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
15 --------------------------------------------------------------------------------
17 2010-07-15 Oracle, Inc.
19 * Server support for kernels:
20 2.6.18-164.11.1.el5 (RHEL 5)
21 2.6.18-164.11.1.0.1.el5 (OEL 5)
22 * Client support for unpatched kernels:
23 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
24 2.6.18-164.11.1.el5 (RHEL 5),
25 2.6.18-164.11.1.0.1.el5 (OEL 5)
26 2.6.16.60-0.42.8 (SLES 10),
28 2.6.29.4-167.fc11 (FC11)
29 * Recommended e2fsprogs version: 1.41.10-sun2
31 Severity : enhancement
33 Description: Distribute lfs_migrate script with lustre
34 Details : The lfs_migrate mentioned in the manual is now included in the
35 lustre tree in lustre/scripts and will distributed in lustre RPMs.
38 Severity : enhancement
40 Description: Failover nid list added to import proc file.
42 Severity : enhancement
44 Description: "som_preview" mount option required to use SOM
45 Details : The clients will need to use som_preview mount option
49 Description: Lustre does not do 1MB IOs to HW RAID
50 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
52 Severity : enhancement
54 Description: lfs quota output is not very convenient for awk/sed-parsing
55 Details : Some positions in lfs quota output table could be empty or
56 non-empty which made it hard to parse it with scripts, now
57 a dash is put instead of space where there is not supposed
62 Description: Update to OFED 1.5.1
63 Details : OFED 1.5.1 will be used in environments that don't use
68 Description: Remove lnet. prefix from lctl params display
69 Details : Also renamed lnet.memused to lnet_memused to
74 Description: Use conf_param -d to remove permanant settings.
78 Description: Add list_param -R
79 Details : lctl list_param now supports recursive mode.
82 Bugzilla : 21829, 22312
83 Description: Fix llobdstat.
84 Details : 1. Now updating the write_bytes counter
85 2. Port script fixes from 1.8.
89 Description: decrease the usage of memory on clients.
90 Details : 1. On clients, recycle dentries and inodes unused.
91 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
96 Description: llobdstat fix and enhancement
97 Details : add a counter to set a limit to how many samples will be returned
98 fix a wildcard in the path to limit to obdfilter stats only
102 Description: lfs check only allowed for root
103 Details : Code cleanups and added a regresion test of non-root use of
108 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
109 Details : Fix hr->hr_index overload.
113 Description: Fix FIEMAP functionality.
117 Description: FC11 patchless client support added
121 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
122 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
126 Description: New acc-sm script and various test-framework enhancments.
130 Description: Previously landed patch was reverted.
134 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
138 Description: Add ldiskfs maxdirsize mount option.
142 Description: Fix mkfs.lustre support for 16TB LUNs.
146 Description: Print mmp_check_interval
147 Details : print mmp_check_interval and make it possible to abort
148 mount operation in case it takes too long.
152 Description: Add cascading_rw to lustre mpi tests.
156 Description: Store SOM data in EA.
157 Details : Also several SOM related cleanups.
161 Description: Add refcount to md_open_data
165 Description: Use refcount on lov_request_set to prevent memory corruption.
169 Description: Statahead: use parent dir's i_mutex for synchronization.
173 Description: loadgen improvements.
177 Description: Enable peer health detection in ksocklnd and ko2iblnd.
181 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
185 Description: Fix race on obd_devcie:md_stats
189 Description: Fix SLES 11 client builds.
190 Details : Resolved uninitalized pointer gcc warnings.
194 Description: Allow non-root access for "lfs check".
195 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
199 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
201 Severity : enhancement
203 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
204 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
208 Description: Sanity test_27p fails.
209 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
210 Sanity tests 27 were cleaned up and improved.
214 Description: Unify req format on client/servers
215 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
216 buffers and swabbers approach..
221 Description: drop join file stuff.
223 Severity : enhancement
225 Description: Add 2.6.27 and SLES11 (patchless client) support
229 Description: Fix in ptlrpc_expire_one_request() to print the signed time
232 Severity : enhancement
234 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
237 Severity : enhancement
239 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
241 Severity : enhancement
243 Description: correctly handle big reply message.
244 Details : send LNet event if reply is bigger then buffer and adjust this buffer
249 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
251 Severity : enhancement
253 Description: Add support for OEL5.
257 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
261 Description: client could not reconnect to OST because of an active request.
262 Details : The client now retries to reconnect to the same server, if a
263 connect request failed with EBUSY or -EAGAIN. Test case 35b
264 conf-sanity.sh has been created.
268 Description: Don't allow make backward step on assignin osc next id.
269 Details : race between allocation next id and ll_sync thread can be cause
270 of set wrong osc next id and can be kill valid ost objects.
274 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
278 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
282 Description: Error handling in osc_statfs_interpret() has been improved.
283 Details : Check in osc_statfs_interpret() for EBADR.
287 Description: Do not send statfs() requests to OSTs disabled by administrator.
288 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
290 Severity : enhancement
292 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
294 Severity : enhancement
296 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
298 Severity : enhancement
300 Description: Enhancement for lfs(1) command to use numeric uid/gid.
304 Description: Improved cleanup procedure for conf-sanity test_46a.
308 Description: Do not start bulk transfer, if client tried to reconnect.
309 Details : Do not start bulk transfer, if client tried to reconnect,
310 but failed due to active rpcs. Check for exp_abort_active_req
311 has been implemented.
315 Description: Prevent inconsistences between linux and lustre mount structures.
316 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
317 Make the sleep interruptible.
319 Severity : enhancement
321 Description: provide server to client comms path
322 Details : server to client communications path via new LDLM_SET_INFO rpc
326 Description: truncate starts GFP_FS allocation under transaction causing deadlock
327 Details : ldiskfs_truncate calls grab_cache_page which may start page
328 allocation under an open transaction. This may lead to
329 calling prune_icache with consequent lustre reentrance.
333 Description: send correctly lsm on open replay
334 Details : MDS is trust to LSM size on replay open, but client can set wrong size
339 Description: enable client interface failover
340 Details : When a child reconnects from another NID, properly update export
341 nid hash position and ldlm reverse import.
343 Severity : enhancement
345 Description: Communicate OST degraded/readonly state via statfs to MDS
346 Details : Flags in the statfs returned from OSTs indicate whether the
347 OST is in a degraded RAID state, or if the filesystem has
348 turned read-only after a filesystem error is detected.
353 Description: don't panic if EPROTO was hit when reading symlink
354 Details : correctly handling request reference in error cases.
356 Severity : enhancement
358 Description: Add LustreNetLink, a kernel-userspace communcation path.
360 Severity : enhancement
362 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
367 Description: access to llog context before init.
368 Details : move handling CATALOGS file at osc layer and forbid access to llog
372 Frequency : rare, only when client is writing to OST on local node
374 Description: Avoid deadlock for local client writes
375 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
376 memory freeing context. This allows OST threads to set the
377 PF_MEMALLOC flag on task structures in order to allocate memory
378 from reserved pools and complete IO.
379 Use GFP_HIGHUSER for OST allocations for non-local client writes,
380 so that the OST threads generate memory pressure and allow
381 inactive pages to be reclaimed.
383 Severity : enhancement
385 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
390 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
391 Details : move ldlm namespace creation in setup phase to avoid grab
392 _lprocfs_lock with cli_sem held.
396 Description: Temporarily disable grant shrink.
397 Details : Disable the feature for debugging.
400 Frequency : only when formatting test filesystems
402 Description: Unable to run several mkfs.lustre on loop devices at the same time
403 Details : mkfs.lustre returns error 256 on the concurrent loop devices
404 formatting. The solution is to proper handle the error.
406 Severity : enhancement
408 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
410 Severity : enhancement
412 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
417 Description: resolve race between obd_disconnect and class_disconnect_exports
418 Details : if obd_disconnect will be called to already disconnected export he
419 forget release one reference and osc module can't unloaded.
421 Severity : enhancement
423 Description: Allow tuning service thread via /proc
424 Details : For each service a new
425 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
426 created that can be used to set min/max thread counts, and get the
427 current number of running threads.
432 Descriptoin: don't return error if have partially created objects for file.
433 Details : lov_update_create_set uses set->set_success as index for created
434 objects, so if some requests will be failed, they will have hole
435 at end of array and qos_shrink_lsm for allocate correct lsm.
437 Severity : enhancement
439 Description: Update OFED support to 1.4.1
444 Description: Reduce small size read RPC
445 Details : Set read-ahead limit for every file and only do read-ahead when
446 available read-ahead pages are bigger than 1M to avoid small size
449 Severity : enhancement
451 Description: move AT tunable parameters for more consistent usage
452 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
454 Severity : enhancement
456 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
457 Details : allow skip disconnected ost for send statfs request and hide error
463 Description: fix racy locking of mballoc block bitmaps causing BUG
464 Details : The locking of the mballoc buddy bitmap and the in-memory
465 block bitmap was using two different spin locks in some
466 cases. This made it possible to incorrectly access the
467 mballoc bitmap while another process was modifying it,
468 causing a sanity assertion to fail. While no on-disk corruption
469 was reported, there was some risk of this happening.
472 Frequency : rare, on llog test 6
474 Descriptoin: don't allow connect to already connected import
475 Details : allowing connect to already connected import is hide connecting problem.
480 Descriptoin: allow kill process which wait statahead result
481 Details : for some reasons 'ls' can stick in waiting result from statahead,
482 in this case need way for kill this process.
484 Severity : enhancement
486 Description: Add state history info file, enhance import info file
487 Details : Track import connection state changes in a new osc/mdc proc file;
488 add overview-type data to the osc/mdc import proc file.
