Details : Wrap kms updates into a helper function.
2249 Description: Constify instances of struct lsm_operations.
2250 Details : Constify instances of struct lsm_operations.
2254 Description: lu_conf support.
2255 Details : On a server, a file system object is uniquely identified
2256 by a fid, which is sufficient to locate and load all object
2257 state (inode). On a client, on the other hand, more data are
2258 necessary instantiate an object. Change lu_object_find() and
2259 friends to take additional `lu_conf' argument describing object.
2260 Typically this includes layout information.
2264 Description: lu_context fixes.
2265 Details : Introduce new lu_context functions that are needed on the client
2266 side, where some system threads (ptlrpcd) are shared by multiple
2267 modules, and so cannot be stopped during module shutdown.
2271 Description: Add start and stop methods to lu_device_type_operations.
2272 Details : Introduce two new methods in lu_device_type_operations, that are
2273 invoked when first instance of a given type is created and last one
2274 is destroyed respectively. This is need by CLIO.
2278 Description: Add lu_ref support to struct lu_device.
2279 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2280 track leaked references.
2284 Description: Introduce lu_kmem_descr.
2285 Details : lu_kmem_descr and its companion interface allow to create
2286 and destroy a number of kmem caches at once.
2290 Description: Fix lu_object finalization race.
2291 Details : Fix a race between lu_object_find() finding an object and its
2292 concurrent finalization. This race is (most likely) not possible
2293 on the server, but might happen on the client.
2297 Description: Introduce lu_ref interface.
2298 Details : lu_ref is a debugging module allowing to track references to
2299 a given object. It is quite cpu expensive, and has to be
2300 explicitly enabled with --enable-lu_ref. See usage description
2305 Description: Factor lu_site procfs stats into a separate function.
2306 Details : Separate lu_site stats printing code into a separate function
2307 that can be reused on a client.
2311 Description: Constify instances of struct {lu,dt,md}_device_operations.
2312 Details : Constify instances of struct {lu,dt,md}_device_operations.
2316 Description: Introduce struct md_site and move meta-data specific parts of
2317 struct lu_site here.
2318 Details : Move md-specific fields out of struct lu_site into special struct
2319 md_site, so that lu_site can be used on a client.
2323 Description: Kill mdd_lov_destroy().
2324 Details : Remove unused mdd code.
2328 Description: Add st_block checking to multistat.c.
2329 Details : Add st_block checking to multistat.c.
2333 Description: Add lu_ref support to struct obd_device.
2334 Details : Add lu_ref tracking to obd_device.
2338 Description: Kill obd_set_fail_loc().
2339 Details : Remove unused code.
2343 Description: Add special type for ptlrpc_request interpret functions.
2344 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2345 there. Actual usage will be in CLIO.
2349 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2350 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2351 doesn't work with lockdep.
2355 Description: Add rwv.c test program.
2356 Details : New testing program exercising readv(2) and writev(2) (Qian).
2360 Description: Add sendfile.c test program.
2361 Details : New testing program exercising sendfile(2) (Jay).
2365 Description: Ratelimit a message that can be very frequent.
2366 Details : Ratelimit a memory allocation failure message that can
2371 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2372 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2376 Description: Liblustre build fixes.
2377 Details : Liblustre build fixes.
2381 Description: libcfs: add cfs_{need,cond}_resched() interface.
2382 Details : libcfs: add cfs_{need,cond}_resched() definition and
2383 implementations for Linux, NT, and liblustre.
2385 Severity : enhancement
2387 Description: More exported tunables for mballoc
2388 Details : Add support for tunable preallocation window and new tunables for
2389 large/small requests
2393 Description: Detect corruption of block bitmap and checking for preallocations
2394 Details : Checks validity of on-disk block bitmap. Also it does better
2395 checking of number of applied preallocations. When corruption is
2396 found, it turns filesystem readonly to prevent further corruptions.
2400 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2401 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2402 window to 0. If it does hit cache miss, and read-pattern is still
2403 stride-io mode, does not reset the stride window, but also does not
2404 increase the stride window length in this case.
2408 Frequency : only for big-endian servers
2409 Description: Check if system is big-endian while mounting fs with extents feature
2410 Details : Mounting a filesystem with extents feature will fail on big-endian
2411 systems since ext3-based ldiskfs is not supported on big-endian
2412 systems. This can be over-riden with "bigendian_extents" mount option.
2414 Severity : enhancement
2416 Description: The root squash functionality
2417 Details : A security feature, which is to prevent users from being able
2418 to mount lustre on their desktop, run as root, and delete
2419 all of the files in the filesystem. The goal is accomplished by
2420 remapping user id (UID) and group id (GID) of the root user to
2421 a UID and GID specified by the system administartor via Lustre
2422 configuration management server (MGS). The functionality also
2423 allows to specify sets of clients for which the remapping does
2428 Description: Excessive recovery window
2429 Details : With AT enabled, the recovery window can be excessively long (6000+
2430 seconds). To address this problem, we no longer use
2431 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2432 timeout no longer depends on the service time, it is set to
2433 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2434 time via pb_service_time.
2438 Description: Watchdog triggered on MDS failover
2439 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2440 the OSTs know that the MDS "UUID" can be reused for the same export
2441 from a different NID, so we do not need to wait for the export to be
2445 Frequency : rare, only if using MMP with Linux RAID
2447 Description: MMP doesn't work with Linux RAID
2448 Details : While using HA for Lustre servers with Linux RAID, it is possible
2449 that MMP will not detect multiple mounts. To make this work we
2450 need to unplug the device queue in RAID when the MMP block is being
2451 written. Also while reading the MMP block, we should read it from
2452 disk and not the cached one.
2454 Severity : enhancement
2456 Description: open file using fid
2457 Details : A file can be opened using just its fid, like
2458 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2461 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2464 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2465 Details : add proper locking for creation of jbd_4k slab cache
2469 Description: MMP check in ext3_remount() fails without displaying any error
2470 Details : When multiple mount protection fails during remount, proper error
2473 Severity : enhancement
2475 Description: Allow stripe size to be up to 4G-64k
2476 Details : Fix math logic to allow large stripe sizes.
2480 Description: add check for >8TB ldiskfs filesystems
2481 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2482 Don't allow >8TB ldiskfs filesystems to be mounted without
2483 force_over_8tb mount option
2488 Descriptoin: OSS complains "too many blocks" and crashes
2489 Details : In filter_commitrw_write():
2490 1. lquota_chkquota();
2491 2. obdo_from_inode();
2492 3. lquota_pending_commit();
2493 "2." has changed uid and gid in this case, then caused
2494 lquota_chkquota() and lquota_pending_commit() to use
2495 different uid and gid. That is the root of the bug.
2499 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2500 Details : Detect and print out the 'stripe count = -1' case properly.
2504 Description: Out or order replies might be lost on replay
2505 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2506 request with tid smaller then one currently being added, add it
2507 to the start, not end of the list.
2509 --------------------------------------------------------------------------------
2511 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2513 * Support for kernels:
2514 2.6.5-7.283 (SLES 9),
2515 2.6.9-55.EL (RHEL 4),
2516 2.6.16.46-0.14 (SLES 10),
2517 2.6.18.8 vanilla (kernel.org)
2518 * Client support for unpatched kernels:
2519 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2520 2.6.16 - 2.6.22 vanilla (kernel.org)
2521 * Due to recently discovered recovery problems, we do not recommend
2522 using patchless RHEL 4 clients with this or any earlier release.
2523 * Recommended e2fsprogs version: 1.39.cfs8
2524 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2525 * Starting with this release, the ldiskfs backing filesystem required
2526 by Lustre is now in its own package, lustre-ldiskfs. This package
2527 should be installed. It is versioned separately from Lustre and
2528 may be released separately in future.
2532 Description: block reactivating mgc import until all deactivates complete
2533 Details : Fix race when failing back MDT/MGS to itself (testing)
2536 Frequency : at statup only
2538 Description: mds_lov_synchronize race leads to various problems
2539 Details : simultaneous MDT->OST connections at startup can cause the
2540 sync to abort, leaving the OSC in a bad state.
2542 Severity : enhancement
2544 Description: add optional extra BUILD_VERSION info
2545 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2546 to override the lustre version.
2549 Frequency : 2.6.18 servers only
2551 Description: ll_kern_mount() doesn't release the module reference
2552 Details : The ldiskfs module reference count never drops down to 0
2553 because ll_kern_mount() doesn't release the module reference.
2558 Description: server LBUG when using old ost_num_threads parameter
2559 Details : Accept the old ost_num_threads parameter but warn that it
2560 is deprecated, and fix an off-by-one error that caused an LBUG.
2565 Description: Transient SCSI error results in persistent IO issue
2566 Details : iobuf->dr_error is not reinitialized to 0 between two
2570 Frequency : sometimes when underlying device returns I/O errors
2572 Description: OSTs not going read-only during write failures
2573 Details : OSTs are not remounted read-only when the journal commit threads
2574 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2575 instead of the ext3 wrappers.
2579 Description: poor connect scaling with increasing client count
2580 Details : Don't run filter_grant_sanity_check for more than 100 exports
2581 to improve scaling for large numbers of clients.
2584 Frequency : SLES10 only
2586 Description: sanity-quota.sh quotacheck failed: rc = -22
2587 Details : Quotas cannot be enabled on SLES10.
2590 Frequency : liblustre clients only
2592 Description: getdirentries does not give error when run on compute nodes
2593 Details : getdirentries does not fail when the size specified as an argument
2594 is too small to contain at least one entry
2596 Severity : enhancement
2598 Description: Add LNET router traceability for debug purposes
2599 Details : If a checksum failure occurs with a router as part of the
2600 IO path, the NID of the last router that forwarded the bulk data
2601 is printed so it can be identified.
2606 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2607 Details : Due to a race condition, liblustre clients were occasionally
2608 evicted incorrectly.
2610 Severity : enhancement
2612 Description: lfs setstripe use optional parameters instead of postional
2615 Severity : enhancement
2617 Description: Nanosecond timestamp support for ldiskfs
2618 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2619 resolution timestamps. There is not yet support for this at
2620 the Lustre filesystem level.
2623 Frequency : during server recovery
2625 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2626 Details : request with rq_no_resend flag not awake l_wait_event if they get
2630 Frequency : nfs export on patchless client
2632 Description: connectathon hang when test nfs export over patchless client
2633 Details : Disconnected dentry cannot be found with lookup, so we do not need
2634 to unhash it or make it invalid
2637 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2638 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2639 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2644 Description: open req refcounting wrong on reconnect
2645 Details : If reconnect happened between getting open reply from server and
2646 call to mdc_set_replay_data in ll_file_open, we will schedule
2647 replay for unreferenced request that we are about to free.
2648 Subsequent close will crash in variety of ways.
2649 Check that request is still eligible for replay in
2650 mdc_set_replay_data().
2655 Description: disable writes to filesystem when reading health_check file
2656 Details : the default for reading the health_check proc file has changed
2657 to NOT do a journal transaction and write to disk, because this
2658 can cause reads of the /proc file to hang and block HA state
2659 checking on a healthy but otherwise heavily loaded system. It
2660 is possible to return to the previous behaviour during configure
2661 with --enable-health-write.
2663 Severity : enhancement
2665 Description: 64-bit inode version
2666 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2667 to the inode. This will be required for version-based recovery.
2672 Description: MDS fails to start if a duplicate client export is detected
2673 Details : in some rare cases it was possible for a client to connect to
2674 an MDS multiple times. Upon recovery the MDS would detect this
2675 and fail during startup. Handle this more gracefully.
2677 Severity : enhancement
2679 Description: Add -o localflock option to simulate old noflock
2681 Details : This will achieve local-only flock/fcntl locks
2687 Description: log_commit_thread vs filter_destroy race leads to crash
2688 Details : Take import reference before releasing llog record semaphore
2693 Description: Wrong request locking in request set processing
2694 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2695 imp_delayed_list, in this place should be used imp_lock.
2698 Frequency : when reconnection
2700 Description: Grant Leak when osc reconnect to OST
2701 Details : When osc reconnect ost, OST(filter) should check whether it
2702 should grant more space to client by comparing fed_grant and
2703 cl_avail_grant, and return the granted space to client instead
2704 of "new granted" space, because client will call osc_init_grant
2705 to update the client grant space info.
2708 Frequency : when client reconnect to OST
2710 Description: Grant Leak when osc do resend and replay bulk write
2711 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2712 bulk write request, because the grant info while be sync between
2713 OSC and OST when reconnect, and we should ignore the grant info
2714 these of resend/replay write req.
2719 Description: Grant space more than avaiable left space sometimes.
2720 Details : When then OST is about to be full, if two bulk writing from
2721 different clients came to OST. Accord the avaliable space of the
2722 OST, the first req should be permitted, and the second one
2723 should be denied by ENOSPC. But if the seconde arrived before
2724 the first one is commited. The OST might wrongly permit second
2725 writing, which will cause grant space > avaiable space.
2728 Frequency : when client is evicted
2730 Description: Grant might be wrongly erased when osc is evicted by OST
2731 Details : when the import is evicted by server, it will fork another
2732 thread ptlrpc_invalidate_import_thread to invalidate the
2733 import, where the grant will be set to 0. While the original
2734 thread will update the grant it got when connecting. So if
2735 the former happened latter, the grant will be wrongly errased
2736 because of this race.
2741 Description: Checking Stale with correct fid
2742 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2743 to check whether it is stale, because sometimes, we want the
2744 enqueue happened anyway, and op_data.fid2 will not be initialized.
2746 Severity : enhancement
2748 Description: update patchless client
2749 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2752 Frequency : only with 2.4 kernel
2754 Description: random memory corruption
2755 Details : size of struct ll_inode_info is to big for union inode.u and this
2756 can be cause of random memory corruption.
2761 Description: Memory leak in recovery
2762 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2763 It should also check obd_fail before fsfilt_start, otherwise if
2764 fsfilt_start return -EROFS,(failover mds during mds recovery).
2765 then the req will return with repmsg->transno = 0 and rc = EROFS.
2766 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2767 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2768 be freed no matter whether fsfilt_commit success or not.
2771 Frequency : only with huge count clients
2773 Description: Prevents from taking the superblock lock in llap_from_page for
2775 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2776 for llap_from_page call in ll_removepage prevents from taking the
2777 superblock lock for a soon died page.
2782 Description: Not check open intent error before release open handle
2783 Details : in some rare cases, the open intent error is not checked before
2784 release open handle, which may cause
2785 ASSERTION(open_req->rq_transno != 0), because it tries to release
2786 the failed open handle.
2791 Description: Set cat log bitmap only after create log success.
2792 Details : in some rare cases, the cat log bitmap is set too early. and it
2793 should be set only after create log success.
2797 Description: Accessing a block bevice can re-enable I/O when Lustre is
2798 tearing down a device.
2799 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2804 Description: service threads may hog cpus when there are a lot of requests
2805 Details : Insert cond_resched to give other threads a chance to use some CPU
2810 Description: the cat log was not initialized in recovery
2811 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2812 unlink log on mds will not be initialized until mds post
2813 recovery. And also in mds post recovery, the unlink log will
2814 initialization will be done asynchronausly, so there will be race
2815 between add unlink log and unlink log initialization.
2819 Description: brw_stats were being printed incorrectly
2820 Details : brw_stats were being printed as log2 but all of them were not
2821 recorded as log2. Also remove some code duplication arising from
2822 filter_tally_{read,write}.
2826 Frequency : rare, only in recovery.
2827 Description: ASSERTION(req->rq_type != LI_POISON) failed
2828 Details : imp_lock should be held while iterating over imp_sending_list for
2829 prevent destroy request after get timeout in ptlrpc_queue_wait.
2833 Description: replay-single.sh test 52 fails
2834 Details : A lock's skiplist need to be cleanup when it being unlinked
2835 from its resource list.
2839 Description: Short directio read returns full requested size rather than
2841 Details : Direct I/O operations should return actual amount of bytes
2842 transferred rather than requested size.
2844 Severity : enhancement
2846 Description: metadata RPC reduction (e.g. for rm performance)
2847 Details : decrease the amount of synchronous RPC between clients and servers
2848 by canceling conflicing lock before the operation on the client
2849 and packing thier handles into the main operation RPC to server.