490 Severity : enhancement
492 Description: MDS create should not wait for statfs RPC while holding DLM lock.
494 Severity : enhancement
496 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
501 Descriptoin: OST grant too much space to client even there are not enough space.
502 Details : Client will shrink its grant cache to OST if there are no write
503 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
504 this grant cache if there are already not enough avaible space
505 (left_space < total_clients * 32M).
508 Frequency : start MDS on uncleanly shutdowned MDS device
510 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
511 Details : stay in waiting mds<>ost recovery finished produce random bugs
512 due race between two ll_sync thread for one lov target. send
513 ACTIVATE event only if connect realy finished and import have
517 Frequency : rare, connect and disconnect target at same time
519 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
520 Details : don't call obd_disconnect under lov_lock. this long time
521 operation and can block ptlrpcd which answer to connect request.
526 Descriptoin: don't lose wakeup for imp_recovery_waitq
527 Details : recover_import_no_retry or invalidate_import and import_close can
528 both sleep on imp_recovery_waitq, but we was send only one wakeup
532 Frequency : always with long access acl
534 Descriptoin: mds can't pack reply with long acl.
535 Details : mds don't control size of acl but they limited by reint/getattr
538 Severity : enhancement
540 Description: Update to SLES10 kernel-2.6.16.60-0.33.
542 Severity : enhancement
544 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
547 Frequency : start MDS on uncleanly shutdowned MDS device
549 Descriptoin: aborting recovery hang on MDS
550 Details : don't throttle destroy RPCs for the MDT.
553 Frequency : on remount
555 Description: external journal device not working after the remount
556 Details : clear dev_rdonly flag for external journal devices in
562 Description: shutdown vs evict race
563 Details : client_disconnect_export vs connect request race.
564 if client will evicted at this time - we start invalidate
565 thread without referece to import and import can be freed
569 Frequency : rare, need acl's on inode.
571 Description: client can't handle ost additional correctly
572 Details : if ost was added after client connected to mds client can have
573 hit lnet_try_match_md ... to big messages to wide striped files.
574 in this case need teach client to handle config events about add
575 lov target and update client max ea size at that event.
577 Severity : enhancement
579 Description: Changelogs
580 Details : Changelogs are a lightweight mechanism to track filesystem
581 metadata and namespace changes. The changelog is recorded
582 permanently on the MDTs, and is periodically "consumed" / purged
583 when records are no longer needed.
585 Severity : enhancement
587 Description: compact fld format with extents
588 Details : Store range of seq rather than every seq in FLD. Seq
589 controller update FLD rather than clients. In Case of CMD, mdt0
590 has FLD, all other metadata server act as non persistent proxy
591 for FLD queries and cache fld entries in fld cache.
596 Description: don't skip ost target if they assigned to file
597 Details : Drop slow OSCs if we can, but not for requested start idx.
598 This means "if OSC is slow and it is not the requested
599 start OST, then it can be skipped, otherwise skip it only
600 if it is inactive/recovering/out-of-space.
604 Description: more cleanup in mds_lov
605 Details : not send LOV EA under replay, we can't know about they size at this
606 time. Don't allow client connect to mds before any ost connected,
607 for avoid problems with LOV EA size and returning EIO to client.
609 Severity : enhancement
611 Description: Interoperability at server side (Disk interoperability)
613 Severity : enhancement
615 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
617 Severity : enhancement
619 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
621 Severity : enhancement
623 Description: New client IO stack (CLIO).
625 Severity : enhancement
627 Description: Commit on sharing. Eliminate inter-client dependencies between
628 uncommitted transactions by doing transaction commits.
629 Thereby clients may recovery independently.
632 Frequency : Create a symlink file with a very long name
634 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
635 Details : If there is no extra space in the request for early cancels,
636 ldlm_req_handles_avail() returns 0 instead of a negative value.
638 Severity : enhancement
640 Description: Add /proc entry for import status
641 Details : The mdc, osc, and mgc import directories now have
642 an import directory that contains useful import data for debugging
645 Severity : enhancement
647 Description: Re-disable certain /proc logging
648 Details : Enable and disable client's offset_stats, extents_stats and
649 extents_stats_per_process stats logging on the fly.
652 Frequency : Only on FC kernels 2.6.22+
654 Description: oops in statahead
655 Details : Do not drop reference count for the dentry from VFS when lookup,
656 VFS will do that by itself.
658 Severity : enhancement
660 Description: Generic /proc file permissions
661 Details : Set /Proc file permissions in a more generic way to enable non-
662 root users operate on some /proc files.
666 Description: Hitting mdc_commit_close() ASSERTION
667 Details : Properly handle request reference release in
668 ll_release_openhandle().
672 Description: quota recovery deadlock during mds failover
673 Details : This patch includes att18982, att18236, att18237 in bz14840.
675 1. fix osts hang when mds does failover with quotaon
676 2. prevent watchdog storm when osts threads wait for the
681 Frequency : only patchless client
682 Description: add workaround for race between add/remove dentry from hash
684 Severity : enhancement
686 Description: Allow OST glimpses to return PW locks
690 Description: LBUG when llog conf file is full
691 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
695 Description: Prevent import from entering FULL state when server in recovery
699 Description: service mount cannot take device name with ":"
700 Details : Only when device name contains ":/" will mount treat it as
706 Description: replace ptlrpcd with the statahead thread to interpret the async
707 statahead RPC callback
711 Frequency : on recovery
712 Description: I/O failures after umount during fail back
713 Details : if client reconnected to restarted server we need join to recovery
714 instead of find server handler is changed and process self
715 eviction with cancel all locks.
717 Severity : enhancement
719 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
721 Severity : enhancement
723 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
725 Severity : enhancement
727 Description: Upcall on Lustre log has been dumped
728 Details : Allow for a user mode script to be called once a Lustre log has
729 been dumped. It passes the filename of the dumped log to the
730 script, the location of the script can be specified via
731 /proc/sys/lnet/debug_log_upcall.
736 Description: avoid idr_remove called for id which is not allocated.
737 Details : Move assigment s_dev for clustered nfs to end of initialization,
738 to avoid problem with error handling.
743 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
744 Details : When connection is reused this not moved from CONN_UNUSED_HASH
745 into CONN_USED_HASH and this prodice warning when put connection
746 again in unused hash.
749 Severity : enhancement
751 Description: File striping can now be set to use an arbitrary pool of OSTs.
753 Severity : enhancement
755 Description: Export bytes_read/bytes_write count on OSC/OST.
759 Description: Early reply size mismatch, MGC loses connection
760 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
761 the connect flags are properly negotiated.
766 Description: quotas are not honored with O_DIRECT
767 Details : all writes with the flag O_DIRECT will use grants which leads to
768 this problem. Now using OBD_BRW_SYNC to guard this.
772 Description: add quota statistics
773 Details : 1. sort out quota proc entries and proc code.
774 2. add quota statistics
776 Severity : enhancement
778 Description: enable quota support for HEAD.
782 Description: Properly propagate oinfo flags from lov to osc for statfs
783 Details : restore missing copy oi_flags to lov requests.
785 Severity : enhancement
787 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
788 : l_getgroups(8), lst(8), routerstat(8)
789 Details : included man pages for llobdstat(8), llstat(8),
790 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
792 Severity : enhancement
794 Description: configure's --enable-quota should check the
795 : kernel .config for CONFIG_QUOTA
796 Details : configure is terminated if --enable-quota is passed but
797 : no quota support is in kernel
801 Description: Remove portals compatibility
802 Details : Remove portals compatibility, not interoperable with releases
807 Description: Resolve device initialization race
808 Details : Prevent proc handler from accessing devices added to the
809 obd_devs array but yet be intialized.
811 Severity : enhancement
813 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
815 Severity : enhancement
817 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
822 Description: Using wrong pointer in osc_brw_prep_request
823 Details : Access to array[-1] can produce panic if kernel compiled with
824 CONFIG_PAGE_ALLOC enabled
828 Description: Client runs out of low memory
829 Details : Consider only lowmem when counting initial number of llap pages
833 Description: *optional* service tags registration
834 Details : if the "service tags" package is installed on a Lustre node
835 When the filesystem is mounted, a local-node service tag will
836 be created. See http://inventory.sun.com/ for more information
837 about the Service Tags asset management system.
841 Description: Kernel BUG tries to release flock
842 Details : Lustre does not destroy flock lock before last reference goes
843 away. So always drop flock locks when client is evicted and
844 perform unlock regardless of successfulness of speaking to MDS.
848 Description: add refcount protection for osc callbacks, avoid panic on shutdown
852 Description: sanity test 65a fails if stripecount of -1 is set
853 Details : handle -1 striping on filesystem in ll_dirstripe_verify
858 Description: ASSERTION(CheckWriteback(page,cmd)) failed
859 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
862 Severity : enhancement
864 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
868 Description: do not process already freed flock
869 Details : flock can possibly be freed by another thread before it reaches
870 to ldlm_flock_completion_ast.
874 Description: LBUG during stress test
875 Details : Need properly lock accesses the flock deadlock detection list.
879 Description: oops in page fault handler
880 Details : kernel page fault handler can return two special 'pages' in error
881 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
885 Description: timeout with invalidate import.
886 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
887 should be handled by ptlrpcd. This produce long age waiting and
888 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
890 Severity : enhancement
892 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
896 Description: enable MGS and MDT services start separately
897 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
898 a co-located MGS without starting the MGS, which is a complement
899 to 'nosvc' mount option.
903 Frequency : after recovery
904 Description: precreate to many object's after del orphan.
905 Details : del orphan st in oscc last_id == next_id and this triger growing
906 count of precreated objects. Set flag LOW to skip increase count
907 of precreated objects.
911 Frequency : rare, on clear nid stats
912 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
913 Details : when clean nid stats sometimes try destroy live entry,
914 and this produce panic in free.
918 Description: Stack overflow during MDS log replay
919 ease stack pressure by using a thread dealing llog_process.
923 Description: wait until IO finished before start new when do lock cancel.
924 Details : VM protocol want old IO finished before start new, in this case
925 need wait until PG_writeback is cleared until check dirty flag and
926 call writepages in lock cancel callback.
928 Severity : enhancement
930 Description: using special macro for print time and cleanup in includes.
934 Description: mds_mfd_close() ASSERTION(rc == 0)
935 Details : In mds_mfd_close(), we need protect inode's writecount change
936 within its orphan write semaphore to prevent possible races.
940 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
944 Frequency : rare, on shutdown ost
945 Description: don't hit live lock with umount ost.
946 Details : shrink_dcache_parent can be in long loop with destroy dentries,
947 use shrink_dcache_sb instead.
951 Description: don't panic with use echo client
952 Details : echo client pass NULL as client nid pointer and this produce null
957 Description: fix build on ppc32
958 Details : compile code with -m64 flag produce wrong object file for ppc32.
962 Description: add message levels for liblustreapi
966 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
967 Details : Move the 'good_osts' check before the 'total_bavail' check. This
968 will result in an -EAGAIN and in the exit call path we call
969 alloc_rr() which will with increasing aggressiveness attempt to
970 aquire precreated objects on the minimum number of required OSCs.
974 Description: Use old size assignment to avoid deadlock
975 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
976 the scheduling while holding a spinlock. We do not need locking
977 for size in ll_update_inode() because size is only updated from
978 the MDS for directories or files without objects, so there is no
979 other place to do the update, and concurrent access to such inodes
980 are protected by the inode lock.
984 Description: resolve "_IOWR redefined" build error on SLES10
988 Description: dump the memory debugging after all modules are unloaded to
989 suppress false negative in conf_sanity test 39
991 Severity : enhancement
993 Description: build kernel-ib packages for OFED 1.3 in our release cycle
998 Description: fix SLES kernel versioning
999 Details : the kernel version for our SLES 10 kernel did not include a "-"
1000 before the "smp" at the end. while this was not a problem in
1001 general, it did mean that software trying to use the kernel
1002 version to try to detect a vendor specific kernel would fail.
1003 this was most evident by the OFED build scripts.
1007 Description: Don't update lov_desc members until making sure they are valid
1008 Details : When updating lov_desc members via proc fs, need fix their
1009 validities before doing the real update.
1013 Description: don't put request into delay list while invalidate in flight.
1014 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1015 import in flight. this produce timeout for invalidate and sometimes
1016 can cause stale data.
1020 Frequency : on ppc only
1021 Description: not convert ost objects for directory because it's not exist.
1022 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1026 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1028 Severity : enhancement
1030 Description: Update to support 2.6.22.14 vanilla kernel.
1034 Frequency : rare, on recovery
1035 Description: read procfs can produce deadlock in some situation
1036 Details : Holding lprocfs lock which send rpc can produce block for destroy
1037 obd objects and this also block reconnect with -EALREADY. This
1038 isn't fix all lprocfs bugs - but make it rare.
1040 Severity : enhancement
1042 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1045 Frequency : frequent on X2 node
1047 Description: mdc_set_open_replay_data LBUG
1048 Details : Set replay data for requests that are eligible for replay.
1052 Description: lustre_mgs: operation 101 on unconnected MGS
1053 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1054 MGC, and late on MGC cannot successfully connect to MGS and a lot
1055 of the error messages complaining that MGS is not connected.
1059 Frequency : on network error
1060 Description: panic with double free request if network error
1061 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1062 but it's true only for synchronus enqueue, for async enqueue
1063 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1066 Severity : enhancement
1068 Description: client-side metadata stat-ahead during readdir(directory readahead)
1069 Details : perform client-side metadata stat-ahead when the client detects
1070 readdir and sequential stat of dir entries therein
1073 Frequency : on start mds
1075 Description: Implement get_info(last_id) in obdfilter.
1078 Frequency : occasional
1080 Description: Correctly check stale fid, not start epoch if ost not support SOM
1081 Details : open with flag O_CREATE need set old fid in op_fid3 because
1082 op_fid2 was overwritten with new generated fid, but mds can answer
1083 with one of these two fids and both is not stale. Setattr
1084 incorrectly started an epoch and assume will be called
1085 done_writing, but without SOM done_writing ever being called.
1088 Frequency : rare, depends on device drivers and load
1090 Description: MDS or OSS nodes crash due to stack overflow
1091 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1092 In some cases, in conjunction with device drivers that use a lot
1093 of stack the MDS (or possibly OSS) service threads could overflow
1094 the stack. One change which was identified to consume additional
1095 stack has been reworked to avoid the extra stack usage.
1098 Frequency : occasional
1100 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1101 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1102 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1103 oscc_has_objects and see OSC has no objests and call
1104 oscc_internal_create to resend create request.
1106 Severity : enhancement
1108 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1110 Severity : enhancement
1112 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1115 Frequency : very rare
1117 Description: Fix replay if there is an un-replied request and open
1118 Details : In some cases, older replay request will revert the
1119 mcd->mcd_last_xid on MDS which is used to record the client's
1120 latest sent request.
1122 Severity : enhancement
1124 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1126 Severity : enhancement
1128 Description: Add rhel5 support to HEAD.
1130 Serverity : enhancement
1132 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1137 Description: Don't allow skipping OSTs if index has been specified.
1138 Details : Don't allow skipping OSTs if index has been specified, make
1139 locking in internal create lots better.
1143 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1144 Details : access to bitfield in structure is always rounded to long
1145 and this produce problem with not atomic change any bit.
1149 Description: Lustre make rpms failed.
1150 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1151 builds Lustre rpms from tarball.
1154 Frequency : rare on shutdown ost
1156 Description: If llog cancel was not send before clean_exports phase, this can
1157 produce deadlock in llog code.
1158 Details : If llog thread has last reference to obd and call class_import_put
1159 this produce deadlock because llog_cleanup_commit_master wait when
1160 last llog_commit_thread exited, but this never success because was
1161 called from llog_commit_thread.
1165 Description: allow userland application know is lost one of stripes.
1166 Details : fill lvb_blocks with error code on ost and return it to
1167 application if error flag found.
1171 Description: NULL lov_tgts causing MDS oops
1172 Details : more safe checks for NULL lov_tgts for avoid oops.
1174 Severity : enhancement
1176 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1180 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1181 Details : If an OST is down the MDS will hang indefinitely in
1182 obd_statfs_rqset() waiting for the statfs data. While for
1183 MDS QOS usage of statfs, it should not stuck in waiting.
1185 Severity : enhancement
1187 Description: remote_acl support
1188 Details : Support ACL-based permission check for remote user.
1189 Support setfacl/getfacl for remote user with the utils
1190 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1191 the system "{s,g}etfacl" utils.
1193 Severity : enhancement
1195 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1197 Severity : enhancement
1199 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1203 Description: make mgs_setparam() handle fsname containing dash
1204 Details : fsname containing a dash does not work with lctl conf_param
1206 Severity : enhancement
1208 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1210 Severity : enhancement
1212 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1216 Frequency : rare, at shutdown
1217 Description: access already free / zero obd_namespace.
1218 Details : if client_disconnect_export was called without force flag set,
1219 and exist connect request in flight, this can produce access to
1220 NULL pointer (or already free pointer) when connect_interpret
1221 store ocd flags in obd_namespace.
1225 Frequency : only at startup
1226 Description: not alloc memory with spinlock held.
1227 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1228 if any spinlock held.
1230 Severity : enhancement
1232 Description: make lustre randomly fail allocating memory
1233 Details : Make lustre randomly failed allocating memory for testing purpose.
1235 Severity : enhancement
1237 Description: lost problems with lov objid file
1238 Details : Fixes some scability and access to not inited memory problems
1239 in work with lov objdid file.
1244 Description: lfs find does not continue on file error
1245 Details : Continue other files processing when a file/dir is absent.
1249 Description: Inconsistent usage of lustre_pack_reply()
1250 Details : Standardize the usage of lustre_pack_reply() such that it
1251 always generate a CERROR on failure.
1254 Frequency : occasional
1256 Description: MDS hang or stay in waiting lock
1257 Details : If client receive lock with CBPENDING flag ldlm need send lock
1258 cancel as separate rpc, to avoid situation when cancel request
1259 can't processed due all i/o threads stay in wait lock.
1263 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1267 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1271 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1275 Description: Update to RHEL4 latest kernel.
1277 Severity : enhancement
1279 Description: Build SLES10 patchless client fails
1280 Details : The configure was broken by run ./configure with
1281 --with-linux-obj=.... argument for patchless client. When the
1282 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1283 can't search header adequately. Use absolute path such as
1284 -I($LINUX)/include instead.
1288 Description: interrupt oig_wait produce painc on resend.
1289 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1290 private set, and this produce situation when rq_ptlrpcd_data not
1291 copyed to new allocated request and triggered LBUG on assert
1292 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1293 ptlrpcd set if request is changed and can be safety changed to use
1298 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1299 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1300 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1301 these macros and replace them with their difinitions instead.
1305 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1306 Details : Update which_patch & target file for SLES10 latest kernel.
1309 Description: Remove client patches from SLES 10 kernel.
1310 Details : This causes SLES 10 clients to behave as patchless clients
1311 even on a Lustre-patched (server) kernel.