2851 Severity : enhancement
2853 Description: Async OSC create to avoid the blocking unnecessarily.
2854 Details : If a OST has no remain object, system will block on the creating
2855 when need to create a new object on this OST. Now, ways use
2856 pre-created objects when available, instead of blocking on an
2857 empty osc while others are not empty. If we must block, we block
2858 for the shortest possible period of time.
2862 Description: improve handling recoverable errors
2863 Details : if request processig with error which can be recoverable on server
2864 request should be resend, otherwise page released from cache and
2867 Severity : enhancement
2869 Description: refine locking for avoid write wrong info into lov_objid file
2870 Details : fix possible races with add new target and write/update data in
2873 --------------------------------------------------------------------------------
2875 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2880 Frequency : on some architectures
2882 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2883 Details : Uninitialized flags sometimes cause configuration commands to
2887 Frequency : patchless clients only
2889 Description: missing __iget() symbol export
2890 Details : The __iget() symbol export is missing. To avoid the need for
2891 this on patchless clients the deathrow inode reaper is turned
2892 off, and we depend on the VM to clean up old inodes. This
2893 dependency was during via the fix for bug 12181.
2895 --------------------------------------------------------------------------------
2897 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2899 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2900 INTEROPERATE with older versions automatically. In many cases a
2901 special upgrade step is needed. Please read the
2902 user documentation before upgrading any part of a live system.
2903 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2904 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2905 * WARNING: Lustre configuration and startup changes are required with
2906 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2910 Severity : enhancement
2912 Description: MountConf
2913 Details : Lustre configuration is now managed via mkfs and mount
2914 commands instead of lmc and lconf. New obd types (MGS, MGC)
2915 are added for dynamic configuration management. See
2916 https://mail.clusterfs.com/wikis/lustre/MountConf for
2919 Severity : enhancement
2921 Description: dynamic OST addition
2922 Details : OSTs can now be added to a live filesystem
2924 Severity : enhancement
2926 Description: startup order invariance
2927 Details : MDTs and OSTs can be started in any order. Clients only
2928 require the MDT to complete startup.
2930 Severity : enhancement
2932 Description: parallel, asynchronous orphan cleanup
2933 Details : orphan cleanup is now performed in separate threads for each
2934 OST, allowing parallel non-blocking operation.
2936 Severity : enhancement
2938 Description: optimized stripe assignment
2939 Details : stripe assignments are now made based on ost space available,
2940 ost previous usage, and OSS previous usage, in order to try
2941 to optimize storage space and networking resources.
2943 Severity : enhancement
2945 Description: Permanently set tunables
2946 Details : All writable /proc/fs/lustre tunables can now be permanently
2947 set on a per-server basis, at mkfs time or on a live system.
2949 Severity : enhancement
2951 Description: Lustre message v2
2952 Details : Add lustre message format v2.
2954 Severity : enhancement
2956 Description: client OST exclusion list
2957 Details : Clients can be started with a list of OSTs that should be
2958 declared "inactive" for known non-responsive OSTs.
2962 Description: SPEC SFS validation failure on NFS v2 over lustre.
2963 Details : Changes the blocksize for regular files to be 2x RPC size,
2964 and not depend on stripe size.
2966 Severity : enhancement
2968 Description: Multiple MD RPCs in flight.
2969 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2970 To avoid overly-overloading MDS, introduce a limit on number of
2971 MDS RPCs in flight for a single client and add /proc controls
2972 to adjust this limit.
2974 Severity : enhancement
2976 Description: client read/write statistics
2977 Details : Add client read/write call usage stats for performance
2978 analysis of user processes.
2979 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2980 file access. extents_stats shows chunk size distribution.
2981 extents_stats_per_process show chunk size distribution per
2984 Severity : enhancement
2986 Description: mds statistics
2987 Details : Add detailed mds operations statistics in
2988 /proc/fs/lustre/mds/*/stats.
2992 Description: Failure of copying files with lustre special EAs.
2993 Details : Client side always return success for setxattr call for lustre
2994 special xattr (currently only "trusted.lov").
2999 Description: Refcount LNET uuids
3000 Details : The global LNET uuid list grew linearly with every startup;
3001 refcount repeated list entries instead of always adding to
3004 Severity : enhancement
3006 Description: Dynamic service threads
3007 Details : Within a small range, start extra service threads
3008 automatically when the request queue builds up.
3011 Frequency : mixed-endian client/server environments
3013 Description: mixed-endian crashes
3014 Details : The new msg_v2 system had some failures in mixed-endian
3017 Severity : enhancement
3019 Description: Easy OST removal
3020 Details : OSTs can be permanently deactivated with e.g. 'lctl
3021 conf_param lustre-OST0001.osc.active=0'
3023 Severity : enhancement
3025 Description: MGS proc entries
3026 Details : Added basic proc entries for the MGS showing what filesystems
3029 Severity : enhancement
3031 Description: provide MGS failover
3032 Details : Added config lock reacquisition after MGS server failover.
3034 Severity : enhancement
3036 Description: add Linux 2.4 support
3037 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3041 Description: a large application tries to do I/O to the same resource and dies
3042 in the middle of it.
3043 Details : Check the req->rq_arrival time after the call to
3044 ost_brw_lock_get(), but before we do anything about
3045 processing it & sending the BULK transfer request. This
3046 should help move old stale pending locks off the queue as
3047 quickly as obd_timeout.
3050 Frequency : when an incorrect nid is specified during startup
3052 Description: ptlrpc connect to non-existant node causes kernel crash
3053 Details : LNET can't be re-entered from an event callback, which
3054 happened when we expire a message after the export has been
3055 cleaned up. Instead, hand the zombie cleanup off to another
3058 Severity : enhancement
3060 Description: plain/inodebits lock performance improvement
3061 Details : Grouping plain/inodebits in granted list by their request modes
3062 and bits policy, thus improving the performance of search through
3066 Frequency : only if OST filesystem is corrupted
3068 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3069 Details : for a short time RPCs with bulk IO are in the replay list,
3070 but replay of bulk IOs is unimplemented. If the OST filesystem
3071 is corrupted due to disk cache incoherency and then replay is
3072 started it is possible to trip an assertion. Avoid putting
3073 committed RPCs into the replay list at all to avoid this issue.
3076 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3078 Description: System hang on startup
3079 Details : This bug allowed the liblustre (e.g. catamount) client to
3080 return to the app before handling all startup RPCs. This
3081 could leave the node unresponsive to lustre network traffic
3082 and manifested as a server ptllnd timeout.
3084 Severity : enhancement
3086 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3087 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3088 boolean to control whether to print peer debug info when a
3089 client's RPC times out.
3092 Frequency : only for kernels with patches from Lustre below 1.4.3
3094 Description: Remove old rdonly API
3095 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3098 Frequency : only for devices with external journals
3100 Description: Set external device read-only also
3101 Details : During a commanded failover stop, we set the disk device
3102 read-only while the server shuts down. We now also set any
3103 external journal device read-only at the same time.
3106 Frequency : when upgrading from 1.4 while trying to change parameters
3108 Description: The wrong (new) MDC name was used when setting parameters for
3109 upgraded MDT's. Also allows changing of OSC (and MDC)
3110 parameters if --writeconf is specified at tunefs upgrade time.
3113 Frequency : when setting specific ost indicies
3115 Description: QOS code breaks on skipped indicies
3116 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3117 created with --index need not be sequential.
3119 Severity : enhancement
3121 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3122 Details : The uninit_groups feature works in conjunction with the kernel
3123 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3124 up the pass1 processing of e2fsck. This is a read-only feature
3125 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3126 mount filesystems that have had this feature enabled.
3128 Severity : enhancement
3130 Description: Improve multi-block allocation algorithm to avoid fragmentation
3131 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3132 allocation locality and avoid filesystem fragmentation.
3134 ------------------------------------------------------------------------------
3136 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3138 * Support for kernels:
3139 2.4.21-47.0.1.EL (RHEL 3)
3140 2.6.5-7.283 (SLES 9)
3141 2.6.9-42.0.10.EL (RHEL 4)
3142 2.6.12.6 vanilla (kernel.org)
3143 2.6.16.27-0.9 (SLES 10)
3144 * Recommended e2fsprogs version: 1.39.cfs5
3146 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3150 Frequency : occasional, depends on client load and configuration
3151 Bugzilla : 12181, 12203
3152 Description: data loss for recently-modified files
3154 Details : In some cases it is possible that recently written or created
3155 files may not be written to disk in a timely manner (this should
3156 normally be within 30s unless client IO load is very high).
3157 The problem appears as zero-length files or files that are a
3158 multiple of 1MB in size after a client crash or client eviction
3159 that are missing data at the end of the file.
3161 This problem is more likely to be hit on clients where files are
3162 repeatedly created and unlinked in the same directory, clients
3163 have a large amount of RAM, have many CPUs, the filesystem has
3164 many OSTs, the clients are rebooted frequently, and/or the files
3165 are not accessed by other nodes after being written.
3167 The presence of the problem can be detected by looking at
3168 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3169 smaller than the second (nr_unused) then dirty files will not
3170 be flushed automatically to disk. "sync; sleep 10" should be
3171 run several times on the node before unmounting it to update
3172 Lustre (this is also safe to run on nodes without this problem).
3174 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3175 that can cause this same problem, so customers using that kernel
3176 also need to update the kernel in addition to Lustre. In order
3177 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3179 It is normal that files written just before a client crash (less
3180 than 30s) may not yet have been flushed to disk, even for local
3184 Frequency : frequent on thin XT3 nodes
3186 Description: UUID collision on thin XT3 Linux nodes
3187 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3188 randomly, since we relied on an insufficiently-seeded PRNG.
3193 Description: OSS hangs after "All ost request buffers busy"
3194 Details : A deadlock between quota and journal operations caused OSS
3195 hangs after printing "All ost request buffers busy."
3198 Frequency : always on liblustre builds
3200 Description: Cleanup compiler warnings on liblustre
3203 Frequency : always on liblustre builds on XT3
3205 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3209 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3210 > 32000 subdirectories
3211 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3212 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3213 subdirectory count crosses 32000. This will aid e2fsck to
3214 correctly handle more than 32000 subdirectories.
3217 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3219 Description: System hang on startup
3220 Details : This bug allowed the liblustre (e.g. catamount) client to
3221 return to the app before handling all startup RPCs. This
3222 could leave the node unresponsive to lustre network traffic
3223 and manifested as a server ptllnd timeout.
3225 Severity : enhancement
3227 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3228 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3229 boolean to control whether to print peer debug info when a
3230 client's RPC times out.
3235 Description: make O_SYNC working on 2.6 kernels
3236 Details : 2.6 kernels use different method for mark pages for write,
3237 so need add a code to lustre for O_SYNC work.
3242 Description: Failure to close file and release space on NFS
3243 Details : Put inode details into lock acquired in ll_intent_file_open.
3244 Use mdc_intent_lock in ll_intent_open to properly
3245 detect all kind of errors unhandled by mdc_enqueue.
3250 Description: proc file read during shutdown sometimes raced obd removal,
3252 Details : Add lock to prevent obd access after proc file removal.
3255 Frequency : Only for files larger than 4GB on 32-bit clients.
3257 Description: improperly doing page alignment of locks
3258 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3259 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3264 Description: RPCs being resent when they shouldn't be
3265 Details : Some RPCs that should not be resent are being resent. This
3266 can cause inconsistencies in the RPC state machine. Do not
3267 resend such requests.
3270 Frequency : rare, only with NFS export
3272 Description: Crash on NFS re-export node
3273 Details : under very unusual load conditions an assertion is hit in
3274 ll_intent_file_open()
3277 Frequency : only if OST filesystem is corrupted
3279 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3280 Details : for a short time RPCs with bulk IO are in the replay list,
3281 but replay of bulk IOs is unimplemented. If the OST filesystem
3282 is corrupted due to disk cache incoherency and then replay is
3283 started it is possible to trip an assertion. Avoid putting
3284 committed RPCs into the replay list at all to avoid this issue.
3289 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3290 Details : Large single O_DIRECT read and write calls can fail to allocate
3291 a sufficiently large buffer to process the request. In case of
3292 allocation failure the allocation is retried with a smaller
3293 buffer and broken into smaller requests.
3295 Severity : enhancement
3297 Description: Add -o localflock option to simulate old noflock behaviour.
3298 Details : This will achieve local-only flock/fcntl locks coherentness.
3303 Description: versioning check is incomplete
3304 Details : Checking the version difference of client vs. server, report
3305 error if the gap is too big.
3310 Description: add support PG_writeback bit
3311 Details : add support for PG_writeback bit for Lustre, for more carefull
3312 work with page cache in 2.6 kernel. This also fix some deadlocks
3313 and remove hack for work O_SYNC with 2.6 kernel.
3315 Severity : enhancement
3317 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3318 Details : The uninit_groups feature works in conjunction with the kernel
3319 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3320 up the pass1 processing of e2fsck. This is a read-only feature
3321 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3322 mount filesystems that have had this feature enabled.
3324 Severity : enhancement
3326 Description: Improve multi-block allocation algorithm to avoid fragmentation
3327 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3328 allocation locality and avoid filesystem fragmentation.
3330 ------------------------------------------------------------------------------
3332 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3334 * Support for kernels:
3335 2.6.9-42.0.3.EL (RHEL 4)
3336 2.6.5-7.276 (SLES 9)
3337 2.4.21-47.0.1.EL (RHEL 3)
3338 2.6.12.6 vanilla (kernel.org)
3339 2.6.16.21-0.8 (SLES10)
3340 * Recommended e2fsprogs version: 1.39.cfs2-0
3342 * The backwards-compatible /proc/sys/portals symlink has been removed
3343 in this release. Before upgrading, please ensure that you change
3344 any configuration scripts or /etc/sysctl.conf files that access
3345 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3346 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3347 in advance of the upgrade on any system running Lustre 1.4.6 or
3348 newer, since /proc/sys/lnet was added in that version.
3349 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3353 Frequency : only when quota is used
3355 Description: avoid scanning export list for quota master
3356 Details : Change the algorithms to avoid scanning export list in order
3357 to improve the efficiency.
3360 Frequency : MDS failover only, very rarely
3362 Description: "went back in time" messages on mds failover
3363 Details : The greatest transno may be lost when the current operation
3364 finishes with an error (transno==0) and the client's last_rcvd
3365 record is over-written. Save the greatest transno in the
3366 mds_last_transno for this case.
3369 Frequency : always for specific kernels and striping counts
3371 Description: client may get "Matching packet too big" without ACL support
3372 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3373 when trying to access files in certain configurations. The
3374 clients should in fact be denied when mounting because they do
3375 not understand ACLs.
3378 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3380 Description: many clients connecting with IO in progress causes connect timeouts
3381 Details : Avoid synchronous journal commits to avoid delays caused by many
3382 clients connecting/disconnecting when bulk IO is in progress.
3383 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3384 OST_IO_PORTAL to avoid delays behind potentially many pending
3388 Frequency : occasionally with multiple writers to a single file
3390 Description: shared writes to file may result in wrong size reported by stat()
3391 Details : Allow growing of kms when extent lock is cancelled
3394 Frequency : always with random mmap IO to multi-striped file
3396 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3397 Details : Only if the hole is at the end of OST object so that kms is too
3398 small. Fix is to increase kms accordingly in ll_nopage.
3401 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3403 Description: writes to a missing object would leak memory on the OST
3404 Details : If there is an inconsistency between the MDS and OST filesystems,
3405 such that the MDS references an object that doesn't exist, writes
3406 to that object will leak memory due to incorrect cleanup in the
3407 error handling path, eventually running out of memory on the OST.
3412 Description: Creating too long symlink causes lustre errors
3413 Details : Check symlink and name lengths before sending requests to MDS.
3416 Frequency : only if flock is enabled (not on by default)
3418 Description: posix locks not released on fd closure on 2.6.9+
3419 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3420 kernels, this caused such locks not to be released on fd close and
3421 then assertions on fs unmount about still used locks.