1313 Severity : enhancement
1315 Description: self-adjustable client's lru lists
1316 Details : use adaptive algorithm for managing client cached locks lru
1317 lists according to current server load, other client's work
1318 pattern, memory activities, etc. Both, server and client
1319 side namespaces provide number of proc tunables for controlling
1322 Severity : enhancement
1324 Description: light-weight GSS support
1325 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1326 and reduce performance overhead.
1328 Severity : enhancement
1330 Description: Linux keyring support
1331 Details : Support using service of Linux keyring for Lustre GSS internal
1332 context refresh/cache mechanism.
1336 Description: Fix errors in lfs documentation
1337 Details : Fixes man pages
1341 Description: don't use GFP_* in generic Lustre code.
1342 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1346 Description: obdclass is limited by single OBD_ALLOC(idarray)
1347 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1351 Description: warnings with build patchless client with vanila 2.6.19 and up
1352 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1353 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1357 Frequency : Only for SLES
1358 Description: sanity_quota fail test_1
1359 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1360 file, which makes TSTID[2] inunique.
1364 Frequency : Only for SLES9
1365 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1366 Details : Update target/ChangeLog/which_patch .
1370 Frequency : Only for test_10 in sanity-quota.sh
1371 Description: a bug in quota test code
1372 Details : checking a test flag in a wrong place
1376 Frequency : Only for test_13 in sanity-quota.sh
1377 Description: a bug in quota test script
1378 Details : shouldn't check overfull allocated quota
1382 Description: jbd statistics
1383 Details : Port older jbd statistics patch for sles10
1385 Severity : enhancement
1387 Description: Add printing inode info into message about error in writepage.
1392 Description: reply_lock_interpret crash due to race with it and lock cancel.
1393 Details : Do not replay locks that are being cancelled. Do not reference
1394 locks by their address during replay, just by their handle.
1398 Frequency : When flocks are used.
1399 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1400 Details : Flock locks might destroy just granted lock if it could be merged
1401 with another existing flock, this is done in completion handler,
1402 so teach ldlm_cli_enquque_fini that this is a valid case for
1408 Description: Oops in read and write path when failing to allocate lock.
1409 Details : Check if lock allocation failed and return error back.
1413 Description: lstripe command fails for valid OST index
1414 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1415 instead of lov->desc.ld_active_tgt_count.
1419 Description: sanity.sh failed test 103
1420 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1425 Description: df doesn't work properly if diskfs blocksize != 4K
1426 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1430 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1431 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1433 Severity : enhancement
1435 Description: Adaptive timeouts
1436 Details : RPC timeouts adapt to changing server load and network
1437 conditions to reduce resend attempts and improve recovery time.
1441 Description: llapi_file_create() does not allow some changes
1442 Details : add llapi_file_open() that allows specifying the mode and
1443 open flags, and also returns an open file handle.
1447 Description: merge and cleanup kernel patches.
1448 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1452 Description: Add journal checksum support.(Kernel part)
1453 Details : The journal checksum feature adds two new flags i.e
1454 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1455 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1456 indicates that the commit block contains the checksum for
1457 the blocks described by the descriptor blocks. Now commit
1458 record can be sent to disk without waiting for descriptor
1459 blocks to be written to disk. This behavior is controlled
1460 using JBD2_FEATURE_ASYNC_COMMIT flag.
1464 Description: OSS needs mutliple precreate threads
1465 Details : Add ability to start more than one create thread per OSS.
1469 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1470 Details : execute lfs setstripe on client
1474 Description: mds_obd_create error creating tmp object
1475 Details : When the user sets quota on root, llog will be affected and can't
1476 create files and write files.
1479 Frequency : Always on ia64 patchless client, and possibly others.
1481 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1482 Details : This allows the patchless client to be loaded on architectures
1483 without this export.
1487 Description: RedHat Update kernel for RHEL5
1488 Details : Add the kernel config file for RHEL5.
1492 Description: RedHat Update kernel for RHEL5
1493 Details : Modify the kernel config file more closer RHEL5.
1497 Description: Build failure against Centos5 (RHEL5)
1498 Details : Define PAGE_SIZE when it isn't present.
1502 Description: test_7 on sanity-quota.sh can't be used on separate machines
1503 Details : add facet to handle it
1507 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1508 Details : with NFS, the anon dentry's parent was set to itself in
1509 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1510 corresponding dentry other than use rec->ur_name.
1512 Severity : enhancement
1514 Description: lfs setstripe enhancement
1515 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1519 Description: enable data checksumming by default
1520 Details : enable checksum by default, allow --disable-checksum
1521 configure option and "-o nochecksum" mount option. Checksums
1522 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1526 Description: lustre support for RHEL5
1527 Details : Add support for RHEL5.
1531 Description: Client eviction due to failover config
1532 Details : after a connection loss, the lustre client should attempt to
1533 reconnect to the last active server first before trying the
1534 other potential connections.
1538 Description: when mds and osts use different quota unit(32bit and 64bit),
1539 quota will be released repeatly.
1540 Details : void sending multiple quota reqs to mds, which will keep the
1541 status between the reqs.
1545 Description: osts not allocated evenly to files
1546 Details : change the condition to increase offset_idx
1549 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1551 Description: Data corruption for OSTs that are formatted larger than 2TB
1553 Details : When generating the bio request for lustre file writes the
1554 sector number would overflow a temporary variable before being
1555 used for the IO. The data reads correctly from Lustre (which
1556 will overflow in a similar manner) but other file data or
1557 filesystem metadata may be corrupted in some cases.
1561 Description: Tune the kernel for good SCSI performance.
1562 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1563 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1568 Description: rewrite ext2-derived code in obdclass/uuid.c
1569 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1570 from scratch preserving functionality.
1574 Description: "lfs find -obd UUID" prints directories
1575 Details : "lfs find -obd UUID" will return all directory names instead
1576 of just file names. It is incorrect because the directories
1577 do not reside on the OSTs.
1581 Description: use i_size_read and i_size_write in 2.6 port
1582 Details : replace inode->i_size access with i_size_read/write()
1585 Frequency : when removing large files
1587 Description: scheduling issue during removal of large Lustre files
1588 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1589 It causes scheduling issues when removing large files (17TB in the
1593 Frequency : only with liblustre clients on XT3
1595 Description: evictions taking too long
1596 Details : allow llrd to evict clients directly on OSTs
1599 Frequency : only on ppc
1601 Description: /proc/fs/lustre/devices broken on ppc
1602 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1603 We should make sure the type of 'index' is loff_t and then cast
1604 explicitly as needed below. Do not assign an explicitly cast
1608 Frequency : only for rhel5
1610 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1611 Details : Modify the target file & which_kernel.
1613 Severity : enhancement
1615 Description: omit set fsid for export NFS
1616 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1617 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1621 Frequency : Always, for kernels after 2.6.16
1622 Description: Fix warning idr_remove called for id=.. which is not allocated.
1623 Details : Last kernels save old s_dev before kill super and not allow
1624 to restore from callback - restore it before call kill_anon_super.
1628 Description: buffer overruns could theoretically occur
1629 Details : llapi_semantic_traverse() modifies the "path" argument by
1630 appending values to the end of the origin string, and a
1631 overrun may occur. Adding buffer overrun check in liblustreapi.
1635 Description: Fix error on 'ls .' at the top of the Lustre mount.
1636 Details : Don't revalidate dentry if it is a root dentry.
1640 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1641 Details : Modify vm-tunables-rhel4.patch.
1645 Description: Kernel config for 2.6.18-vanilla.
1646 Details : Modify targets/2.6-vanilla.target.in.
1647 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1648 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1649 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1650 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1654 Description: improve handling recoverable errors
1655 Details : if request processig with error which can be recoverable on server
1656 request should be resend, otherwise page released from cache and
1661 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1662 Details : Modify target file & which_patch.
1663 A flaw was found in the IA32 system call emulation provided
1664 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1665 value could be stored in the %RAX register, which could trigger an
1666 out-of-bounds system call table access. An untrusted local user
1667 could exploit this flaw to run code in the kernel
1668 (ie a root privilege escalation). (CVE-2007-4573).
1672 Description: O_DIRECT bypasses client statistics.
1673 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1674 read_ahead_stats were updated. copied the stats section from
1675 osc_send_oap_rpc() into async_internal().
1679 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1683 Description: change order of libsysio includes
1684 Details : '#include sysio.h' should always come before '#include xtio.h'
1688 Description: handle "serious error: objid * already exists" more gracefully
1689 Details : If LAST_ID value on disk is smaller than the objects existing in
1690 the O/0/d* directories, it indicates disk corruption and causes an
1691 LBUG(). If the object is 0-length, then we should use the existing
1692 object. This will help to avoid a full fsck in most cases.
1694 Severity : enhancement
1696 Description: adapt the lustre_config script to support the upgrade case
1697 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1698 server targets to 1.6 in parallel.
1702 Description: To avoid grant space > avaible space when the disk is almost
1703 full. Without this patch you might see the error "grant XXXX >
1704 available" or some LBUG about grant, when the disk is almost
1706 Details : In filter_check_grant, for non_grant cache write, we should
1707 check the left space by if (*left > ungranted + bytes), instead
1708 of (*left > ungranted), because only we are sure the left space
1709 is enough for another "bytes", then the ungrant space should be
1710 increase. In client, we should update cl_avail_grant only there
1711 is OBD_MD_FLGRANT in the reply.
1715 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1716 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1717 described in CVE-2007-4573. This problem could allow untrusted
1718 local users to gain root access.
1721 Frequency : when using O_DIRECT and quotas
1723 Description: Incorrect file ownership on O_DIRECT output files
1724 Details : block usage reported by 'lfs quota' does not take into account
1725 files that have been written with O_DIRECT.
1730 Description: touch file failed when fs is not full
1731 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1732 otherwise we can get ENOSP while fs is not full.
1736 Description: parallel lock callbacks
1737 Details : Instead of sending blocking and completion callbacks as separated
1738 requests, adding them to a set and sending in parallel.
1741 Frequency : only for Cray XT3
1742 Bugzilla : 12829/13455
1743 Description: Changing primary group doesn't change the group lustre assigns to
1745 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1746 overridden with the primary group provided by the group upcall,
1747 whereas the supplied fsgid can be trusted if it is in the list of
1748 supplementary groups returned by the group upcall.
1750 Severity : enhancement
1752 Description: Allow masking D_WARNING, D_ERROR messages from console
1753 Details : Console messages can now be disabled via lnet.printk.
1757 Description: User code with malformed file open parameter crashes client node
1758 Details : Before packing join_file req, all the related reference should be
1759 checked carefully in case some malformed flags cause fake
1760 join_file req on client.
1764 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1765 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1766 mode is equal to the granted mode. However, it should also check
1767 that LDLM_FL_CP_REQD is not set.
1771 Description: Heavy nfs access might result in deadlocks
1772 Details : After ELC code landed, it is now improper to enqueue any mds
1773 locks under och_sem, because enqueue might want to decide to
1774 cancel open locks for same inode we are holding och_sem for.
1778 Description: Client eviction while running blogbench
1779 Details : A lot of unlink operations with concurrent I/O can lead to a
1780 deadlock causing evictions. To address the problem, the number of
1781 oustanding OST_DESTROY requests is now throttled to
1782 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1783 callbacks are processed in priority.
1787 Description: enable ACLs on MDS by default
1788 Details : ACLs must be enabled on MDS by default.
1791 Frequency : PPC/PPC64 only
1793 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1794 Details : fix duplicated definitions between asm-ppc64/types.h and
1795 lustre_types.h on PPC.
1798 Frequency : PPC/PPC64 only
1800 Description: asm-ppc/segment.h does not exist
1801 Details : fix compile issue on PPC.
1805 Description: better handle error messages in extents code
1808 Frequency : RHEL4 only
1810 Description: mkfs is very slow on IA64/RHEL4
1811 Details : A performance regression has been discovered in the MPT Fusion
1812 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1813 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1814 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1816 Severity : enhancement
1818 Description: SNMP support enhancement
1819 Details : Adding total number of sampled request for an MDS node in snmp
1822 Severity : enhancement
1824 Description: Optimize ldlm waiting list processing for PR extent locks
1825 Details : When processing waiting list for read extent lock and meeting read
1826 lock that is same or wider to it that is not contended, skip
1827 processing rest of the list and immediatelly return current
1828 status of conflictness, since we are guaranteed there are no
1829 conflicting locks in the rest of the list.
1833 Description: Time out and refuse to reconnect
1834 Details : When the failover node is the primary node, it is possible
1835 to have two identical connections in imp_conn_list. We must
1836 compare not conn's pointers but NIDs, otherwise we can defeat
1837 connection throttling.
1841 Description: port llog fixes from b1_6 into HEAD
1842 Details : Port llog reference couting and some llog cleanups from b1_6
1843 (bug 10800) into HEAD, for protect from panic and access to already
1844 free llog structures.
1848 Description: Detect stride IO mode in read-ahead
1849 Details : When a client does stride read, read-ahead should detect that and
1850 read-ahead pages according to the detected stride pattern.
1854 Description: data checksumming impacts single node performance
1855 Details : add support for several checksum algorithm. Currently, only CRC32
1856 and Adler-32 are supported. The checksum type can be changed on
1857 the fly via /proc/fs/lustre/osc/*/checksum_type.
1861 Description: use adler32 for page checksums
1862 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1867 Description: build for x2 fails
1868 Details : fix compile issue on Cray systems.
1872 Description: Properly match for duplicate locks
1873 Details : Due to different lock order from skiplists code, we need to
1874 traverse entire list for now
1877 Frequency : only on PPC/SLES10
1879 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1880 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1881 be defined. Add a hack in mkfs_lustre.c to work around this compile
1886 Description: LASSERT on MDS when client holding flock lock dies
1887 Details : ldlm pool logic depends on number of granted locks equal to
1888 number of released locks which is not true for flock locks, so
1889 just exclude such locks from consideration.
1893 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1894 Details : Use fsfilt_sync() for both the whole filesystem sync and
1895 individual file sync to eliminate dangerous inode locking
1896 with I_LOCK that can lead to a deadlock.
1900 Description: performance in 1.6.3
1901 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1902 sure that 1MB requests are not fragmented by the block layer.
1904 Severity : enhancement
1906 Description: organize the server-side client stats on per-nid basis
1907 Details : Change the structure of stats under obdfilter and mds to
1915 The "uuid"s file would list the uuids of _active_ exports.
1916 And the clear entry is to clear all stats and stale nids.
1918 Severity : enhancement
1920 Description: eliminate client locks in face of contention
1921 Details : file contention detection and lockless i/o implementation
1922 for contended files.
1926 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1929 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1931 Description: MDS may lose file striping (and hence file data) in some cases
1932 Details : If there are additional extended attributes stored on the MDS,
1933 in particular ACLs, SELinux, or user attributes (if user_xattr
1934 is specified for the client mount options) then there is a risk
1935 of attribute loss. Additionally, the Lustre file striping
1936 needs to be larger than default (e.g. striped over all OSTs),
1937 and an additional attribute must be stored initially in the
1938 inode and then increase in size enough to be moved to the
1939 external attribute block (e.g. ACL growing in size) for file
1944 Description: skiplist implementation simplification
1945 Details : skiplists are used to group compatible locks on granted list
1946 that was implemented as tracking first and last lock of each
1947 lock group the patch changes that to using doubly linked lists
1951 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1952 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1953 because LASSERT on both the data supplied by a client, and the
1954 data on disk is dangerous and incorrect.
1956 Severity : enhancement
1958 Description: Slow truncate/writes to huge files at high offsets.
1959 Details : Directly associate cached pages to lock that protect those pages,
1960 this allows us to quickly find what pages to write and remove
1961 once lock callback is received.
1965 Description: more ldlm soft lockups
1966 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1967 starve other threads from the resource lock for a long time in
1968 case of long waiting queue, so change the debug level from
1969 D_OTHER to the less frequently used D_INFO.
1971 Severity : enhancement
1973 Description: add -gid, -group, -uid, -user options to lfs find
1977 Description: Hung threads in invalidate_inode_pages2_range
1978 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1979 once one is completed. As a result, some RPCs are stuck in the
1980 queue and are never sent.
1984 Description: filter threads hungs on waiting journal commit
1985 Details : Cleanup filter group llog code, then only filter group llog will
1986 be only created in the MDS/OST syncing process.
1990 Description: Procfs and llog threads access destoryed import sometimes.
1991 Details : Sync the import destoryed process with procfs and llog threads by
1992 the import refcount and semaphore.
1994 Severity : enhancement
1996 Description: openlock cache of b1_6 port to HEAD
2001 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2002 Details : If insertion of an extent fails, then discard the inode
2003 preallocation and free data blocks else it can lead to duplicate
2008 Description: don't always update ctime in ext3_xattr_set_handle()
2009 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2010 In some cases the ctime should not be updated, for example for
2011 2.0->1.8 compatibility it is necessary to delete an xattr and it
2012 should not update the ctime.
2016 Bugzilla : 15713/16362
2017 Description: Assertion in iopen_connect_dentry in 1.6.3
2018 Details : looking up an inode via iopen with the wrong generation number can
2019 populate the dcache with a disconneced dentry while the inode
2020 number is in the process of being reallocated. This causes an
2021 assertion failure in iopen since the inode's dentry list contains
2022 both a connected and disconnected dentry.
2026 Description: assertion failure in ldlm_handle2lock()
2027 Details : fix a race between class_handle_unhash() and class_handle2object()
2028 introduced in lustre 1.6.5 by bug 13622.
2033 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2034 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2039 Description: oops in sd_iostats_seq_show()
2040 Details : unloading/reloading the scsi low level driver triggers a kernel
2041 bug when trying to access the sd iostat file.
2046 Description: Kernel panics during QLogic driver reload
2047 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2048 patch, causing memory corruption.
2053 Description: journal_dev option does not work in b1_6
2054 Details : pass mount option during pre-mount.
2056 Severity : enhancement
2058 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2059 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2060 extent information of a file. It can be used to map logical blocks
2061 in a file to physical blocks in the block device.
2065 Description: LDLM soft lockups - improvement
2066 Details : It is be possible to send the lock handle along with each read
2067 or write request because the client is already doing a lock match
2068 itself so there isn't any reason the OST should have to re-do that
2072 Frequency : only on Cray X2
2074 Description: X2 build failures
2075 Details : fix build failures on Cray X2.
2079 Description: xid & resent requests
2080 Details : Initialize RPC XID from clock at startup (randomly if clock is
2083 Severity : enhancement
2085 Description: Add lustre_start utility to start or stop multiple Lustre servers
2090 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2091 Details : In case of memory pressure, list_del() can be called twice on
2092 req->rq_history_list, causing a kernel oops.
2096 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2097 Details : fix stack overflow in the distributed lock manager by defering
2098 export eviction after a failed AST to the elt thread instead of
2099 handling it in the dlm interpret routine.
2103 Description: Convert some comments to new format.
2104 Details : Update documenting comments to match doxygen conventions.
2108 Description: Grammar fixes.
2109 Details : A couple of trivial spelling fixes.
2113 Description: OSD_COUNTERS-mandatory
2114 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2115 to simplify the code.
2119 Description: simplify cmm_device freeing logic.
2120 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2124 Description: Add lockdep support to dt_object_operations locking interface.