3424 Frequency : MDS failover only, very rarely
3426 Description: clients may get ASSERTION(granted_lock != NULL)
3427 Details : When request was taking a long time, and a client was resending
3428 a getattr by name lock request. The were multiple lock requests
3429 with the same client lock handle and
3430 mds_getattr_name->fixup_handle_for_resent_request found one of the
3431 lock handles but later failed with ASSERTION(granted_lock != NULL).
3436 Description: handle->h_buffer_credits > 0, assertion failure
3437 Details : h_buffer_credits is zero after truncate, causing assertion
3438 failure. This patch extends the transaction or creates a new
3442 Frequency : NFS re-export or patchless client
3443 Bugzilla : 11179, 10796
3444 Description: Crash on NFS re-export node (__d_move)
3445 Details : We do not want to hash the dentry if we don't have a lock.
3446 But if this dentry is later used in d_move, we'd hit uninitialised
3447 list head d_hash, so we just do this to init d_hash field but
3448 leave dentry unhashed.
3451 Frequency : NFS re-export or patchless client
3453 Description: NFS exports has problem with symbolic link
3454 Details : lustre client didn't properly install dentry when re-exported
3455 to NFS or running patchless client.
3458 Frequency : NFS re-export or patchless client
3460 Description: Various nfs/patchless fixes.
3461 Details : fixes reuse disconected alias for lookup process - this fixes
3462 warning "find_exported_dentry: npd != pd",
3463 fix permission error with open files at nfs.
3464 fix apply umask when do revalidate.
3467 Frequency : occasional
3469 Description: Crash on NFS re-export node
3470 Details : calling clear_page() on the wrong pointer triggered oops in
3471 generic_mapping_read().
3474 Frequency : rarely, using O_DIRECT IO
3476 Description: unaligned directio crashes client with LASSERT
3477 Details : check for unaligned buffers before trying any requests.
3480 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3482 Description: stale data returned from RAID cache
3483 Details : If only a small amount of IO is done to the RAID device before
3484 reading it again it is possible to get stale data from the RAID
3485 cache instead of reading it from disk.
3488 Frequency : always for sles10 kernel
3490 Description: sles10 support
3491 Details : ll_follow_link: compile fixes and using of nd_set_link
3492 under newer kernels.
3495 Frequency : depends on arch, kernel and compiler version, always on sles10
3498 Description: recursive or deep enough symlinks cause stack overflow
3499 Details : getting rid of large stack-allocated variable in
3503 Frequency : depends on hardware
3505 Description: lustre write performance loss in the SLES10 kernel
3506 Details : the performance loss is caused by using of write barriers in the
3507 ext3 code. The SLES10 kernel turns barrier support on by
3508 default. The fix is to undo that change for ldiskfs.
3510 ------------------------------------------------------------------------------
3512 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3514 * Support for kernels:
3515 2.6.9-42.0.3EL (RHEL 4)
3516 2.6.5-7.276 (SLES 9)
3517 2.4.21-47.0.1.EL (RHEL 3)
3518 2.6.12.6 vanilla (kernel.org)
3522 Frequency : quota enabled and large files being deleted
3524 Description: releasing more than 4GB of quota at once hangs OST
3525 Details : If a user deletes more than 4GB of files on a single OST it
3526 will cause the OST to spin in an infinite loop. Release
3527 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3532 Description: statfs data retrieved from /proc may be stale or zero
3533 Details : When reading per-device statfs data from /proc, in the
3534 {kbytes,files}_{total,free,avail} files, it may appear
3535 as zero or be out of date.
3538 Frequency : systems with MD RAID1 external journal devices
3540 Description: lconf's call to blkid is confused by RAID1 journal devices
3541 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3542 of returning all block devices that match the journal UUID.
3545 Frequency : always, for aggregate stripe size over 4GB
3547 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3548 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3549 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3552 Frequency : always if "lfs find" used on a local file/directory
3554 Description: "lfs find" segfaults if used on a local file/directory
3555 Details : The case where a directory component was not specified wasn't
3556 handled correctly. Handle this properly.
3559 Frequency : always on ppc64
3561 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3562 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3563 perform endianness conversion.
3566 Frequency : rarely (truncate to non-zero file size after write under load)
3567 Bugzilla : 10730, 10687
3568 Description: Files padded with zeros to next 4K multiple
3569 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3570 it is possible that files that are truncated to a non-zero size
3571 immediately after being written are filled with zero bytes beyond
3572 the truncated size. No file data is lost.
3574 Severity : enhancement
3576 Description: Allow recovery/failover for liblustre clients.
3577 Details : liblustre clients were unaware of failover configurations until
3580 Severity : enhancement
3582 Description: user file locks should fail when not mounting with flock option
3583 Details : Set up an error-returning stub in ll_file_operations.lock field
3584 to prevent incorrect behaviour when client is mounted without
3585 flock option. Also, set up properly f_op->flock field for
3589 Frequency : always on ia64
3591 Description: "lfs df" loops on printing out MDS statfs information
3592 Details : The obd_ioctl_data was not initialized and in some systems
3593 this caused a failure during the ioctl that did not return
3594 an error. Initialize the struct and return an error on failure.
3597 Frequency : SLES 9 only
3599 Description: Error of copying files with lustre special EAs as root
3600 Details : Client side always return success for setxattr call for lustre
3601 special xattr (currently only "trusted.lov").
3604 Frequency : rarely on clusters with both ia64+i386 clients
3606 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3607 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3608 clients are doing concurrent writes to the same file it is
3609 possible that the ia64 clients may LASSERT because the OST
3610 extent locks are not PAGE_SIZE aligned. Ensure that grown
3611 locks are always aligned on the request boundary.
3614 Frequency : specific use, occasional
3616 Description: Overwriting in use executable truncates on-disk binary image
3617 Details : If one node attempts to overwrite an executable in use by
3618 another node, we now correctly return ETXTBSY instead of
3619 truncating the file.
3621 Severity : enhancement
3623 Description: Async OSC create to avoid the blocking unnecessarily.
3624 Details : If a OST has no remain object, system will block on the creating
3625 when need to create a new object on this OST. Now, ways use
3626 pre-created objects when available, instead of blocking on an
3627 empty osc while others are not empty. If we must block, we block
3628 for the shortest possible period of time.
3633 Description: chmod on Lustre root is propagated to other clients
3634 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3635 invalid by the follow_mount time.
3638 Frequency : liblustre clients only
3640 Description: Race in 'instant cancel' lock handling could lead to such locks
3641 never to be granted in case of SMP MDS
3642 Details : Do not destroy not yet granted but cbpending locks in
3646 Frequency : replay/resend of open
3648 Description: non null lock assetion failure in mds_intent_policy
3649 Details : Trying to replay/resend lockless open requests resulted in
3650 mds_open() returning 0 with no lock. Now it sets a flag if
3651 a lock is going to be returned.
3653 Severity : enhancement
3655 Description: Checksum enhancements
3656 Details : New checksum enhancements allow for resending RPCs that failed
3659 Severity : enhancement
3661 Description: Tunables on number of dirty pages in cacche
3662 Details : Allow to set limit on number of dirty pages cached.
3667 Description: client crash on unmount - lock still has references
3668 Details : In some error handling cases it was possible to leak a lock
3669 reference on a client while accessing a file. This was not
3670 harmful to the client during operation, but would cause the
3671 client to crash when the filesystem is unmounted.
3674 Frequency : specific case, rare
3676 Description: ETXTBSY on mds though file not in use by client
3677 Details : ETXTBSY is no longer incorrectly returned when attempting to
3678 chmod or chown a directory that the user previously tried to
3679 execute or a currently-executing binary.
3682 Frequency : extremely rare except on liblustre-based clients
3684 Description: Lustre space not freed when files are deleted
3685 Details : Clean up open-unlinked files after client eviction. Previously
3686 the unlink was skipped and the files remained as orphans.
3691 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3692 Details : In some cases it was possible to send a blocking callback to
3693 a client doing a glimpse, even though that client didn't get
3694 a lock granted. When the glimpse lock is cancelled on the OST
3695 the freed lock is left on the waiting list and corrupted the list.
3698 Frequency : all core dumps
3700 Description: Broke core dumps to lustre
3701 Details : Negative dentry may be unhashed if parent does not have UPDATE
3702 lock, but some callers, e.g. do_coredump, expect dentry to be
3703 hashed after successful create, hash it in ll_create_it.
3705 ------------------------------------------------------------------------------
3707 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3709 * Support for kernels:
3710 2.6.9-42.0.2.EL (RHEL 4)
3711 2.6.5-7.276 (SLES 9)
3712 2.4.21-40.EL (RHEL 3)
3713 2.6.12.6 vanilla (kernel.org)
3717 Frequency : always on RHEL 3
3719 Description: Number of open files grows over time
3720 Details : The number of open files grows over time, whether or not
3721 Lustre is started. This was due to a filp leak introduced
3722 by one of our kernel patches.
3724 ------------------------------------------------------------------------------
3726 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3728 * Support for kernels:
3729 2.6.9-42.EL (RHEL 4)
3730 2.6.5-7.276 (SLES 9)
3731 2.4.21-40.EL (RHEL 3)
3732 2.6.12.6 vanilla (kernel.org)
3737 Bugzilla : 5719, 9635, 9792, 9684
3738 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3739 Details : If a server is under heavy load and cannot reply to new
3740 connection requests before the client resends the (re)connect,
3741 the connection handling code can behave badly if two service
3742 threads are concurrently handing separate (re)connections from
3743 the same client. Add better locking to the connection handling
3744 code, and ensure that only a single connection will be processed
3745 for a given client UUID, even if the lock is dropped.
3747 Severity : enhancement
3749 Description: add TCP zero-copy support to kernel
3750 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3751 sends if the hardware supports scatter-gather and checksumming.
3752 This allows socklnd to do client-write and server-read more
3753 efficiently and reduce CPU utilization from skbuf copying.
3756 Frequency : only if NFS exporting from client
3758 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3759 Details : If setting attributes on a file created under NFS that had
3760 never been opened it would be possible to oops the client
3761 if the file had no objects.
3764 Frequency : always for liblustre
3766 Description: liblustre client does MDS+OSTs setattr RPC for each write
3767 Details : When doing a write from a liblustre client, the client
3768 incorrectly issued an RPC to the MDS and each OST the file was
3769 striped over in order to update the timestamps. When writing
3770 with small chunks and many clients this could overwhelm the MDS
3771 with RPCs. In all cases it would slow down the write because
3772 these RPCs are unnecessary.
3774 Severity : enhancement
3776 Description: allow number of MDS service threads to be changed at module load
3777 Details : It is now possible to change the number of MDS service threads
3778 running. Adding "options mds mds_num_threads={N}" to the MDS's
3779 /etc/modprobe.conf will set the number of threads for the next
3780 time Lustre is restarted (assuming the "mds" module is also
3781 reloaded at that time). The default number of threads will
3782 stay the same, 32 for most systems.
3787 Description: OST crash if filesystem is unformatted or corrupt
3788 Details : If an OST is started on a device that has never been formatted
3789 or if the filesystem is corrupt and cannot even mount then the
3790 error handling cleanup routines would dereference a NULL pointer.
3795 Description: NULL pointer deref in llap_from_page.
3796 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3797 incorrect) mapping if the page was truncated/reclaimed while it was
3798 searched for. Check for this condition and skip such pages when
3799 doing readahead. Introduce extra check to llap_from_page() to
3800 verify page->mapping->host is non-NULL (so page is not anonymous).
3803 Frequency : Sometimes when using sys_sendfile
3805 Description: "page not covered by a lock" warnings from ll_readpage
3806 Details : sendfile called ll_readpage without right page locks present.
3807 Now we introduced ll_file_sendfile that does necessary locking
3808 around call to generic_file_sendfile() much like we do in
3812 Frequency : with certain MDS communication failures at client mount time
3814 Description: NULL pointer deref after failed client mount
3815 Details : a client connection request may delayed by the network layer
3816 and not be sent until after the PTLRPC layer has timed out the
3817 request. If the client fails the mount immediately it will try
3818 to clean up before the network times out the request. Add a
3819 reference from the request import to the obd device and delay
3820 the cleanup until the network drops the request.
3823 Frequency : occasionally during client (re)connect
3825 Description: assertion failure during client (re)connect
3826 Details : processing a client connection request may be delayed by the
3827 client or server longer than the client connect timeout. This
3828 causes the client to resend the connection request. If the
3829 original connection request is replied in this interval, the
3830 client may trip an assertion failure in ptlrpc_connect_interpret()
3831 which thought it would be the only running connect process.
3834 Frequency : only with obd_echo servers and clients that are rebooted
3836 Description: kernel BUG accessing uninitialized data structure
3837 Details : When running an obd_echo server it did not start the ping_evictor
3838 thread, and when a client was evicted an uninitialized data
3839 structure was accessed. Start the ping_evictor in the RPC
3840 service startup instead of the OBD startup.
3842 Severity : enhancement
3843 Bugzilla : 10193 (patchless)
3844 Description: Remove dependency on various unexported kernel interfaces.
3845 Details : No longer need reparent_to_init, exit_mm, exit_files,
3846 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3849 Frequency : rare (only users of deprecated and unsupported LDAP config)
3851 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3852 Details : LDAP apparently contains 'lustreName' attributes instead of
3853 'name'. A simple remapping of the name is sufficient.
3856 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3858 Description: waiting_locks_callback trips kernel BUG if client is evicted
3859 Details : Running with the dump_on_timeout debug flag turned on makes
3860 it possible that the waiting_locks_callback() can try to dump
3861 the Lustre kernel debug logs from an interrupt handler. Defer
3862 this log dumping to the expired_lock_main() thread.
3864 Severity : enhancement
3866 Description: Support NFS exporting on 2.6 kernels.
3867 Details : Implement non-rawops metadata methods for NFS server to use without
3868 changing NFS server code.
3871 Frequency : very rare (synthetic metadata workload only)
3873 Description: two racing renames might cause an MDS thread to deadlock
3874 Details : Running the "racer" program may cause one MDS thread to rename
3875 a file from being the source of a rename to being the target of
3876 a rename at exactly the same time that another thread is doing
3877 so, and the second thread has already enqueued these locks after
3878 doing a lookup of the target and is trying to relock them in
3879 order. Ensure that we don't try to re-lock the same resource.
3882 Frequency : only very large systems with liblustre clients
3884 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3885 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3886 all OSTs in parallel. This allows the eviction of stale liblustre
3887 clients to proceed much faster than if they were done in series,
3888 and also offers similar improvements for other set_info RPCs.
3893 Description: excessive CPU usage during initial read phase on client
3894 Details : During the initial read phase on a client, it would agressively
3895 retry readahead on the file, consuming too much CPU and impacting
3896 performance (since 1.4.5.8). Improve the readahead algorithm
3897 to avoid this, and also improve some other common cases (read
3898 of small files in particular, where "small" is files smaller than
3899 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3904 Description: MDS crash when receiving packet with unknown intent.
3905 Details : Do not LBUG in unknown intent case, just return -EFAULT
3907 Severity : enhancement
3908 Bugzilla : 9293, 9385
3909 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3910 Details : Do not serialize getattr (non-intent version) and statfs.
3913 Frequency : occasional, when OST network is overloaded/intermittent
3915 Description: client evicted by OST after bulk IO timeout
3916 Details : If a client sends a bulk IO request (read or write) the OST
3917 may evict the client if it is unresposive to its data GET/PUT
3918 request. This is incorrect if the network is overloaded (takes
3919 too long to transfer the RPC data) or dropped the OST GET/PUT
3920 request. There is no need to evict the client at all, since
3921 the pinger and/or lock callbacks will handle this, and the
3922 client can restart the bulk request.
3925 Frequency : Always when mmapping file with no objects
3927 Description: client crashes when mmapping file with no objects
3928 Details : Check that we actually have objects in a file before doing any
3929 operations on objects in ll_vm_open, ll_vm_close and
3935 Description: Request leak when working with deleted CWD
3936 Details : Introduce advanced request refcount tracking for requests
3937 referenced from lustre intent.
3939 Severity : Enhancement
3941 Description: Cache open file handles on client.
3942 Details : MDS now will return special lock along with openhandle, if
3943 requested and client is allowed to hold openhandle, even if unused,
3944 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3945 closing files for every read/write openration.