2125 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2126 parameter indicating lock ordering. Update mdd code to use new
2131 Description: Introduce failloc constants for lockless IO tests.
2132 Details : Add two new failloc constants to test lockless IO. Only one of
2133 them in implemented---another is checked in yet to be landed
2138 Description: Add lockdep support for inode mutex.
2139 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2140 in the situations where Lustre has to lock more than one inode
2145 Description: Add optional invariants checking support.
2146 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2147 configure switch. This macro is to be used for consistency and
2148 sanity checks that are too expensive to be left in `production'
2153 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2154 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2159 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2160 Details : Use spin_lock_nested() in (the only) situation where more than
2161 one ldlm_lock is locked simultaneously. Also, fix possible
2162 dead-lock in ldlm_lock_change_resource() by enforcing particular
2167 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2168 Details : Instead of specifying each ldlm_lock call-back through separate
2169 parameter, wrap them into struct ldlm_callback_suite.
2173 Description: Kill join_lru obd method and its callers.
2174 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2175 in memory---a requirement implemented through ->o_join_lru() obd
2176 method in HEAD. Remove this method and its users.
2180 Description: Add asynchronous ldlm ENQUEUE completion handler.
2181 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2182 case that stock ldlm_completion_ast() cannot handle as it waits
2183 until lock is granted. Introduce new ldlm_completion_ast_async()
2184 for this. Also comment ldlm_completion_ast().
2188 Description: ldlm_error <-> errno conversion.
2189 Details : Add functions to map (rather arbitrary) between LDLM error codes
2190 and standard errno values. CLIO needs this to prevent LDLM specific
2191 constants from escaping ldlm and osc.
2195 Description: Kill unused ldlm_handle2lock_ns() function.
2196 Details : Kill unused ldlm_handle2lock_ns() function.
2200 Description: Add lu_ref support to ldlm_lock
2201 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2202 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2203 and ldlm_resource. LDLM interface has to be changed, because code that
2204 releases a reference on a lock, has to "know" what reference this is.
2205 In the most frequent case
2207 lock = ldlm_handle2lock(handle);
2209 LDLM_LOCK_PUT(lock);
2211 no changes are required. When any other reference (received _not_ from
2212 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2213 instead of LDLM_LOCK_PUT().
2215 Arguably, changes are pervasive, and interface requires some discipline
2216 for proper use. On the other hand, it was very instrumental in finding
2217 a few leaked lock references.
2221 Description: Add ldlm_lock_addref_try().
2222 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2223 attempts to addref a lock that might be being canceled
2228 Description: Add ldlm_weigh_callback().
2229 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2230 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2231 hard-coded `number of pages' logic.
2235 Description: Add lockdep annotations to llog code.
2236 Details : Use appropriately tagged _nested() locking calls in the places
2237 where llog takes more than one ->lgh_lock lock.
2241 Description: Add loi_kms_set().
2242 Details : Wrap kms updates into a helper function.
2246 Description: Constify instances of struct lsm_operations.
2247 Details : Constify instances of struct lsm_operations.
2251 Description: lu_conf support.
2252 Details : On a server, a file system object is uniquely identified
2253 by a fid, which is sufficient to locate and load all object
2254 state (inode). On a client, on the other hand, more data are
2255 necessary instantiate an object. Change lu_object_find() and
2256 friends to take additional `lu_conf' argument describing object.
2257 Typically this includes layout information.
2261 Description: lu_context fixes.
2262 Details : Introduce new lu_context functions that are needed on the client
2263 side, where some system threads (ptlrpcd) are shared by multiple
2264 modules, and so cannot be stopped during module shutdown.
2268 Description: Add start and stop methods to lu_device_type_operations.
2269 Details : Introduce two new methods in lu_device_type_operations, that are
2270 invoked when first instance of a given type is created and last one
2271 is destroyed respectively. This is need by CLIO.
2275 Description: Add lu_ref support to struct lu_device.
2276 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2277 track leaked references.
2281 Description: Introduce lu_kmem_descr.
2282 Details : lu_kmem_descr and its companion interface allow to create
2283 and destroy a number of kmem caches at once.
2287 Description: Fix lu_object finalization race.
2288 Details : Fix a race between lu_object_find() finding an object and its
2289 concurrent finalization. This race is (most likely) not possible
2290 on the server, but might happen on the client.
2294 Description: Introduce lu_ref interface.
2295 Details : lu_ref is a debugging module allowing to track references to
2296 a given object. It is quite cpu expensive, and has to be
2297 explicitly enabled with --enable-lu_ref. See usage description
2302 Description: Factor lu_site procfs stats into a separate function.
2303 Details : Separate lu_site stats printing code into a separate function
2304 that can be reused on a client.
2308 Description: Constify instances of struct {lu,dt,md}_device_operations.
2309 Details : Constify instances of struct {lu,dt,md}_device_operations.
2313 Description: Introduce struct md_site and move meta-data specific parts of
2314 struct lu_site here.
2315 Details : Move md-specific fields out of struct lu_site into special struct
2316 md_site, so that lu_site can be used on a client.
2320 Description: Kill mdd_lov_destroy().
2321 Details : Remove unused mdd code.
2325 Description: Add st_block checking to multistat.c.
2326 Details : Add st_block checking to multistat.c.
2330 Description: Add lu_ref support to struct obd_device.
2331 Details : Add lu_ref tracking to obd_device.
2335 Description: Kill obd_set_fail_loc().
2336 Details : Remove unused code.
2340 Description: Add special type for ptlrpc_request interpret functions.
2341 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2342 there. Actual usage will be in CLIO.
2346 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2347 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2348 doesn't work with lockdep.
2352 Description: Add rwv.c test program.
2353 Details : New testing program exercising readv(2) and writev(2) (Qian).
2357 Description: Add sendfile.c test program.
2358 Details : New testing program exercising sendfile(2) (Jay).
2362 Description: Ratelimit a message that can be very frequent.
2363 Details : Ratelimit a memory allocation failure message that can
2368 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2369 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2373 Description: Liblustre build fixes.
2374 Details : Liblustre build fixes.
2378 Description: libcfs: add cfs_{need,cond}_resched() interface.
2379 Details : libcfs: add cfs_{need,cond}_resched() definition and
2380 implementations for Linux, NT, and liblustre.
2382 Severity : enhancement
2384 Description: More exported tunables for mballoc
2385 Details : Add support for tunable preallocation window and new tunables for
2386 large/small requests
2390 Description: Detect corruption of block bitmap and checking for preallocations
2391 Details : Checks validity of on-disk block bitmap. Also it does better
2392 checking of number of applied preallocations. When corruption is
2393 found, it turns filesystem readonly to prevent further corruptions.
2397 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2398 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2399 window to 0. If it does hit cache miss, and read-pattern is still
2400 stride-io mode, does not reset the stride window, but also does not
2401 increase the stride window length in this case.
2405 Frequency : only for big-endian servers
2406 Description: Check if system is big-endian while mounting fs with extents feature
2407 Details : Mounting a filesystem with extents feature will fail on big-endian
2408 systems since ext3-based ldiskfs is not supported on big-endian
2409 systems. This can be over-riden with "bigendian_extents" mount option.
2411 Severity : enhancement
2413 Description: The root squash functionality
2414 Details : A security feature, which is to prevent users from being able
2415 to mount lustre on their desktop, run as root, and delete
2416 all of the files in the filesystem. The goal is accomplished by
2417 remapping user id (UID) and group id (GID) of the root user to
2418 a UID and GID specified by the system administartor via Lustre
2419 configuration management server (MGS). The functionality also
2420 allows to specify sets of clients for which the remapping does
2425 Description: Excessive recovery window
2426 Details : With AT enabled, the recovery window can be excessively long (6000+
2427 seconds). To address this problem, we no longer use
2428 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2429 timeout no longer depends on the service time, it is set to
2430 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2431 time via pb_service_time.
2435 Description: Watchdog triggered on MDS failover
2436 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2437 the OSTs know that the MDS "UUID" can be reused for the same export
2438 from a different NID, so we do not need to wait for the export to be
2442 Frequency : rare, only if using MMP with Linux RAID
2444 Description: MMP doesn't work with Linux RAID
2445 Details : While using HA for Lustre servers with Linux RAID, it is possible
2446 that MMP will not detect multiple mounts. To make this work we
2447 need to unplug the device queue in RAID when the MMP block is being
2448 written. Also while reading the MMP block, we should read it from
2449 disk and not the cached one.
2451 Severity : enhancement
2453 Description: open file using fid
2454 Details : A file can be opened using just its fid, like
2455 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2458 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2461 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2462 Details : add proper locking for creation of jbd_4k slab cache
2466 Description: MMP check in ext3_remount() fails without displaying any error
2467 Details : When multiple mount protection fails during remount, proper error
2470 Severity : enhancement
2472 Description: Allow stripe size to be up to 4G-64k
2473 Details : Fix math logic to allow large stripe sizes.
2477 Description: add check for >8TB ldiskfs filesystems
2478 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2479 Don't allow >8TB ldiskfs filesystems to be mounted without
2480 force_over_8tb mount option
2485 Descriptoin: OSS complains "too many blocks" and crashes
2486 Details : In filter_commitrw_write():
2487 1. lquota_chkquota();
2488 2. obdo_from_inode();
2489 3. lquota_pending_commit();
2490 "2." has changed uid and gid in this case, then caused
2491 lquota_chkquota() and lquota_pending_commit() to use
2492 different uid and gid. That is the root of the bug.
2496 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2497 Details : Detect and print out the 'stripe count = -1' case properly.
2501 Description: Out or order replies might be lost on replay
2502 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2503 request with tid smaller then one currently being added, add it
2504 to the start, not end of the list.