3947 Severity : Enhancement
3949 Description: Cache open negative dentries on client when possible.
3950 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3951 negative dentries on lock revocation.
3956 Description: Remounting a client read-only wasn't possible with a zconf mount
3957 Details : It wasn't possible to remount a client read-only with llmount.
3959 Severity : enhancement
3960 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3961 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3962 MPICH is included to add Lustre-specific ADIO interfaces.
3963 This is based closely on the UFS ADIO layer and only differs
3964 in file creation, in order to allow the OST striping to be set.
3965 This is user-contributed code and not supported by CFS.
3970 Description: extended inode attributes (immutable, append-only) work improperly
3971 when 2.4 and 2.6 kernels are used on client/server or vice versa
3972 Details : Introduce kernel-independent values for these flags.
3974 Severity : enhancement
3977 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3978 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3979 granularity. Now, Fractional MB granularity is be supported,
3980 this is very useful for low memory system.
3982 Severity : enhancement
3984 Description: Getattr by fid
3985 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3986 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3987 have locks and avoids getattr rpc requests for it on every stat.
3990 Frequency : Always, for filesystems larger than 2TB
3992 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3993 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3994 filesystems larger than 2TB to be mounted with mballoc enabled.
3997 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3999 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4000 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4001 filesystems larger than 2TB is mounted then the kernel will
4002 silently corrupt the start of the filesystem. CONFIG_LBD is
4003 enabled for all CFS-supported kernels, but the possibility of
4004 this happening with a modified kernel config exists.
4006 Severity : enhancement
4008 Description: add client O_DIRECT support for 2.6 kernels
4009 Details : It is now possible to do O_DIRECT reads and writes to files
4010 in the Lustre client mountpoint on 2.6 kernel clients.
4012 Severity : enhancement
4014 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4015 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4016 parallel, not waiting for response from every OST before sending
4017 a rpc to the next OST.
4022 Description: setattr vs write race when updating file timestamps
4023 Details : Client processes that update a file timestamp into the past
4024 right after writing to the file (e.g. tar) it is possible that
4025 the updated file modification time can be reset to the current
4026 time due to a race between processing the setattr and write RPC.
4028 Severity : enhancement
4030 Description: Bring 'lfs find' closer in line with regular Linux find.
4031 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4032 -print0 options and obtains all the needed info through the lustre
4035 Severity : enhancement
4037 Description: support up to 1024 configured devices on one node
4038 Details : change obd_dev array from statically allocated to dynamically
4039 allocated structs as they are first used to reduce memory usage
4044 Description: Flush dirty partially truncated pages during truncate
4045 Details : Immediatelly flush partially truncated pages in filter_setattr,
4046 this way we completely avoid having any pages in page cache on OST
4047 and can retire ugly workarounds during writes to flush such pages.
4052 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4053 Details : protect inode from truncation within vfs_unlink() context
4054 just take a reference before calling vfs_unlink() and release it
4055 when parent's i_sem is free.
4060 Description: last_id value checked outside lock on OST caused LASSERT failure
4061 Details : If there were multiple MDS->OST object precreate requests in
4062 flight, it was possible that the OST's last object id was checked
4063 outside a lock and incorrectly tripped an assertion. Move checks
4064 inside locks, and discard old precreate requests.
4067 Frequency : always, if extents are used on OSTs
4069 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4070 Details : OSTs using the extents format would not zero the high 16 bits of
4071 the index physical block number. This is not a problem for any
4072 OST filesystems smaller than 16TB, and no kernels support ext3
4073 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4074 new/modified files) and can be fixed for existing filesystems
4075 with e2fsprogs-1.39-cfs1.
4080 Description: import connection selection may be incorrect if timer wraps
4081 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4082 import connections to be ignored if the 32-bit jiffies counter
4083 wraps. Use a 64-bit jiffies counter.
4086 Frequency : very large clusters immediately after boot
4088 Description: LNET request buffers exhausted under heavy short-term load
4089 Details : If a large number of client requests are generated on a service
4090 that has previously never seen so many requests it is possible
4091 that the request buffer growth cannot keep up with the spike in
4092 demand. Instead of dropping incoming requests, they are held in
4093 the LND until the RPC service can accept more requests.
4096 Frequency : Sometimes during replay
4098 Description: Assertion failure in ll_local_open after replay.
4099 Details : If replay happened on an open request reply before we were able
4100 to set replay handler, reply will become not swabbed tripping the
4101 assertion in ll_local_open. Now we set the handler right after
4102 recognising of open request
4105 Frequency : very rare
4107 Description: kernel reports "badness in vsnprintf"
4108 Details : Reading from the "recovery_status" /proc file in small chunks
4109 may cause a negative length in lprocfs_obd_rd_recovery_status()
4110 call to vsnprintf() (which is otherwise harmless). Exit early
4111 if there is no more space in the output buffer.
4113 Severity : enhancement
4115 Description: clear OBD RPC statistics by writing to them
4116 Details : It is now possible to clear the OBD RPC statistics by writing
4117 to the "stats" file.
4122 Description: Client mtime is not the same on different clients after utimes
4123 Details : In some cases, the client was using the utimes() syscall on
4124 a file cached on another node. The clients now validate the
4125 ctime from the MDS + OSTs to determine which one is right.
4130 Description: Inability to activate failout mode
4131 Details : lconf script incorrectly assumed that in pythong string's numeric
4132 value is used in comparisons.
4135 Frequency : always with multiple stripes per file
4137 Description: Inefficient object allocation for mutli-stripe files
4138 Details : When selecting which OSTs to stripe files over, for files with
4139 a stripe count that divides evenly into the number of OSTs,
4140 the MDS is always picking the same starting OST for each file.
4141 Return the OST selection heuristic to the original design.
4146 Description: mount failures may take full timeout to return an error
4147 Details : Under some heavy load conditions it is possible that a
4148 failed mount can wait for the full obd_timeout interval,
4149 possibly several minutes, before reporting an error.
4150 Instead return an error as soon as the status is known.
4152 Frequency : quota enabled and large files being deleted
4154 Description: releasing more than 4GB of quota at once hangs OST
4155 Details : If a user deletes more than 4GB of files on a single OST it
4156 will cause the OST to spin in an infinite loop. Release
4157 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4162 Description: statfs data retrieved from /proc may be stale or zero
4163 Details : When reading per-device statfs data from /proc, in the
4164 {kbytes,files}_{total,free,avail} files, it may appear
4165 as zero or be out of date.
4168 Frequency : systems with MD RAID1 external journal devices
4170 Description: lconf's call to blkid is confused by RAID1 journal devices
4171 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4172 of returning all block devices that match the journal UUID.
4175 Frequency : always, for aggregate stripe size over 4GB
4177 Description: assertion fails when trying to use 4GB stripe size
4178 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4179 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4182 Frequency : always on ppc64
4184 Description: the first write on an ext3 filesystem with mballoc got stuck
4185 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4186 perform endianness conversion.
4188 ------------------------------------------------------------------------------
4190 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4192 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4193 INTEROPERATE with older versions automatically. Please read the
4194 user documentation before upgrading any part of a live system.
4195 * WARNING: Lustre networking configuration changes are required with
4196 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4199 * Support for kernels:
4200 2.6.9-22.0.2.EL (RHEL 4)
4201 2.6.5-7.244 (SLES 9)
4202 2.6.12.6 vanilla (kernel.org)
4205 Severity : enhancement
4206 Bugzilla : 7981/8208
4207 Description: Introduced Lustre Networking (LNET)
4208 Details : LNET is new networking infrastructure for Lustre, it includes
4209 a reorganized network configuration mode (see the user
4210 documentation for full details) as well as support for routing
4211 between different network fabrics. Lustre Networking Devices
4212 (LNDs) for the supported network fabrics have also been
4213 created for this new infrastructure.
4215 Severity : enhancement
4216 Description: Introduced Access control lists
4217 Details : clients can set ACLs on files and directories in order to have
4218 more fine-grained permissions than the standard Unix UGO+RWX.
4219 The MDS must be started with the "-o acl" mount option.
4221 Severity : enhancement
4222 Description: Introduced filesystem quotas
4223 Details : Administrators may now establish per-user quotas on the
4226 Severity : enhancement
4228 Description: Configuration change for the XT3
4229 The PTLLND is now used to run Lustre over Portals on the XT3
4230 The configure option(s) --with-cray-portals are no longer used.
4231 Rather --with-portals=<path-to-portals-includes> is used to
4232 enable building on the XT3. In addition to enable XT3 specific
4233 features the option --enable-cray-xt3 must be used.
4238 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4239 Details : A race between allocating a new llcd and re-getting the llcd_lock
4240 allowed another thread to grab newly-allocated llcd.
4242 Severity : enhancement
4244 Description: 2.6 OST async journal commit and locking fix to improve performance
4245 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4246 2.6 kernels are now async as they already were in 2.4 kernels so
4247 that they can commit concurrently with the network bulk transfer.
4248 For block-allocated files the filter allocation semaphore is held
4249 to avoid filesystem fragmentation during allocation. BKL lock
4250 removed for 2.6 xattr operations where it is no longer needed.
4255 Description: lconf incorrectly determined whether two IP networks could talk
4256 Details : In some more complicated routing and multiple-network
4257 configurations, lconf will avoid trying to make a network
4258 connection to a disjoint part of the IP space. It was doing the
4259 math incorrectly for one set of cases.
4264 Description: Fix for potential infinite loop processing records in an llog.
4265 Details : If an llog record is corrupted/zeroed, it is possible to loop
4266 forever in llog_process(). Validate the llog record length
4267 and skip the remainder of the block on error.
4270 Frequency : occasional (liblustre only)
4272 Description: liblustre could not open files whose last component is a symlink
4273 Details : sysio_path_walk() would incorrectly pass the open intent to
4274 intermediate path components.
4277 Frequency : rare (liblustre only with non-standard tuning)
4278 Bugzilla : 7201 (7350)
4279 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4280 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4281 after it has been released, only do this for referenced locks
4283 Severity : enhancement
4285 Description: specifying an (invalid) directory default stripe_size of -1
4286 would reset the directory default striping
4287 Details : stripe_size -1 was used internally to signal directory stripe
4288 removal, now use "all default" to signal dir stripe removal
4289 as a directory striping of "all default" is not useful
4292 Frequency : common for large clusters running liblustre clients
4294 Description: doing an ls when liblustre clients are running is slow
4295 Details : sending a glimpse AST to a liblustre client waits for every AST
4296 to time out, as liblustre clients will not respond. Since they
4297 cannot cache data we refresh the OST lock LVB from disk instead.
4299 Severity : enhancement
4301 Description: doing an ls at the same time as file IO can be slow
4302 Details : enqueue and other "small" requests can be blocked behind many
4303 large IO requests. Create a new OST IO portal for non-IO
4304 requests so they can be processed faster.
4307 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4309 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4310 created file with mode 000
4311 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4312 allow an ftruncate() as a non-root user to a file with mode 000.
4313 Lustre now respects this flag to disable mode checks when
4314 truncating a file owned by the user
4317 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4320 Description: Revoking locks from clients that went dead or catatonic might take
4322 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4323 cancellation of such locks instant on servers without waiting for
4324 any reply from clients. Clients drops these locks when cancel
4325 notification from server is received without replying.
4328 Frequency : liblustre-only, when liblustre client dies or becomes busy
4330 Description: Doing ls on Linux clients can take a long time with active
4332 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4333 granting such locks to it in the first place if possible. Locks
4334 are taken by proxy on the OST during the read or write and
4335 dropped immediately afterward. Add connect flags handling, do
4336 not grant locks to liblustre clients for glimpse ASTs.
4338 Severity : enhancement
4340 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4341 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4342 sequential reads and behaves badly with applications doing random
4343 reads. Improve it by reading ahead at least read region, and
4344 avoiding excessive large RPC for small reads.
4346 Severity : enhancement
4348 Description: Creating more than 1000 files for a single job may cause a load
4349 imbalance on the OSTs if there are also a large number of OSTs.
4350 Details : qos_prep_create() uses an OST index reseed value that is an
4351 even multiple of the number of available OSTs so that if the
4352 reseed happens in the middle of the object allocation it will
4353 still utilize the OSTs as uniformly as possible.
4358 Description: OST or MDS may oops in ping_evictor_main()
4359 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4360 but doesn't restart at beginning of obd_exports_timed list
4363 Severity : enhancement
4365 Description: improve by-nid export eviction on the MDS and OST
4366 Details : allow multiple exports with the same NID to be evicted at one
4367 time without re-searching the exports list.
4370 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4372 Description: MDS may oops in groups_free()
4373 Details : in rare race conditions a newly allocated group_info struct is
4374 freed again, and this can be NULL. The 2.4 compatibility code
4375 for groups_free() checked for a NULL pointer, but 2.6 did not.
4378 Frequency : common for liblustre clients doing little filesystem IO
4379 Bugzilla : 9352, 7313
4380 Description: server may evict liblustre clients accessing contended locks
4381 Details : if a client is granted a lock or receives a completion AST
4382 with a blocking AST already set it would not reply to the AST
4383 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4386 Frequency : lfs setstripe, only systems with more than 160 OSTs
4388 Description: unable to set striping with a starting offset beyond OST 160
4389 Details : llapi_create_file() incorrectly limited the starting stripe
4390 index to the maximum single-file stripe count.
4393 Frequency : LDAP users only
4395 Description: lconf did not handle in-kernel recovery with LDAP properly
4396 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4398 Severity : enhancement
4400 Description: bind OST threads to NUMA nodes to improve performance
4401 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4402 node and do their allocations there to localize memory access
4404 Severity : enhancement
4406 Description: llmount can determine client NID directly from Myrinet (GM)
4407 Details : the client NID code from gmnalnid was moved directly into
4408 llmount, removing the need to use this or specifying the
4409 client NID explicitly when mounting GM clients with zeroconf
4412 Frequency : if client is started with down MDS
4414 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4415 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4416 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4417 ptlrpc_import_delayed_req() because we didn't check if the
4418 request was interrupted, and we also didn't break out of the
4419 event loop if there was no timeout
4424 Description: data loss during non-page-aligned writes to a single file from
4425 both multiple nodes and multiple threads on one node at same time
4426 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4427 inconsistency led to false short-reads, that were cached and later
4428 used by ->prepare_write() to fill in partially written page,
4429 leading to data loss.
4432 Frequency : always, if lconf --abort_recovery used
4434 Description: lconf --abort_recovery fails with 'Operation not supported'
4435 Details : lconf was attempting to abort recovery on the MDT device and not
4438 Severity : enhancement
4440 Description: remove cleanup logs
4441 Details : replace lconf-generated cleanup logs with lustre internal
4442 cleanup routines. Eliminates the need for client-cleanup and
4445 Severity : enhancement
4447 Description: add support for EAs (user and system) on lustre filesystems
4448 Details : it is now possible to store extended attributes in the Lustre
4449 client filesystem, and with the user_xattr mount option it
4450 is possible to allow users to store EAs on their files also
4452 Severity : enhancement
4454 Description: Add possibility (config option) to show minimal available OST free
4456 Details : When compiled with --enable-mindf configure option, statfs(2)
4457 (and so, df) will return least minimal free space available from
4458 all OSTs as amount of free space on FS, instead of summary of
4459 free spaces of all OSTs.
4461 Severity : enhancement
4463 Description: do not expand extent locks acquired on OST-side
4464 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4465 by server: they are not cached anyway.
4468 Frequency : when mmap is used/binaries executed from Lustre
4470 Description: Unmmap pages before throwing them away from read cache.
4471 Details : llap_shrink cache now attempts to unmap pages before discarding
4472 them (if unmapping failed - do not discard). SLES9 kernel has
4473 extra checks that trigger if this unmapping is not done first.
4478 Description: lconf didn't resolve symlinks before checking to see whether a
4479 given mountpoint was already in use
4482 Frequency : when migrating failover services
4483 Bugzilla : 6395, 9514
4484 Description: When migrating a subset of services from a node (e.g. failback
4485 from a failover service node) the remaining services would
4486 time out and evict clients.
4487 Details : lconf --force (implied by --failover) sets the global obd_timeout
4488 to 5 seconds in order to quickly disconnect, but this caused
4489 other RPCs to time out too quickly. Do not change the global
4490 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4492 Severity : enhancement
4493 Frequency : if MDS is started with down OST
4494 Bugzilla : 9439,5706
4495 Description: Allow startup/shutdown of an MDS without depending on the
4496 availability of the OSTs.
4497 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4498 Add appropriate locking and lov-osc refcounts for safe
4499 cleaning. Add osc abort_inflight calls in case the
4500 synchronize never started.
4503 Frequency : occasional (Cray XT3 only)
4505 Description: root not authorized to access files in CRAY_PORTALS environment
4506 Details : The client process capabilities were not honoured on the MDS in
4507 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4508 been accessed by an authorized user then root was able to access
4509 the file on the local client also. The root user capabilities
4510 are now allowed on the MDS, as this environment has secure UID.
4513 Frequency : occasional
4515 Description: ldiskfs "too long searching" message happens too often
4516 Details : A debugging message (otherwise harmless) prints too often on
4517 the OST console. This has been reduced to only happen when
4518 there are fragmentation problems on the filesystem.
4523 Description: Division by zero in statfs when all OSCs are inactive
4524 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4525 lov_statfs divides by value returned by lov_get_stripecnt().
4529 Bugzilla : 9489, 3273
4530 Description: First write from each client to each OST was only 4kB in size,
4531 to initialize client writeback cache, which caused sub-optimal
4532 RPCs and poor layout on disk for the first writen file.
4533 Details : Clients now request an initial cache grant at (re)connect time
4534 and so that they can start streaming writes to the cache right
4535 away and always do full-sized RPCs if there is enough data.
4536 If the OST is rebooted the client also re-establishes its grant
4537 so that client cached writes will be honoured under the grant.
4542 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4543 Details : Now I/O RPCs go to different portal number and (presumably) fast
4544 lock requests (and glimses) and other RPCs get their own service
4545 threads pool that should be able to service those RPCs
4548 Severity : enhancement
4550 Description: Ability to exchange lustre version between client and servers and
4551 issue warnings at client side if client is too old. Also for
4552 liblustre clients there is ability to refuse connection of too old
4554 Details : New 'version' field is added to connect data structure that is
4555 filled with version info. That info is later checked by server and
4559 Frequency : rare, liblustre only.
4560 Bugzilla : 9296, 9581
4561 Description: Two simultaneous writes from liblustre at offset within same page
4562 might proceed at the same time overwriting eachother with stale
4564 Details : I/O lock withing llu_file_prwv was released too early, before data
4565 actually was hitting the wire. Extended lock-holding time until
4566 server acknowledges receiving data.
4569 Frequency : extremely rare. Never observed in practice.
4571 Description: avoid generating lustre_handle cookie of 0.
4572 Details : class_handle_hash() generates handle cookies by incrementing
4573 global counter, and can hit 0 occasionaly (this is unlikely, but
4574 not impossible, because initial value of cookie counter is
4575 selected randonly). Value of 0 is used as a sentinel meaning
4576 "unassigned handle" --- avoid it. Also coalesce two critical
4577 sections in this function into one.
4579 Severity : enhancement
4581 Description: allow liblustre clients to delegate truncate locking to OST
4582 Details : To avoid overhead of locking, liblustre client instructs OST to
4583 take extent lock in ost_punch() on client's behalf. New connection
4584 flag is added to handle backward compatibility.
4586 Severity : enhancement
4587 Bugzilla : 4928, 7341, 9758
4588 Description: allow number of OST service threads to be specified
4589 Details : a module parameter allows the number of OST service threads
4590 to be specified via "options ost ost_num_threads={N}" in the
4591 OSS's /etc/modules.conf or /etc/modprobe.conf.
4595 Bugzilla : 6146, 9635, 9895
4596 Description: servers crash with bad pointer in target_handle_connect()
4597 Details : In rare cases when a client is reconnecting it was possible that
4598 the connection request was the last reference for that export.
4599 We would temporarily drop the export reference and get a new
4600 one, but this may have been the last reference and the export
4601 was just destroyed. Get new reference before dropping old one.
4603 Severity : enhancement
4604 Frequency : if client is started with failover MDS
4606 Description: Allow multiple MDS hostnames in the mount command
4607 Details : Try to read the configuration from all specified MDS
4608 hostnames during a client mount in case the "primary"
4611 Severity : enhancement
4613 Description: Stop sending data to evicted clients as soon as possible.
4614 Details : Check if the client we are about to send or are sending data to
4615 was evicted already. (Check is done every second of waiting,
4616 for which l_wait_event interface was extended to allow checking
4617 of exit condition at specified intervals).
4620 Frequency : rare, normally only when NFS exporting is done from client
4622 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4624 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4627 Frequency : always, if srand() is called before liblustre initialization
4629 Description: Liblustre uses system PRNG disturbing its usage by user application
4630 Details : Introduce internal to lustre fast and high-quality PRNG for
4631 lustre usage and make liblustre and some other places in generic
4632 lustre code to use it.
4634 Severity : enhancement
4635 Bugzilla : 9477, 9557, 9870
4636 Description: Verify that the MDS configuration logs are updated when xml is
4637 Details : Check if the .xml configuration logs are newer than the config
4638 logs stored on the MDS and report an error if this is the case.
4639 Request --write-conf, or allow starting with --old_conf.
4641 Severity : enhancement
4643 Description: Handle symlinks in the path when checking if Lustre is mounted.
4644 Details : Resolve intermediate symlinks when checking if a client has
4645 mounted a filesystem to avoid duplicate client mounts.
4650 Description: lconf can hit an error exception but still return success.
4651 Details : The lconf command catches the Command error exception at the top
4652 level script context and will exit with the associated exit
4653 status, but doesn't ensure that this exit status is non-zero.
4658 Description: failure of ptlrpc thread startup can cause oops
4659 Details : Starting a ptlrpc service thread can fail if there are a large
4660 number of threads or the server memory is very fragmented.
4661 Handle this without oopsing.
4664 Frequency : always, only if liblustre and non-default acceptor port was used
4666 Description: liblustre cannot connect to servers with non-default acceptor port
4667 Details : tcpnal_set_default_params() was not called and was therefore
4668 ignoring the environment varaible TCPNAL_PORT, as well as other
4669 TCPNAL_ environment variables
4674 Description: two objects could be created on the same OST for a single file
4675 Details : If an OST is down, in some cases it was possible to create two
4676 objects on a single OST for a single file. No problems other
4677 than potential performance impact and spurious error messages.
4681 Bugzilla : 5681, 9562
4682 Description: Client may oops in ll_unhash_aliases
4683 Details : Client dcache may become inconsistent in race condition.
4684 In some cases "getcwd" can fail if the current directory is
4690 Description: Inode refcounting problems in NFS export code
4691 Details : link_raw functions used to call d_instantiate without obtaining
4692 extra inode reference first.
4696 Bugzilla : 9942, 9903
4697 Description: Referencing freed requests leading to crash, memleaks with NFS.
4698 Details : We used to require that call to ll_revalidate_it was always
4699 followed by ll_lookup_it. Also with revalidate_special() it is
4700 possible to call ll_revalidate_it() twice for the same dentry
4701 even if first occurence returned success. This fix changes semantic
4702 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4703 reference on a request referred from the intent.
4704 ll_intent_release() then releases such a request.
4707 Frequency : rare, normally benchmark loads only
4709 Description: unlinked inodes were kept in memory on the client
4710 Details : If a client is repeatedly creating and unlinking files it
4711 can accumulate a lot of stale inodes in the inode slab cache.
4712 If there is no other client load running this can cause the
4713 client node to run out of memory. Instead flush old inodes
4714 from client cache that have the same inode number as a new inode.
4717 Frequency : SLES9 2.6.5 kernel and long filenames only
4718 Bugzilla : 9969, 10379
4719 Description: utime reports stale NFS file handle
4720 Details : SLES9 uses out-of-dentry names in some cases, which confused
4721 the lustre dentry revalidation. Change it to always use the
4725 Frequency : rare, unless heavy write-truncate concurrency is continuous
4726 Bugzilla : 4180, 6984, 7171, 9963, 9331
4727 Description: OST becomes very slow and/or deadlocked during object unlink
4728 Details : filter_destroy() was holding onto the parent directory lock
4729 while truncating+unlinking objects. For very large objects this
4730 may block other threads for a long time and slow overall OST
4731 responsiveness. It may also be possible to get a lock ordering
4732 deadlock in this case, or run out of journal credits because of
4733 the combined truncate+unlink. Solution is to do object truncate
4734 first in one transaction without parent lock, and then do the
4735 final unlink in a new transaction with the parent lock. This
4736 reduces the lock hold time dramatically.
4739 Frequency : rare, 2.4 kernels only
4741 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4742 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4743 before unmounting their filesystem in order to avoid deadlock.
4744 The kernel_locked() function in 2.4 kernels only checks whether
4745 the kernel lock is held, not whether it is this process that is
4746 holding it as 2.6 kernels do.
4751 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4752 Details : The client ptlrpc code may be trying to reconnect to a down
4753 server before a previous connection attempt has timed out.
4754 Increase the reconnect interval to be longer than the connection
4755 timeout interval to avoid sending duplicate connections to
4759 Frequency : echo_client brw_test command
4761 Description: fix echo_client to work with OST preallocated code
4762 Details : OST preallocation code (5137) didn't take echo_client IO path
4763 into account: echo_client calls filter methods outside of any
4764 OST thread and, hence, there is no per-thread preallocated
4765 pages and buffers to use. Solution: hijack pga pages for IO. As
4766 a byproduct, this avoids unnecessary data copying.
4770 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4771 Description: Client can oops in mdc_commit_close() after open replay
4772 Details : It was possible for the MDS to return an open request with no
4773 transaction number in mds_finish_transno() if the client was
4774 evicted, but without actually returning an error. Clients
4775 would later try to replay that open and may trip an assertion
4776 Simplify the client close codepath, and always return an error
4777 from the MDS in case the open is not successful.
4780 Frequency : rare, 2.6 OSTs only
4782 Description: OST may deadlock under high load on fragmented files
4783 Details : If there was a heavy load and highly-fragmented OST filesystems
4784 it was possible to have all the OST threads deadlock waiting on
4785 allocation of biovecs, because the biovecs were not released
4786 until the entire RPC IO was completed. Instead, release biovecs
4787 as soon as they are complete to ensure forward IO progress.
4789 Severity : enhancement
4791 Description: Support for specifying external journal device at mount
4792 Details : If an OST or MDS device is formatted with an external journal
4793 device, this device major/minor is stored in the ext3 superblock
4794 and may not be valid for failover. Allow detecting and
4795 specifying the external journal at mount time.
4800 Description: Mounting an MDS with pending unlinked files may cause oops
4801 Details : target_finish_recovery() calls mds_postrecov() which returned
4802 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4803 considers this an error and immediately begins cleaning up the
4804 lov, just after starting the mds_lov process
4806 Severity : enhancement
4808 Description: Implement 'lfs df' to report actual free space on per-OST basis
4809 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4810 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4811 usage of objects; '-h' to report in human readable format.
4813 ------------------------------------------------------------------------------
4815 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4822 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4823 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4824 at startup during orphan unlinking. Instead, skip buddy update
4825 before setup, as it will be regenerated after recovery is complete.
4830 Description: If an OST is inactive, its locks might reference stale inodes.
4831 Details : lov_change_cbdata() must iterate over all namespaces, even if
4832 they are inactive to clear inode references from the lock.
4834 Severity : enhancement
4835 Frequency : occasional, if non-standard max_dirty_mb used
4837 Description: Client will block write RPCs if not enough grant
4838 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4839 then the client will block writes while waiting for another RPC
4840 to complete instead of consuming its dirty limit. With change
4841 we get improved performance when max_dirty_mb is small.
4843 Severity : enhancement
4844 Bugzilla : 3389, 6253
4845 Description: Add support for supplementary groups on the MDS.
4846 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4847 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4848 lookups for user supplementary groups into a cache.
4852 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4853 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4854 for liblustre O_EXCL is enforced by MDS.
4858 Description: readdir never returns NULL in liblustre.
4859 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4860 dentry in struct dirent.
4864 Frequency : liblustre only, depends on application IO pattern
4865 Description: liblustre clients evicted if not contacting servers
4866 Details : Don't put liblustre clients into the ping_evictor list, so
4867 they will not be evicted by the pinger ever.
4869 Severity : enhancement
4871 Description: Add ability to evict clients by NID from MDS.
4872 Details : By echoing "nid:$NID" string into
4873 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4874 $NID would be instantly evicted from this MDS and from all active
4875 OSTs connected to it.
4879 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4880 Details : lookup_it_finish() used to query file size from OSTs that was not
4885 Description: service threads change working directory to that of init
4886 Details : Starting lustre service threads may pin the working directory
4887 of the parent thread, making that filesystem busy. Threads
4888 now change to the working directory of init to avoid this.
4892 Frequency : during shutdown only
4893 Description: shutdown with a failed MDS or OST can cause unmount to hang
4894 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4899 Frequency : 2.6 only
4900 Description: chmod/chown may include an extra supplementary group
4901 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4902 supplementary group and if none is specified this is used.
4905 Bugzilla : 5479 (6816)
4907 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4908 Details : If lookup is in progress on a file that is unlinked we might try
4909 to revalidate the inode and fail in revalidate after lookup is
4910 complete and ll_file_open() enqueues the open again but
4911 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4914 Frequency : always, if lconf --abort_recovery used
4916 Description: lconf --abort_recovery fails with 'Operation not supported'
4917 Details : lconf was attempting to abort recovery on the MDT device and not
4920 ------------------------------------------------------------------------------
4922 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4927 Frequency : rare (only unsupported configurations with a node running as an
4929 Bugzilla : 6514, 5137
4930 Description: Mounting a Lustre file system on a node running as an OST could
4932 Details : OSTs now preallocates memory needed to write out data at
4933 startup, instead of when needed, to avoid having to
4934 allocate memory in possibly low memory situations.
4935 Specifically, if the file system is mounted on on OST,
4936 memory pressure could force it to try to write out data,
4937 which it needed to allocate memory to do. Due to the low
4938 memory, it would be unable to do so and the node would
4939 become unresponsive.
4941 Severity : enhancement
4943 Description: Addition of lconf --service command line option
4944 Details : lconf now accepts a '--service <arg>' option, which is
4945 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4947 Severity : enhancement
4949 Description: Failover mode is now the default for OSTs.
4950 Details : By default, OSTs will now run in failover mode. To return to
4951 the old behaviour, add '--failout' to the lmc line for OSTs.
4953 Severity : enhancement
4955 Description: Health checks are now provided for MDS and OSTs
4956 Details : Additional detailed health check information on MSD and OSTs
4957 is now provided through the procfs health_check value.
4960 Frequency : occasional, depends on IO load
4962 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4963 after numerous create/delete cycles
4964 Details : The ext3 inode allocation policy would not allocate new inodes
4965 very well on the OSTs because there are no new directories
4966 being created. Instead we look for groups with free space if
4967 the parent directories are nearly full.
4972 Description: Network or server problems during mount may cause partially
4973 mounted clients instead of returning an error.
4974 Details : The config llog parsing code may overwrite the error return
4975 code during mount error handling, returning success instead
4981 Description: MDS can fail to allocate large reply buffers
4982 Details : After long uptimes the MDS can fail to allocate large reply
4983 buffers (e.g. zconf client mount config records) due to memory
4984 fragmentation or consumption by the buffer cache. Preallocate
4985 some large reply buffers so that these replies can be sent even
4986 under memory pressure.
4990 Frequency : rare (liblustre)
4991 Description: fsx running with liblustre complained that using truncate() to
4992 extend the file doesn't work. This patch corrects that issue.
4993 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4994 ATTR_SIZE and lsm use in llu_setattr_raw.
4998 Frequency : rare, only 2.6 kernels
4999 Description: Unusual file access patterns on the MDS may result in inode
5000 data being lost in very rare circumstances.
5001 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5002 code in the 2.6 kernel caused the inode and/or EA data not to
5003 be read from disk, causing single-file corruption.