2506 --------------------------------------------------------------------------------
2508 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2510 * Support for kernels:
2511 2.6.5-7.283 (SLES 9),
2512 2.6.9-55.EL (RHEL 4),
2513 2.6.16.46-0.14 (SLES 10),
2514 2.6.18.8 vanilla (kernel.org)
2515 * Client support for unpatched kernels:
2516 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2517 2.6.16 - 2.6.22 vanilla (kernel.org)
2518 * Due to recently discovered recovery problems, we do not recommend
2519 using patchless RHEL 4 clients with this or any earlier release.
2520 * Recommended e2fsprogs version: 1.39.cfs8
2521 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2522 * Starting with this release, the ldiskfs backing filesystem required
2523 by Lustre is now in its own package, lustre-ldiskfs. This package
2524 should be installed. It is versioned separately from Lustre and
2525 may be released separately in future.
2529 Description: block reactivating mgc import until all deactivates complete
2530 Details : Fix race when failing back MDT/MGS to itself (testing)
2533 Frequency : at statup only
2535 Description: mds_lov_synchronize race leads to various problems
2536 Details : simultaneous MDT->OST connections at startup can cause the
2537 sync to abort, leaving the OSC in a bad state.
2539 Severity : enhancement
2541 Description: add optional extra BUILD_VERSION info
2542 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2543 to override the lustre version.
2546 Frequency : 2.6.18 servers only
2548 Description: ll_kern_mount() doesn't release the module reference
2549 Details : The ldiskfs module reference count never drops down to 0
2550 because ll_kern_mount() doesn't release the module reference.
2555 Description: server LBUG when using old ost_num_threads parameter
2556 Details : Accept the old ost_num_threads parameter but warn that it
2557 is deprecated, and fix an off-by-one error that caused an LBUG.
2562 Description: Transient SCSI error results in persistent IO issue
2563 Details : iobuf->dr_error is not reinitialized to 0 between two
2567 Frequency : sometimes when underlying device returns I/O errors
2569 Description: OSTs not going read-only during write failures
2570 Details : OSTs are not remounted read-only when the journal commit threads
2571 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2572 instead of the ext3 wrappers.
2576 Description: poor connect scaling with increasing client count
2577 Details : Don't run filter_grant_sanity_check for more than 100 exports
2578 to improve scaling for large numbers of clients.
2581 Frequency : SLES10 only
2583 Description: sanity-quota.sh quotacheck failed: rc = -22
2584 Details : Quotas cannot be enabled on SLES10.
2587 Frequency : liblustre clients only
2589 Description: getdirentries does not give error when run on compute nodes
2590 Details : getdirentries does not fail when the size specified as an argument
2591 is too small to contain at least one entry
2593 Severity : enhancement
2595 Description: Add LNET router traceability for debug purposes
2596 Details : If a checksum failure occurs with a router as part of the
2597 IO path, the NID of the last router that forwarded the bulk data
2598 is printed so it can be identified.
2603 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2604 Details : Due to a race condition, liblustre clients were occasionally
2605 evicted incorrectly.
2607 Severity : enhancement
2609 Description: lfs setstripe use optional parameters instead of postional
2612 Severity : enhancement
2614 Description: Nanosecond timestamp support for ldiskfs
2615 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2616 resolution timestamps. There is not yet support for this at
2617 the Lustre filesystem level.
2620 Frequency : during server recovery
2622 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2623 Details : request with rq_no_resend flag not awake l_wait_event if they get
2627 Frequency : nfs export on patchless client
2629 Description: connectathon hang when test nfs export over patchless client
2630 Details : Disconnected dentry cannot be found with lookup, so we do not need
2631 to unhash it or make it invalid
2634 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2635 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2636 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2641 Description: open req refcounting wrong on reconnect
2642 Details : If reconnect happened between getting open reply from server and
2643 call to mdc_set_replay_data in ll_file_open, we will schedule
2644 replay for unreferenced request that we are about to free.
2645 Subsequent close will crash in variety of ways.
2646 Check that request is still eligible for replay in
2647 mdc_set_replay_data().
2652 Description: disable writes to filesystem when reading health_check file
2653 Details : the default for reading the health_check proc file has changed
2654 to NOT do a journal transaction and write to disk, because this
2655 can cause reads of the /proc file to hang and block HA state
2656 checking on a healthy but otherwise heavily loaded system. It
2657 is possible to return to the previous behaviour during configure
2658 with --enable-health-write.
2660 Severity : enhancement
2662 Description: 64-bit inode version
2663 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2664 to the inode. This will be required for version-based recovery.
2669 Description: MDS fails to start if a duplicate client export is detected
2670 Details : in some rare cases it was possible for a client to connect to
2671 an MDS multiple times. Upon recovery the MDS would detect this
2672 and fail during startup. Handle this more gracefully.
2674 Severity : enhancement
2676 Description: Add -o localflock option to simulate old noflock
2678 Details : This will achieve local-only flock/fcntl locks
2684 Description: log_commit_thread vs filter_destroy race leads to crash
2685 Details : Take import reference before releasing llog record semaphore
2690 Description: Wrong request locking in request set processing
2691 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2692 imp_delayed_list, in this place should be used imp_lock.
2695 Frequency : when reconnection
2697 Description: Grant Leak when osc reconnect to OST
2698 Details : When osc reconnect ost, OST(filter) should check whether it
2699 should grant more space to client by comparing fed_grant and
2700 cl_avail_grant, and return the granted space to client instead
2701 of "new granted" space, because client will call osc_init_grant
2702 to update the client grant space info.
2705 Frequency : when client reconnect to OST
2707 Description: Grant Leak when osc do resend and replay bulk write
2708 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2709 bulk write request, because the grant info while be sync between
2710 OSC and OST when reconnect, and we should ignore the grant info
2711 these of resend/replay write req.
2716 Description: Grant space more than avaiable left space sometimes.
2717 Details : When then OST is about to be full, if two bulk writing from
2718 different clients came to OST. Accord the avaliable space of the
2719 OST, the first req should be permitted, and the second one
2720 should be denied by ENOSPC. But if the seconde arrived before
2721 the first one is commited. The OST might wrongly permit second
2722 writing, which will cause grant space > avaiable space.
2725 Frequency : when client is evicted
2727 Description: Grant might be wrongly erased when osc is evicted by OST
2728 Details : when the import is evicted by server, it will fork another
2729 thread ptlrpc_invalidate_import_thread to invalidate the
2730 import, where the grant will be set to 0. While the original
2731 thread will update the grant it got when connecting. So if
2732 the former happened latter, the grant will be wrongly errased
2733 because of this race.
2738 Description: Checking Stale with correct fid
2739 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2740 to check whether it is stale, because sometimes, we want the
2741 enqueue happened anyway, and op_data.fid2 will not be initialized.
2743 Severity : enhancement
2745 Description: update patchless client
2746 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2749 Frequency : only with 2.4 kernel
2751 Description: random memory corruption
2752 Details : size of struct ll_inode_info is to big for union inode.u and this
2753 can be cause of random memory corruption.
2758 Description: Memory leak in recovery
2759 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2760 It should also check obd_fail before fsfilt_start, otherwise if
2761 fsfilt_start return -EROFS,(failover mds during mds recovery).
2762 then the req will return with repmsg->transno = 0 and rc = EROFS.
2763 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2764 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2765 be freed no matter whether fsfilt_commit success or not.
2768 Frequency : only with huge count clients
2770 Description: Prevents from taking the superblock lock in llap_from_page for
2772 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2773 for llap_from_page call in ll_removepage prevents from taking the
2774 superblock lock for a soon died page.
2779 Description: Not check open intent error before release open handle
2780 Details : in some rare cases, the open intent error is not checked before
2781 release open handle, which may cause
2782 ASSERTION(open_req->rq_transno != 0), because it tries to release
2783 the failed open handle.
2788 Description: Set cat log bitmap only after create log success.
2789 Details : in some rare cases, the cat log bitmap is set too early. and it
2790 should be set only after create log success.
2794 Description: Accessing a block bevice can re-enable I/O when Lustre is
2795 tearing down a device.
2796 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2801 Description: service threads may hog cpus when there are a lot of requests
2802 Details : Insert cond_resched to give other threads a chance to use some CPU
2807 Description: the cat log was not initialized in recovery
2808 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2809 unlink log on mds will not be initialized until mds post
2810 recovery. And also in mds post recovery, the unlink log will
2811 initialization will be done asynchronausly, so there will be race
2812 between add unlink log and unlink log initialization.
2816 Description: brw_stats were being printed incorrectly
2817 Details : brw_stats were being printed as log2 but all of them were not
2818 recorded as log2. Also remove some code duplication arising from
2819 filter_tally_{read,write}.
2823 Frequency : rare, only in recovery.
2824 Description: ASSERTION(req->rq_type != LI_POISON) failed
2825 Details : imp_lock should be held while iterating over imp_sending_list for
2826 prevent destroy request after get timeout in ptlrpc_queue_wait.
2830 Description: replay-single.sh test 52 fails
2831 Details : A lock's skiplist need to be cleanup when it being unlinked
2832 from its resource list.
2836 Description: Short directio read returns full requested size rather than
2838 Details : Direct I/O operations should return actual amount of bytes
2839 transferred rather than requested size.
2841 Severity : enhancement
2843 Description: metadata RPC reduction (e.g. for rm performance)
2844 Details : decrease the amount of synchronous RPC between clients and servers
2845 by canceling conflicing lock before the operation on the client
2846 and packing thier handles into the main operation RPC to server.