5007 Frequency : rare, only 2.6 filesystems using extents
5008 Description: Heavy concurrent write and delete load may cause data corruption.
5009 Details : It was possible under high-load situations to have an extent
5010 metadata block in the block device cache from a just-unlinked
5011 file overwrite a newly-allocated data block. We now unmap any
5012 metadata buffers that alias just-allocated data blocks.
5016 Frequency : filesystems with default stripe_count larger than 77
5017 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5018 Details : lconf specifies an inode size of 4096 bytes when the default
5019 stripe_count is larger than 77. This conflicts with the default
5020 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5021 this case to avoid pinning too much memory for large EAs.
5023 ------------------------------------------------------------------------------
5025 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5030 Frequency : rare (extremely heavy IO load with hundreds of clients)
5032 Description: Client is evicted, gets IO error writing to file
5033 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5034 caused clients to be evicted for AST timeouts. The fixes in
5035 bug 5192 mean we no longer need to have such short AST timeouts
5036 so ldlm_timeout has been increased.
5039 Frequency : occasional during --force or --failover shutdown under load
5040 Bugzilla : 5949, 4834
5041 Description: Server oops/LBUG if stopped with --force or --failover under load
5042 Details : a collection of import/export refcount and cleanup ordering
5043 issues fixed for safer force cleanup
5046 Frequency : only filesystems larger than 120 OSTs
5047 Bugzilla : 5990, 6223
5048 Description: lfs getstripe would oops on a very large filesystem
5049 Details : lov_getconfig used kfree on vmalloc'd memory
5052 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5053 Bugzilla : 6242, 6243
5054 Description: reading from files that had been truncated to a non-zero size
5055 but never opened returned no data
5056 Details : ll_file_read() reads zeros from no-object files to EOF
5061 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5062 Details : There's little harm in aborting MDS/OSS recovery and letting it
5063 try again, so I removed the LASSERT and return an error instead.
5065 Severity : enhancement
5067 Description: New debugging infrastructure for tracking down data corruption
5068 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5069 (b) cover more of the client-side code path, and (c) try to narrow
5070 down where problems occurred
5074 Bugzilla : 3819, 4364, 4397, 6313
5075 Description: Racing close and eviction MDS could cause assertion in mds_close
5076 Details : It was possible to get multiple mfd references during close and
5077 client eviction, leading to one thread referencing a freed mfd.
5079 Severity: : enhancement
5080 Bugzilla : 3262, 6359
5081 Description: Attempts to reconnect to servers are now more aggressive.
5082 Details : This builds on the enhanced upcall-less recovery that was added
5083 in 1.4.2. When trying to reconnect to servers, clients will
5084 now try each server in the failover group every 10 seconds. By
5085 default, clients would previously try one server every 25 seconds.
5090 Description: After recovery, certain operations trigger a failed
5091 assertion on a client.
5092 Details : Failing over an mds, using lconf -d --failover, while a
5093 client was doing a readdir() call would cause the client to
5094 LBUG after recovery completed and the readdir() was resent.
5096 Severity : enhancement
5098 Description: Default groups are now added by lconf
5099 Details : You can now run lconf --group <servicename> without having to
5100 manually add groups with lmc.
5103 Frequency : occasional
5105 Description: Nodes with an elan id of 0 trigger a failed assertion
5108 Frequency : always when accessing e.g. tty/console device nodes
5110 Description: tty and some other devices nodes cannot be used on lustre
5111 Details : file's private_data field is used by device data and lustre
5112 values in there got lost. New field was added to struct file to
5113 store fs-specific private data.
5116 Frequency : when exporting Lustre via NFS
5118 Description: NFSD failed occasionally when looking up a path component
5119 Details : NFSD is looking up ".." which was broken in ext3 directories
5120 that had grown large enough to become hashed.
5123 Frequency : Clusters with multiple interfaces not on the same subnet
5125 Description: Nodes will repeatedly try to reconnect to an interface which it
5126 cannot reach and report an error to the log.
5127 Details : Extra peer list entries will be created by lconf with some peers
5128 unreachable. lconf now validates the peer before adding it.
5131 Frequency : Only if a default stripe is set on the filesystem root.
5133 Description: Setting a default stripe on the filesystem root prevented the
5134 filesystem from being remounted.
5135 Details : The client was sending extra request flags in the root getattr
5136 request and did not allocate a reply buffer for the dir EA.
5139 Frequency : occasional, higher if lots of files are accessed by one client
5140 Bugzilla : 6159, 6097
5141 Description: Client trips assertion regarding lsm mismatch/magic
5142 Details : While revalidating inodes the VFS looks up inodes with ifind()
5143 and in rare cases can find an inode that is being freed.
5144 The ll_test_inode() code will free the lsm during ifind()
5145 when it finds an existing inode and then the VFS later attaches
5146 this free lsm to a new inode.
5150 Bugzilla : 6422, 7030
5151 Description: MDS deadlock between mkdir and client eviction
5152 Details : Creating a new file via mkdir or mknod (starting a transaction
5153 and getting the ns lock) can deadlock with client eviction
5154 (gets ns lock and trying to finish a synchronous transaction).
5157 Frequency : occasional
5158 Description: While starting a server, the fsfilt_ext3 module could not be
5160 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5161 kernels. Previously, lconf would still use the ext3 name
5162 when trying to load modules. Now, it will correctly use
5163 ext3 on 2.4 and ldiskfs on 2.6.
5165 Severity : enhancement
5166 Description: The default stripe count has been changed to 1
5167 Details : The interpretation of the default stripe count (0, to lfs
5168 or lmc) has been changed to mean striping across a single
5169 OST, rather than all available. For general usage we have
5170 found a stripe count of 1 or 2 works best.
5172 Severity : enhancement
5173 Description: Add support for compiling against Cray portals.
5174 Details : Conditional compiling for some areas that are different
5178 Frequency : occasional
5179 Bugzilla : 6409, 6834
5180 Description: Creating files with an explicit stripe count may lead to
5181 a failed assertion on the MDS
5182 Details : If some OSTs are full or unavailable, creating files may
5183 trigger a failed assertion on the MDS. Now, Lustre will
5184 try to use other servers or return an error to the
5188 Frequency : occasional
5190 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5191 caused lock timeout during readahead (since 1.4.2).
5192 Details : Processes doing readahead might match a lock that hasn't been
5193 granted yet if there are overlapping and conflicting lock
5194 requests. The readahead process waits on ungranted lock
5195 (original lock is CBPENDING), while OST waits for that process
5196 to cancel CBPENDING read lock and eventually evicts client.
5198 Severity : enhancement
5200 Description: Initial enabling of flock support for clients
5201 Details : Implements fcntl advisory locking and file status functions.
5202 This feature is provided as an optional mount flag (default
5203 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5204 locking are implemented yet, and those that are are not guaranteed
5205 to be completely correct in production environments.
5206 mount -t lustre -o [flock|noflock] ...
5209 Frequency : occasional
5211 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5212 assertion in the ext3 JBD (journaling) layer.
5213 Details : The b_committed_data struct is protected by the big kernel lock
5214 in 2.4 kernels, serializing journal_commit_transaction() and
5215 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5216 access to this struct. In 2.6 kernels there is finer grained
5217 locking to improve SMP performance of the JBD layer.
5221 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5226 Description: merge and cleanup kernel patches.
5229 -----------------------------------------------------------------------------
5231 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5233 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5234 versions of Lustre. Please update all servers and clients to
5235 version 1.4.2 or later at the same time. You must also run
5236 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5237 to update the configuration logs.
5239 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5240 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5241 - swab llog records of type '0' so we get proper header size/idx (5861)
5242 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5243 - fix rename of one directory over another leaking an inode (5953)
5244 - avoid SetPageDirty on 2.6 (5981)
5245 - don't re-add just-being-destroyed locks to the waiting list (5653)
5246 - when creating new directories, inherit the parent's custom
5247 striping settings if present parent (3048)
5248 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5249 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5250 - don't hold client locks on temporary worklist from l_lru (5666)
5251 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5252 - automatically evict dead clients (5921)
5253 - Update file size properly in create+truncate+fstat case (6196)
5254 - Do not unhash mountpoint dentries, do not allow removal of
5256 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5257 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5258 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5259 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5260 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5261 - Correctly handle reads of files with no objects (6243)
5262 - lctl recover will also mark a device active if deactivate used (5933)
5264 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5265 - allow --write-conf on an MDS with different nettype than client (5619)
5266 - don't write config llogs to MDS for mounts not from that MDS (5617)
5267 - lconf should create multiple TCP connections from a client (5201)
5268 - init scripts are now turned off by default; run chkconfig --on
5269 lustre and chkconfig --on lustrefs to use them
5270 - upcalls are no longer needed for clients to recover to failover
5272 - add --abort-recovery option to lconf to abort recovery on device
5274 - add support for an arbitrary number of OSTs (3026)
5275 - Quota support protocol changes.
5276 - forward compatibility changes to wire structs (6007)
5277 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5278 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5279 - improved "lustre status" script
5280 - initialize blocksize for non-regular files (6062)
5281 - added --disable-server and --disable-client configure options (5782)
5282 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5283 - Vanilla 2.4.29 support
5284 - increase maximum number of obd devices to 520 (6242)
5285 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5286 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5287 SLES 9 2.6.5 kernels
5288 - limit stripes per file to 160 (the maximum EA size) (6093)
5290 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5293 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5294 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5295 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5296 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5297 - small ext3 extents cleanups and fixes (5733)
5298 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5299 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5300 - avoid needless client->OST connect, fix handle mismatch (5317)
5301 - fix DLM error path that led to out-of-sync client, long delays (5779)
5302 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5303 - fix several locking issues related to i_size (5492,5624,5654,5672)
5304 - don't move pending lock onto export if it is already evicted (5683)
5305 - fix kernel oops when creating .foo in unlinked directory (5548)
5306 - fix deadlock in obdfilter statistics vs. object create (5811)
5307 - use time_{before,after} to avoid timer jiffies wrap (5882)
5308 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5309 - Do not leak request if server was not able to process it (5154)
5310 - If mds_open unable to find parent dir, make that negative lookup(5154)
5311 - don't create new directories with extent-mapping (5909, 5936)
5313 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5314 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5315 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5316 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5317 - support for RHEL4 kernel on i686 (5773)
5318 - provide error messages when incompatible logs are encountered (5898)
5320 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5321 * version 1.4.0.10 (1.4.1 release candidate 1)
5323 - don't keep a lock reference when lock is not granted (4238)
5324 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5325 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5326 - drop import inflight refcount on signal_completed_replay error (5255)
5327 - unlock page after async write error during send (3677)
5328 - handle missing objects in filter_preprw_read properly (5265)
5329 - no transno return for symlink open, don't save no-trasno open (3440)
5330 - don't try to complete elan receive that already failed (4012)
5331 - free RPC server reply state on error (5406)
5332 - clean up thread from ptlrpc_start_thread() on error (5160)
5333 - readahead could read extra page into cache that wasn't ejected (5388)
5334 - prevent races in class_attach/setup/cleanup/detach (5260)
5335 - don't dereference de->d_inode after l_dput of de (5458)
5336 - use "int" for stripe value returned from lock_to_stripe (5544)
5337 - mballoc allocation and error-checking fixes in 2.6 (5504)
5338 - block device patches to fix I/O request sizes in 2.6 (5482)
5339 - look up hostnames for IB nals (5602)
5340 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5341 - don't start multiple acceptors for the same port (5277)
5342 - fix incorrect LASSERT in mds_getattr_name (5635)
5343 - export a proc file for general "ping" checking (5628)
5344 - fix "lfs check" to not block when the MDS is down (5628)
5346 - service request history (4965)
5347 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5348 - create an "evict_client" /proc entry on OSTs, like the MDS has
5349 - fix mount usage message, return errors per mount(8) (5168)
5350 - change grep [] to grep "[]" in tests so they work in more UMLs
5351 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5352 - remove ancient LOV_MAGIC_V0 EA support (5047)
5353 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5354 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5355 - allow readahead allocations to fail when low on memory (5383)
5356 - mmap locking landed again, after considerable improvement (2828)
5357 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5359 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5362 - send OST transaction number in read/write reply to free req (4966)
5363 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5364 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5365 - the watchdog thread now runs as interruptible (5246)
5366 - flock/lockf fixes (but it's still disabled, pending 5135)
5367 - don't use EXT3 constants in llite code (5094)
5368 - memory shortage at startup could cause assertion (5176)
5370 - reorganization of lov code
5371 - single portals codebase
5373 - add extents/mballoc support (5025)
5374 - direct I/O reads in the obdfilter (4048)
5375 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5377 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5380 - send OST transaction number in read/write reply to free req (4966)
5381 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5382 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5383 - the watchdog thread now runs as interruptible (5246)
5384 - handle missing objects in filter_preprw_read properly (5265)
5385 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5386 - drop import inflight refcount on signal_completed_replay error (5255)
5387 - unlock page after async write error during send (3677)
5388 - return original error code on reconstructed replies (3761)
5389 - no transno return for symlink open, don't save no-trasno open (3440)
5391 - add pid to ldlm debugging output (4922)
5392 - bump the watchdog timeouts -- we can't handle 30sec yet
5393 - extra debugging for orphan dentry/inode bug (5259)
5395 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5398 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5399 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5400 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5401 - fix NULL dereference in /proc/sys/portals/routes (4827)
5402 - allow failed mdc_close() operations to be interrupted (4561)
5403 - stop precreate on OST before MDS would time out on it (4778)
5404 - don't send partial-page writes before EOF from client (4410)
5405 - discard client grant for sub-page writes on large-page clients (4520)
5406 - don't free dentries not owned by NFS code, check generation (4806)
5407 - fix lsm leak if mds_create_objects() fails (4801)
5408 - limit debug_daemon file size, always print CERROR messages (4789)
5409 - use transno after validating reply (3892)
5410 - process timed out requests if import state changes (3754)
5411 - update mtime on OST during writes, return in glimpse (4829)
5412 - add mkfsoptions to LDAP (4679)
5413 - use ->max_readahead method instead of zapping global ra (5039)
5414 - don't interrupt __l_wait_event() during strace
5416 - add software watchdogs to catch hung threads quickly (4941)
5417 - make lustrefs init script start after nfs is mounted
5418 - fix CWARN/ERROR duplication (4930)
5419 - return async write errors to application if possible (2248)
5420 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5421 - print NAL number in %x format (4645)
5422 - update barely-supported suse-2.4.21-171 series (4842)
5423 - support for sles 9 %post scripts
5424 - support for building 2.6 kernel-source packages
5425 - support for sles km_* packages
5427 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5430 - ignore -ENOENT errors in osc_destroy (3639)
5431 - notify osc create thread that OSC is being cleaned up (4600)
5432 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5433 - reconstruct ost_handle() like mds_handle() (4657)
5434 - create a new thread to do import eviction to avoid deadlock (3969)
5435 - let lconf resolve symlinked-to devices (4629)
5436 - don't unlink "objects" from directory with default EA (4554)
5437 - hold socknal file ref over connect in case target is down (4394)
5438 - allow more than 32000 subdirectories in a single directory (3244)
5439 - fix blocks count for O_DIRECT writes (3751)
5440 - OST returns ENOSPC from object create when no space left (4539)
5441 - don't send truncate RPC if file size isn't changing (4410)
5442 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5443 - bind to privileged port in socknal and tcpnal (3689)
5445 - rate limit CERROR/CWARN console message to avoid overload (4519)
5446 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5447 - basic mmap support (3918)
5448 - kernel patch series update from b1_4 (4711)
5450 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5453 - avoid crash during MDS cleanup with OST shut down (2775)
5454 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5455 - don't use bad inodes on the MDS (3744)
5456 - dynamic object preallocation to improve recovery speed (4236)
5457 - don't hold spinlock over lock dumping or change debug flags (4401)
5458 - don't zero obd_dev when it is force cleaned (3651)
5459 - print grants to console if they go negative (4431)
5460 - "lctl deactivate" will stop automatic recovery attempts (3406)
5461 - look for existing locks in ldlm_handle_enqueue() (3764)
5462 - don't resolve lock handle twice in recovery avoiding race (4401)
5463 - revalidate should check working dir is a directory (4134)
5465 - don't always mark "slow" obdfilter messages as errors (4418)
5467 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5470 - don't close LustreDB during write_conf until it is done (3860)
5471 - fix typo in lconf for_each_profile (3821)
5472 - allow dumping logs from multiple threads at one time (3820)
5473 - don't allow multiple threads in OSC recovery (3812)
5474 - fix debug_size parameters (3864)
5475 - fix mds_postrecov to initialize import for llog ctxt (3121)
5476 - replace config semaphore with spinlock (3306)
5477 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5478 - don't allow enqueue to complete on a destroyed export (3822)
5479 - down write_lock before checking llog header bitmap (3825)
5480 - recover from lock replay timeout (3764)
5481 - up llog sem before sending rpc (3652)
5482 - reduce ns lock hold times when setting kms (3267)
5483 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5484 - fix NULL deref and obd_dev leak on setup error (3312)
5485 - replace some LBUG about llog ops with error handling (3841)
5486 - don't match INVALID dentries from d_lookup and spin (3784)
5487 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5488 - fix invalid assertion in ptlrpc_set_wait (3880)
5490 - add libwrap support for the TCP acceptor (3996)
5491 - add /proc/sys/portals/routes for non-root route listing (3994)
5492 - allow setting MDS UUID in .xml (2580)
5493 - print the stack of a process that LBUGs (4228)
5495 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5498 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5499 - make sure to unset replay flag from failed open requests (3440)
5500 - if default stripe count is 0, use OST count for inode size (3636)
5501 - update parent mtime/ctime on client for create/unlink (2611)
5502 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5503 - free recovery state on server during a forced cleanup (3571)
5504 - unregister_reply for resent reqs (3063)
5505 - loop back devices mounting and status check on 2.6 (3563)
5506 - fix resource-creation race that can provoke i_size == 0 (3513)
5507 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5508 - more debugging for page-accounting assertion (3746)
5509 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5510 - avoid deadlock after precreation failure (3758)
5511 - fix race and lock order deadlock in orphan handling (3450, 3750)
5512 - add validity checks when grabbing inodes from l_ast_data (3599)
5514 - add /proc/.../recovery_status to obdfilter (3428)
5515 - lightweight CDEBUG infrastructure, debug daemon (3668)
5516 - change default OSC RPC parameters to be better on small clusters
5517 - turn off OST read cache for files smaller than 32MB
5518 - install man pages and include them in rpms (3100)
5519 - add new init script for (un)mounting lustre filesystems (2593)
5520 - run chkconfig in %post for init scripts (3701)
5521 - drop scimac NAL (unmaintained)
5523 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5526 - clean kiobufs before and after use (3485)
5527 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5528 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5529 - change init script to fail more clearly if not run as root (1528)
5530 - allow clients to reconnect during replay (1742)
5531 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5532 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5533 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5534 - fix cond_resched (3554)
5535 - only evict unfinished clients after recovery (3515)
5536 - allow bulk resend, prevent data loss (3570)
5537 - dynamic ptlrpc request buffer allocation (2102)
5538 - don't allow unlinking open directory if it isn't empty (2904)
5539 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5540 - remove extraneous obd dereference causing LASSERT failure (3334)
5541 - don't use get_cycles() when creating temp. files on the mds (3156)
5542 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5543 - handle EEXIST for set-stripe, set proper directory name (3336)
5545 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5546 - fix ksocknal_fmb_callback() error messages (2918)
5548 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5551 - don't copy lvb into (possibly NULL) reply on error (2983)
5552 - don't deref dentry after dput, don't free lvb on error (2922)
5553 - use the kms to determine writeback rpc length (2947)
5554 - increment oti_logcookies when osc is inactive (2948)
5555 - update client's i_blocks count via lvb messages (2543)
5556 - handle intent open/close of special files properly (1557)
5557 - mount MDS with errors=remount-ro, like obdfilter (2009)
5558 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5559 - don't use cancelling-locks' kms values (2947)
5560 - use highest lock extent for kms, not last one (2925)
5561 - don't dereference ERR_PTR() dentry in error handling path (3107)
5562 - fix thread race in portals_debug_dumplog() (3122)
5563 - create lprocfs device entries at setup instead of at attach (1519)
5564 - common AST error handler, don't evict client on completion race (3145)
5565 - zero nameidata in detach_mnt in 2.6 (3118)
5566 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5567 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5568 - initialize RPC timeout timer earlier for 2.6 (3219)
5569 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5570 - print nal/nid for unknown nid (3258)
5571 - additional checks for oscc recovery before doing precreate (3284)
5572 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5573 - don't crash in mdc_close for bad permissions on open (3285)
5574 - zero i_rdev for non-device files (3147)
5575 - clear page->private before handing to FS, better assertion (3119)
5576 - tune the read pipeline (3236)
5577 - fix incorrect decref of invalidated dentry (2350)
5578 - provide read-ahead stats and refine rpc in flight stats (3328)
5579 - don't hold journal transaction open across create RPC (3313)
5580 - update atime on MDS at close time (3265)
5581 - close LDAP connection when recovering to avoid server load (3315)
5582 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5583 - don't leak open file on MDS after open resend (3325)
5584 - serialize filter_precreate and filter_destroy_precreated (3329)
5585 - loop device shouldn't call sync_dev() for nul device (3092)
5586 - clear page cache after eviction (2766)
5587 - resynchronize MDS->OST in background (2824)
5588 - refuse to mount the same filesystem twice on same mountpoint (3394)
5589 - allow llmount to create routes for mounting behind routers (3320)
5590 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5591 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5592 - fix rare problem with rename on htree directories (3417)
5594 - allow default OST striping configuration per directory (1414)
5595 - fix compilation for qswnal for 2.6 kernels (3125)
5596 - increase maximum number of MDS request buffers for large systems
5597 - change liblustreapi to be useful for external progs like lfsck (3098)
5598 - increase local configuration timeout for slow disks (3353)
5599 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5601 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5604 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5605 - don't overwrite extent policy data in reply if lock was blocked (2901)
5606 - drop filter export grants atomically with removal from device (2663)
5607 - del obd_self_export from work_list in class_disconnect_exports (2908)
5608 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5609 - swab reply message in mdc_close, other PPC fixes (2464)
5610 - fix destroying of named logs (2325)
5611 - overwrite old logs when running lconf --write_conf (2264)
5612 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5613 - fix race in target_handle_connect (2898)
5614 - mds_reint_create() should take same inode create lock (2926)
5615 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5616 - don't close files for self_export to avoid uninitialized obd (2936)
5617 - allow MDS with the same name as client node (2939)
5618 - hold dentry reference for closed log files for unlink (2325)
5619 - reserve space for all logs during transactions (2059)
5620 - don't evict page beyond end of stripe extent (2925)
5621 - don't oops on a deleted current working directory (2399)
5622 - handle hard links to targets without a parent properly (2517)
5623 - don't dereference NULL lock when racing during eviction (2867)
5624 - don't grow lock extents when lots of conflicting locks (2919)
5626 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5629 - account for cache space usage on clients to avoid data loss (974)
5630 - lfsck support in lustre kernel code (2349)
5631 - reduce journal credits needed for BRW writes (2370)
5632 - orphan handling to avoid losing space on client/server crashes
5633 - ptlrpcd can be blocked, stopping ALL progress (2477)
5634 - use lock value blocks to assist in proper KMS, faster stat (1021)
5635 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5636 - recovery for initial connections (2355)
5637 - fixes for mds_cleanup_orphans (1934)
5638 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5639 - block all file creations until orphan recovery completes (1901)
5640 - client remove rq_connection from request struct (2423)
5641 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5642 - recovery timer race (2670)
5643 - mdc_close recovey bug (2532)
5644 - ptlrpc cleanup bug (2710)
5645 - mds timeout on local locks (2588)
5646 - namespace lock held during RPCs (2431)
5647 - handle interrupted sync write properly (2503)
5648 - don't try to handle a message that hasn't been replied to (2699)
5649 - client assert failure during cleanup after abort recovery (2701)
5650 - leak mdc device after failed mount (2712)
5651 - ptlrpc_check_set allows timedout requests to complete (2714)
5652 - wait for inflight reqs when ptlrpcd finishes (2710)
5653 - make sure unregistered services are removed from the srv_list
5654 - reset bulk XID's when resending them (caught by 1138 test)
5655 - unregister_bulk after timeout
5656 - fix lconf error (2694)
5657 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5658 - readahead locks pages, leaves pending causing memory pressure (2673)
5659 - increase OST request buffers to 4096 on large machines (2729)
5660 - fix up permission of existing directories in simple_mkdir (2661)
5661 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5662 - don't assign transno to errored transactions (2742)
5663 - don't delete objects on OST if given a bogus objid from MDS (2751)
5664 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5665 - if rq_no_resend, then timeout request after recovery (2432)
5666 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5667 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5668 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5669 - MDS recovery shouldn't delete objects that it has given out (2730)
5670 - if enqueue arrives after completion, don't clobber LVB (2819)
5671 - don't unlock pages twice when trigger_group_io returns error (2814)
5672 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5673 - don't write pages to disk if there was an error (1450)
5674 - don't ping imports that have recovery disabled (2676)
5675 - take buffered bytes into account when balancing socknal conn (2817)
5676 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5677 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5678 - remove little-endian swabbing of llog records (1987)
5679 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5680 - retry reposting request buffers if they fail (1191)
5681 - grow extent at grant time to avoid granting a revoked lock (2809)
5682 - lock revoke doesn't evict page if covered by a second lock (2765)
5683 - disable VM readahead to avoid reading outside lock extents (2805)
5685 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5686 - updated kernel patches for hp-2.4.20 kernel (2681)
5688 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5691 - fix truncated write corruption (2366)
5692 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5694 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5695 - ensure len(uuid) < 37 in lmc (1171)
5696 - fix ia64 OOPS in llog_test (2255)
5697 - zero end of page at obdfilter for partial page writes (2648)
5698 - don't leave stale dentries around after renames (bug 2428)
5699 - fix timeouts when evicting a client with a single lock held (2642)
5700 - set deadline for the initial HELLO message to drain (2634)
5701 - print out dotted-quad IP addresses in the socknal (2302)
5703 - additional debugging for MDS client eviction problem (2443)
5704 - fix mkfsoptions support for osts (2603, 2604)
5706 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5709 - add series for the vanilla 2.6.0 kernel
5710 - add series for the vanilla 2.4.24 kernel
5711 - add series for a cray x86/64 UL kernel drop
5712 - fix xattr patches for the vanilla 2.4.19 series
5714 - generate true UUIDs in lmc (1171)
5715 - have portals stack dumping break in UML (2466)
5716 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5717 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5718 - wake callback waiting threads on client eviction (2460)
5719 - Add --ptldebug and --subsystem to lmc (1719)
5720 - update assertion to allow safe interrupt allocation
5721 - set rq_no_resend for cancel requests (2432)
5722 - recalculate ptlrpcd timeout after resend (2494)
5723 - call vfs_rmdir when removing pending directories (2368)
5724 - fix renaming a file to itself (2429)
5725 - lmc creates a default one-stripe lov (2454)
5726 - expand procfs space to handle large clusters (2326)
5727 - increase UML stack to avoid overflow
5728 - update lconf's list of debug and subsystem masks
5729 - fix lfs find --obd (2510)
5730 - /proc tunable for disabling filter read caching (2591)
5731 - stop rpm packages from altering slapd.conf (2301)
5732 - disable nagle in the socknal under 0conf (2578)
5733 - choose mds inode size based on stripe count (2572)
5734 - fix kernel-source rpm problems (2516)
5736 - add --disable-doc to avoid pdf generation (2421)
5737 - update documentation, tests, type-os, comments
5738 - avoid format warnings on ia64
5739 - remove the TOE NAL
5740 - tiny code cleanups by removing unused fields
5742 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5745 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5746 - remove the most blatant lies from BUILDING file (2371)
5747 - change default debug level to reasonable production setting
5748 - reduce client side cache size to reduce cache flush time
5749 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5750 - make TCP zerocopy and pinger support enabled by default (2476)
5751 - sync writes completed after process exits caused crashes (2319)
5752 - maintain correct mount count on the MDS (2356)
5753 - backout 1557, because 2316 wasn't really fixed
5754 - better file I/O statistics gathering in /proc
5755 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5756 - another kernel patch to fix zero-copy TCP function export
5757 - don't take duplicate lock when processing re-sent getattr (2420)
5758 - lctl uses obd_self_export instead of creating new conn (2353)
5759 - MDS/OST recovery case which requires object creation asserted (2425)
5760 - move lfs from /usr/sbin to /usr/bin in packages
5761 - fix race between mds_client_add and mds_client_free (2417)
5762 - use kmalloc instead of slabs in portals (2430)
5763 - don't create duplicate records when a failover MDS is present (2442)
5764 - remove unnecessary mount age check (2332)
5765 - don't remove directory inodes from locks prematurely (2451)
5766 - don't break if MDS service name is the same as hostname (2103)
5767 - fix races in client write RPC generation when cache full (2482)
5769 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5772 - remove now-unused request->rq_obd (278)
5773 - if an allocation fails, print out how much memory we've used (1933)
5774 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5775 - add the "configurable stack size" patch to most series files (1256)
5776 - ability to write large log records, for 100+ OST configs (2306)
5777 - fix NULL deref when filter_prep fails (2314)
5778 - fix operator precedence error in filter_sync
5779 - dynamic allocation of socknal TX descriptors (2315)
5780 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5781 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5782 - ensure that CWARN messages in Portals always get to the syslog
5783 - __init/__exit are not for prototype decls (ldlm_init/exit)
5784 - x86-64 compile warning fixes
5785 - fix gateway LMC keyword conflict (2318)
5786 - fix MDS lock inversions in getattr/reint paths (1844)
5787 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5788 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5789 - DLM race condition prevented some lock evictions (2328)
5790 - ENOMEM detection and retry on socknal sends (2230)
5791 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5792 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5793 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5794 - better error messages when a client is rejected during recovery (1505)
5795 - avoid cancelling locks which were never granted, after failure (2330)
5796 - fix i_sem/journal inversion in mds_client_add (2333)
5797 - fix truncate/getattr lock cycle deadlock (2334)
5798 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5799 - fix two rare exit paths which could leak an l_lock() ref (2321)
5800 - fencepost error in MDS/OST orphan recovery (2226)
5801 - make log record alignment 8 bytes (1988)
5802 - lstripe now fails when requested offset > ost_count (2237)
5803 - ensure that all kernel series have a complete list.h (1607)
5804 - fix crashes in special-file operations (2316)
5805 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5807 - allow configurable automake binary, for testing new versions
5808 - small update to the lfs documentation
5810 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5812 * fix negative export reference count in fsfilt_sync (2312)
5814 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5815 * release candidate 0.9.1
5817 - orphans are moved into the PENDING directory for possible recovery
5818 - replayed opens now open by fid for orphan/rename safety (1042)
5819 - last close of an orphan inode generates a transno (683)
5820 - chdir() and mount() now pin the directory entry (1020)
5821 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5822 - discard very old requests without processing them (1502)
5823 - remove some common, well-understood CERRORs (1505)
5824 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5825 - clear "grant" flags in OST replies until OST grant code lands (1644)
5826 - fix read performance by not clobbering i_blksize on client (1598)
5827 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5828 - make LRU size a /proc tunable, clears locks when reduced (707)
5829 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5830 - server locks take references on exports now (1558)
5831 - build fixes for 2.4.20-rh trees (1663)
5832 - return an error from lov_create if all OSCs are inactive (1751)
5833 - fix import levels when a reconnect happens without a timeout (1597)
5834 - exit early from mds_open if we get a lookup error (1749)
5835 - partial page read at EOF wouldn't wait for disk before sending (1642)
5836 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5837 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5838 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5839 - write/truncate lock inversion (1639)
5840 - Don't auto-load obdclass, portals modules during cleanup (1495)
5841 - fix timestamps from jumping to "now" (1763)
5842 - extra journal assertions (1648)
5843 - add an extra multiunlink test (1771)
5844 - fix read_record/write_record API (1776)
5845 - fix leak of offset_extent, possible incorrect i_size later (1772)
5846 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5847 - Debugging for the kqswnal_get_idle_tx problems (1820)
5848 - Allow recovery to be attempted multiple times (1536)
5849 - Write out MDS last_rcvd file after it is first created (1600)
5850 - Fix tx_descriptor leak in failed transmit situations (1827)
5851 - ext3 journaling fixes for assertion failure after IO error (1871)
5852 - class_export_put() on freed export after completion AST error (1896)
5853 - Fix revalidate looping in VFS (1322)
5854 - Don't access a freed export during MDS_REINT timeout (1521)
5855 - Add open-unlink recovery support on the MDS (1673,1764)
5856 - Return an error if no MDS data was read from last_rcvd (1946)
5857 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5858 - Don't setup a disk device that doesn't match exported UUID (317)
5859 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5860 - avoid committing NULL handle in force close
5861 - local.sh is now a one-stripe LOV configuration
5862 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5863 - fix ext3 htree duplicate directory entry corruption (1516)
5864 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5865 - update to new LOV EA format (2097)
5866 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5868 2003-06-15 Phil Schwan <phil@clusterfs.com>
5871 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5872 - per-import/export recovery handling (958, 931, 959)
5873 - multiple last-rcvd slots, for serving multiple FSes (949)
5874 - connections are again shared between multiple imp/exports (963, 964)
5875 - "umount -f" would hang if any requests needed to be sent (393, 978)
5876 - avoid pinning large req buffer by copying for queued messages (989)
5877 - add "uuid" to "lctl device" command to help upcalls (991)
5878 - "open" RPCs with transnos would confuse recovery counters (1037)
5879 - do proper endian conversion of all wire messages (288, 340, 891)
5880 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5881 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5882 - fix ptlrpc_connection leak in target_handle_connect (1174)
5883 - fix import refcounting bug in OST and MDS cleanup (1134)
5884 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5885 - fix very unlikely obd_types race condition (501)
5886 - remove osc_open hack for echo_client (1187)
5887 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5888 - a failure in read_inode2 leads to deadlock (1139)
5889 - cancel ack-locks as soon as transaction is committed (1072)
5890 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5891 - make sure to commitrw after any preprw to avoid deadlock (1162)
5892 - failing to execute a file in a lustre FS would lock inode (1203)
5893 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5894 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5895 - obd_uuid2tgt would walk off the end of the list (1255)
5896 - on IA64 the osc would give portals incorrect bulk size (1258)
5897 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5898 - fix lock inversion caused by new llite matching code (1282)
5899 - limit the number of dirty pages on a client to 10MB (1286)
5900 - timed out locks were not being corrected cancelled (1289)
5901 - fix O_DIRECT above 4GB on IA-32 (1292)
5902 * major user-visible changes
5903 - fail out/fail over policy now controlled by the upcall (993)
5905 - add OBD_PING to check server availability and failure (954)
5906 - lustre messages are now sent in sending host order (288, 340, 891)
5907 - add eadatalen to MDS getattr reply (340)
5908 - OST read replies may contain second buffer, with per-page status (593)
5910 2003-03-11 Phil Schwan <phil@clusterfs.com>
5913 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5914 - failed open()s could cause deadlock; fixed (867, 869)
5915 - stop cancelling OST locks when files are closed (481)
5916 - overlapping XID spaces caused network corruption (851, 853)
5917 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5918 - setattr_raw added, to do single-RPC, server-side setattrs
5919 - lmc/lconf syntax change for OST UUIDs
5920 - fix crashy race condition between ptlrpc_free_req and osc_close
5921 - don't use request in mdc_enqueue if we hit a timeout (889)
5922 - don't set the inode i_size for regular files from the MDS (896)
5923 - handle out of order completion AST (842)
5924 - don't LBUG if a lock request times out after receiving AST (913)
5925 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5926 - if a bad lock AST arrives, send an error instead of dropping entirely
5927 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5928 - fix leak in bulk IO when only partially completed (899, 900, 926)
5929 - fix O_DIRECT for ia64 (55)
5930 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5931 - C-z after timeout could hang a process forever; fixed (977)
5933 - client-side I/O cache (678, 924, 929, 941, 970)
5935 - READPAGE and SETATTRs which don't take server-side locks get
5938 2003-02-11 Phil Schwan <phil@clusterfs.com>
5941 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5942 - Fix fsfilt_extN_readpage() to read a full page of directory
5943 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5944 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5945 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5946 - Most intent operations were removed, in favour of a new RPC mode
5947 that does a single RPC to the server and bypasses most of the VFS
5948 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5949 - Aggressively cancel local locks on DLM servers
5950 - mds_reint_unlink sends EA to the client if it's the last nlink.
5951 client uses that EA to unlink OST objects.
5952 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5953 - recursive symlinks were fixed (439)
5954 - fixed NULL deref in DEBUG_REQ
5955 - filter_update_lastobjid no longer calls sync, which annoyed extN
5956 - fixed multi-client small-writes to a single file problem (445)
5957 - fixed mtime updates during file writes (607)
5958 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5959 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5960 - fixed Linux OST size reporting problem (444, 656)
5961 - OST now updates object mtime with writes or setattr (607, 619)
5962 - client verifies file size before zeroing page past EOF (445)
5963 - OST now writes last allocated objid to disk with allocation (108)
5964 - LOV on echo now works (409)
5966 - mds_reint_unlink sends a new buffer, with the EA included. this
5967 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5968 set if a regular file was being unlinked, and it was the last link
5969 - use PtlGet from the target for bulk writes (315)
5970 - OST now updates object mtime with writes or setattr (607, 619)
5971 - LDLM now has a grant-time callback to revalidate locked items, if
5973 - Many MDS operations were reorganized to combat race conditions
5975 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5976 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5978 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5981 - Fully reactivate OST imports after reconnection (512, others)
5982 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5983 - More graceful error handling for truncating on dead OST (515)
5984 - Don't error out unless we're actually accessing dead stripes (474)
5985 - Fix garbage sizes when stripes are missing (410)
5986 - LRU counters were broken, causing constant lock purge (433, 432)
5987 - garbage on read from stripes with failed OSTs (441)
5988 - mark OSCs as active before reconnecting during recovery (438)
5989 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5990 - lfind did not preserve OST order in output (443)
5991 - symlinks cause hung clients, incorrect data (439)
5992 - stop dereferencing request after dropping refcount (457)
5993 - don't LASSERT(spin_is_locked) on non-SMP (455)
5994 - fixes for many rename() bugs
5995 - fstat didn't correctly synchronize attributes (399)
5996 - server must handle lock cancellation during blocking AST prep (487)
5997 - bulk descriptors were free()d too soon (511)
5998 - fix paths in lconf, which would load incorrect modules (451, 507)
5999 - fix confusing lconf 'host not found' error message (386)
6000 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6001 - fix race condition in mdc_blocking_ast() for inode access (526)
6002 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6003 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6004 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6005 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6006 - fix lstripe in several ways (kernel side) (527)
6007 - fix request leak in ldlm_cli_enqueue (262)
6008 - incorrect OSC was marked inactive after OST failure
6009 - call mds_fs_cleanup before unmounting filesystem (524)
6010 - fix races between taking ns_lock and ldlm_lock_change_resource
6011 - fix races updating LOV export open file list
6012 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6013 - fix recovery NULL deref in ldlm_cli_cancel_unused
6014 - fix some DLM races by using new hash table for lock handles (419)
6015 - permit the client to specify desired inodes, at replay
6016 - duplicate requests when we queue them for replay reintegration
6017 - fix last_rcvd offset calculation
6018 - sync after each recovered transaction, so we always make progress
6019 - never, not always, ERESTART requests without transnos
6020 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6022 - skip replay if the MDS says that the client is already connected
6023 - don't check for a recovery-enabled export to match lctl's UUID
6024 - don't INC_USE_COUNT for phantom exports
6025 - don't crash when cleaning up phantom exports (567)
6026 - don't double-finish or set replay data for errored mdc_open requests
6027 - abort requests when they time out, so we don't get old replies
6028 - send/receive replies for AST messages again
6029 - if the client says that it doesn't have the lock, cancel it on the
6031 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6032 mark it as destroyed, it will all work out when decref is called
6033 - fix module use counts (22, 581)
6035 - ASTs now expect a reply (server cancels lock on error reply)
6037 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6040 - fix many simultaneous client startup (392)
6041 - fix dentry->d_it clobbering
6042 - credentials weren't being shipped for readdir/getattr operations
6043 - remove invalid assertions triggered during some concurrent MD
6045 - proper Lustre versions added (336, 389)
6046 - fix memory leak for create error case (398)
6047 - fix LOV locking bug that would get cli/srv out of sync
6048 - fix echo client over LOV (409)
6049 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6050 - fix double-O_EXCL intent crash (424)
6051 - avoid sending multiple lock CANCELs (352)
6053 - MDS can do multi-client recovery (modulo bugs in new code)
6055 - many updates, edits, cleanups
6057 2002-11-18 Phil Schwan <phil@clusterfs.com>
6060 - fix null d_it dereference (346)
6061 - fix full OST/dbench hang (333)
6062 - fix permission problem with file removal (286)
6063 - fix removal of OSCs from LOV when they fail
6064 - fix NULL deref during bulk timeout (214)
6065 - fix problems related to multiple filesystems on one MDS (241)
6066 - fixed serious subtle metadata locking bugs
6067 - free locks on clients when inodes are removed due to memory
6069 - fix inode pointer in lock data (285)
6070 - partial support for multiple MDS on a single host (241)
6071 - data locks weren't cancelled at clear_inode time (290, 311)
6072 - intent locks could lead to unbounded lock growth (205)
6073 - added a maximum lock count, an LRU list, and a flusher
6074 - fix multiple rename (365)
6075 - properly abstracted the echo client
6076 - OSC locked 1 byte too many; fixed
6077 - rewrote brw callback code:
6078 - fixed recovery bugs related to LOVs (306)
6079 - fixed too-many-pages-in-one-write crash (191)
6080 - fixed (again) crash in sync_io_timeout (214)
6081 - probably fixed callback-related race (385)
6083 - Add capability to MDS protocol
6084 - LDLM cancellations and callbacks on different portals
6086 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6089 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6090 - defer open object destruction to close time (601981, 138)
6091 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6092 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6093 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6094 - don't call obd_setattr in ll_file_release for destroyed objects
6095 * protocol change to lustre_msg: move |version| and add |flags|
6096 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6097 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6098 * added replay of create, unlink, link and rename operations during
6099 MDS failover; recovery should be much more robust now
6100 * remove failed OSCs from LOVs (only lov_create uses this so far)
6101 * the lustre-HOWTO was brought (more) up to date (582544)
6103 2002-10-23 Phil Schwan <phil@clusterfs.com>
6106 - in-use dentries weren't being reused properly (617851)
6107 - prevent multiple LDLM setup (599178)
6108 - fix LOV size calculations for truncate (617853)
6109 - fix client handling of MDS intent errors (POSIX)
6110 - fix permission bug in lovstripe.c test (624321)
6111 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6112 - truncate past end of file could corrupt data
6113 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6114 - a race in recovery could return ETIMEDOUT to apps (623947)
6115 - building outside the source directory was fixed
6116 * the lustre-HOWTO was brought (more) up to date (582544)
6117 * major progress was made on recovery functionality
6119 2002-10-10 Phil Schwan <phil@clusterfs.com>
6122 - recovery deadlock fix
6123 - rm -rf causes LBUG fix (617817)
6124 - file open by multiple tasks fix (618962)
6125 - directory permissions bugs (602707 and 620007)
6126 - journal_stop fixed with locking (611313)
6127 - O_APPEND failures resolved (618273, perhaps 614459)
6128 - lconf PATH fix (619770)
6129 - IA64 build fix (621450)
6130 - RPC buffer sizes scale with amount of memory
6132 2002-10-01 Phil Schwan <phil@clusterfs.com>
6135 - locks would be cancelled without throwing away data pages,
6136 resulting in inconsistent data (605627)
6137 - inode attributes were not always being refreshed (605627, 612449)
6138 - lconf now continues to cleanup after lctl reports an error
6139 - MDS now enforces user permissions (602707)
6140 - lprocfs cleanup fixed, but not yet enabled (614157)
6141 - fixed infinite server hang, should a client not respond to an AST
6142 - avoid going into recovery if user calls readlink() with a buffer
6143 that's too small (613941)
6144 - AST RPCs no longer require replies (614867) -- this may be changed
6145 - don't crash server if client sends an IOV that's too big (611336)
6146 - fixed lock conversion deadlock (611892)
6147 - fixed the following of symlinks (614622)
6148 * recovery: the server can remove locks from a client that dies, other
6149 clients can make progress
6150 * more extN patch fixes
6151 * compile-time configurable ptlrpc buffer allocations
6153 - collaborative read cache document
6154 - Lustre Lite Performance CDR document-in-progress
6156 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6159 - fix typo in patch-2.4.18
6161 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6164 - clear ptlrpc request each time in handle_incoming_request()
6165 - unlink of files now destroys the object on the OST
6167 2002-09-19 Peter Braam <braam@clusterfs.com>
6169 * add hard link support
6170 * change obdfile creation method
6171 * kernel patch changed
6173 2002-09-19 Peter Braam <braam@clusterfs.com>
6176 - stack overflow bug in extN fixed
6178 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6180 * documentation updates
6181 - add man pages for config tools
6182 - update tests/README to describe testing with new config tools
6183 - finish metadata API descriptions
6184 * bug fixes and cleanups
6185 - statfs workaround for 16TB limit
6186 - LOV stripe allocation improved, can stripe on subset of OSTs
6187 - LOV file size/IO offset was wrong for files > 4GB in size
6188 - object EA data was being dropped, caused files to be unreadable
6189 - memory overflow with non-LOV OST caused memory corruption
6190 - fixed regression tests to work with new config tools, obdfilter
6191 - fixed bug when directory size became larger than 1 block
6192 - fixed bug (for single client case) when PWD was deleted
6193 - invalidate local directory pages when doing intent-based ops
6194 - avoid LDLM oops when lock callback contained bad data
6196 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6198 * documentation updates
6199 * bug fixes and cleanups
6200 - configuration tools
6203 - 64-bit compile warnings
6204 - 64-bit internal statfs data
6206 * test_brw on persistent OST devices
6208 * lprocfs (disabled)
6210 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6212 * documentation updates
6213 * bug fixes and cleanups
6214 * configuration tools
6216 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6219 * many small fixes to 0_5_4
6220 * io/network handling
6221 * thinkos in MDS operations
6223 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6226 * crucial basic fixes to 0.5.3
6227 * IOR, Iozone work over Elan
6230 2002-08-07 Phil Schwan <phil@clusterfs.com>
6231 * version 0_5_3, our first alpha
6232 * we use the new Portals iovs
6233 * documentation updates
6234 * bug fixes and cleanups
6235 * small changes in the DLM wire protocol
6237 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6238 * version 0_5_1 with some initial stability,
6239 * locking on MD and file I/O.
6240 * documentation updates
6241 * several bug fixes since 0.5.0
6242 * small changes in wire protocol
6244 2002-07-18 Phil Schwan <phil@clusterfs.com>
6246 * delivered as Lustre Light Alpha
6247 * fixed a crash after handling invalid MDS requests
6248 * fixed directory pages for architectures with non-4k pages sizes
6250 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6251 * release version v0_4_4
6252 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6254 * Fixes a number of interruption problems with OST operations.
6255 * Update documentation for portals header changes
6256 * Move all wire protocol structs/defines to lustre_idl.h
6257 * Fixes symlink length bug.
6258 * Add tcpdump to repository.
6260 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6261 * release version v0_4_3
6262 * Fixes statfs for inodes on extN.
6263 * Fixes bug in runtests which would delete /etc/hosts.
6264 * Use 64-bit object IDs wherever possible (not into VFS though)
6265 Remove ost_get_info, which is unused by lustre, and out of date.
6267 2002-07-03 Peter Braam <braam@clusterfs.com>
6268 * release version v0_4_2 Fixes a lookup error (type not passed)
6269 * move forward to head of Portals
6270 * move forward to latest Lustre kernel
6272 2002-06-25 Peter Braam <braam@clusterfs.com>
6273 * release version v0_4_1. Hopefully stable on single node use.