2848 Severity : enhancement
2850 Description: Async OSC create to avoid the blocking unnecessarily.
2851 Details : If a OST has no remain object, system will block on the creating
2852 when need to create a new object on this OST. Now, ways use
2853 pre-created objects when available, instead of blocking on an
2854 empty osc while others are not empty. If we must block, we block
2855 for the shortest possible period of time.
2859 Description: improve handling recoverable errors
2860 Details : if request processig with error which can be recoverable on server
2861 request should be resend, otherwise page released from cache and
2864 Severity : enhancement
2866 Description: refine locking for avoid write wrong info into lov_objid file
2867 Details : fix possible races with add new target and write/update data in
2870 --------------------------------------------------------------------------------
2872 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2877 Frequency : on some architectures
2879 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2880 Details : Uninitialized flags sometimes cause configuration commands to
2884 Frequency : patchless clients only
2886 Description: missing __iget() symbol export
2887 Details : The __iget() symbol export is missing. To avoid the need for
2888 this on patchless clients the deathrow inode reaper is turned
2889 off, and we depend on the VM to clean up old inodes. This
2890 dependency was during via the fix for bug 12181.
2892 --------------------------------------------------------------------------------
2894 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2896 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2897 INTEROPERATE with older versions automatically. In many cases a
2898 special upgrade step is needed. Please read the
2899 user documentation before upgrading any part of a live system.
2900 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2901 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2902 * WARNING: Lustre configuration and startup changes are required with
2903 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2907 Severity : enhancement
2909 Description: MountConf
2910 Details : Lustre configuration is now managed via mkfs and mount
2911 commands instead of lmc and lconf. New obd types (MGS, MGC)
2912 are added for dynamic configuration management. See
2913 https://mail.clusterfs.com/wikis/lustre/MountConf for
2916 Severity : enhancement
2918 Description: dynamic OST addition
2919 Details : OSTs can now be added to a live filesystem
2921 Severity : enhancement
2923 Description: startup order invariance
2924 Details : MDTs and OSTs can be started in any order. Clients only
2925 require the MDT to complete startup.
2927 Severity : enhancement
2929 Description: parallel, asynchronous orphan cleanup
2930 Details : orphan cleanup is now performed in separate threads for each
2931 OST, allowing parallel non-blocking operation.
2933 Severity : enhancement
2935 Description: optimized stripe assignment
2936 Details : stripe assignments are now made based on ost space available,
2937 ost previous usage, and OSS previous usage, in order to try
2938 to optimize storage space and networking resources.
2940 Severity : enhancement
2942 Description: Permanently set tunables
2943 Details : All writable /proc/fs/lustre tunables can now be permanently
2944 set on a per-server basis, at mkfs time or on a live system.
2946 Severity : enhancement
2948 Description: Lustre message v2
2949 Details : Add lustre message format v2.
2951 Severity : enhancement
2953 Description: client OST exclusion list
2954 Details : Clients can be started with a list of OSTs that should be
2955 declared "inactive" for known non-responsive OSTs.
2959 Description: SPEC SFS validation failure on NFS v2 over lustre.
2960 Details : Changes the blocksize for regular files to be 2x RPC size,
2961 and not depend on stripe size.
2963 Severity : enhancement
2965 Description: Multiple MD RPCs in flight.
2966 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2967 To avoid overly-overloading MDS, introduce a limit on number of
2968 MDS RPCs in flight for a single client and add /proc controls
2969 to adjust this limit.
2971 Severity : enhancement
2973 Description: client read/write statistics
2974 Details : Add client read/write call usage stats for performance
2975 analysis of user processes.
2976 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2977 file access. extents_stats shows chunk size distribution.
2978 extents_stats_per_process show chunk size distribution per
2981 Severity : enhancement
2983 Description: mds statistics
2984 Details : Add detailed mds operations statistics in
2985 /proc/fs/lustre/mds/*/stats.
2989 Description: Failure of copying files with lustre special EAs.
2990 Details : Client side always return success for setxattr call for lustre
2991 special xattr (currently only "trusted.lov").
2996 Description: Refcount LNET uuids
2997 Details : The global LNET uuid list grew linearly with every startup;
2998 refcount repeated list entries instead of always adding to
3001 Severity : enhancement
3003 Description: Dynamic service threads
3004 Details : Within a small range, start extra service threads
3005 automatically when the request queue builds up.
3008 Frequency : mixed-endian client/server environments
3010 Description: mixed-endian crashes
3011 Details : The new msg_v2 system had some failures in mixed-endian
3014 Severity : enhancement
3016 Description: Easy OST removal
3017 Details : OSTs can be permanently deactivated with e.g. 'lctl
3018 conf_param lustre-OST0001.osc.active=0'
3020 Severity : enhancement
3022 Description: MGS proc entries
3023 Details : Added basic proc entries for the MGS showing what filesystems
3026 Severity : enhancement
3028 Description: provide MGS failover
3029 Details : Added config lock reacquisition after MGS server failover.
3031 Severity : enhancement
3033 Description: add Linux 2.4 support
3034 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3038 Description: a large application tries to do I/O to the same resource and dies
3039 in the middle of it.
3040 Details : Check the req->rq_arrival time after the call to
3041 ost_brw_lock_get(), but before we do anything about
3042 processing it & sending the BULK transfer request. This
3043 should help move old stale pending locks off the queue as
3044 quickly as obd_timeout.
3047 Frequency : when an incorrect nid is specified during startup
3049 Description: ptlrpc connect to non-existant node causes kernel crash
3050 Details : LNET can't be re-entered from an event callback, which
3051 happened when we expire a message after the export has been
3052 cleaned up. Instead, hand the zombie cleanup off to another
3055 Severity : enhancement
3057 Description: plain/inodebits lock performance improvement
3058 Details : Grouping plain/inodebits in granted list by their request modes
3059 and bits policy, thus improving the performance of search through
3063 Frequency : only if OST filesystem is corrupted
3065 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3066 Details : for a short time RPCs with bulk IO are in the replay list,
3067 but replay of bulk IOs is unimplemented. If the OST filesystem
3068 is corrupted due to disk cache incoherency and then replay is
3069 started it is possible to trip an assertion. Avoid putting
3070 committed RPCs into the replay list at all to avoid this issue.
3073 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3075 Description: System hang on startup
3076 Details : This bug allowed the liblustre (e.g. catamount) client to
3077 return to the app before handling all startup RPCs. This
3078 could leave the node unresponsive to lustre network traffic
3079 and manifested as a server ptllnd timeout.
3081 Severity : enhancement
3083 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3084 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3085 boolean to control whether to print peer debug info when a
3086 client's RPC times out.
3089 Frequency : only for kernels with patches from Lustre below 1.4.3
3091 Description: Remove old rdonly API
3092 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3095 Frequency : only for devices with external journals
3097 Description: Set external device read-only also
3098 Details : During a commanded failover stop, we set the disk device
3099 read-only while the server shuts down. We now also set any
3100 external journal device read-only at the same time.
3103 Frequency : when upgrading from 1.4 while trying to change parameters
3105 Description: The wrong (new) MDC name was used when setting parameters for
3106 upgraded MDT's. Also allows changing of OSC (and MDC)
3107 parameters if --writeconf is specified at tunefs upgrade time.
3110 Frequency : when setting specific ost indicies
3112 Description: QOS code breaks on skipped indicies
3113 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3114 created with --index need not be sequential.
3116 Severity : enhancement
3118 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3119 Details : The uninit_groups feature works in conjunction with the kernel
3120 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3121 up the pass1 processing of e2fsck. This is a read-only feature
3122 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3123 mount filesystems that have had this feature enabled.
3125 Severity : enhancement
3127 Description: Improve multi-block allocation algorithm to avoid fragmentation
3128 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3129 allocation locality and avoid filesystem fragmentation.
3131 ------------------------------------------------------------------------------
3133 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3135 * Support for kernels:
3136 2.4.21-47.0.1.EL (RHEL 3)
3137 2.6.5-7.283 (SLES 9)
3138 2.6.9-42.0.10.EL (RHEL 4)
3139 2.6.12.6 vanilla (kernel.org)
3140 2.6.16.27-0.9 (SLES 10)
3141 * Recommended e2fsprogs version: 1.39.cfs5
3143 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3147 Frequency : occasional, depends on client load and configuration
3148 Bugzilla : 12181, 12203
3149 Description: data loss for recently-modified files
3151 Details : In some cases it is possible that recently written or created
3152 files may not be written to disk in a timely manner (this should
3153 normally be within 30s unless client IO load is very high).
3154 The problem appears as zero-length files or files that are a
3155 multiple of 1MB in size after a client crash or client eviction
3156 that are missing data at the end of the file.
3158 This problem is more likely to be hit on clients where files are
3159 repeatedly created and unlinked in the same directory, clients
3160 have a large amount of RAM, have many CPUs, the filesystem has
3161 many OSTs, the clients are rebooted frequently, and/or the files
3162 are not accessed by other nodes after being written.
3164 The presence of the problem can be detected by looking at
3165 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3166 smaller than the second (nr_unused) then dirty files will not
3167 be flushed automatically to disk. "sync; sleep 10" should be
3168 run several times on the node before unmounting it to update
3169 Lustre (this is also safe to run on nodes without this problem).
3171 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3172 that can cause this same problem, so customers using that kernel
3173 also need to update the kernel in addition to Lustre. In order
3174 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3176 It is normal that files written just before a client crash (less
3177 than 30s) may not yet have been flushed to disk, even for local
3181 Frequency : frequent on thin XT3 nodes
3183 Description: UUID collision on thin XT3 Linux nodes
3184 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3185 randomly, since we relied on an insufficiently-seeded PRNG.
3190 Description: OSS hangs after "All ost request buffers busy"
3191 Details : A deadlock between quota and journal operations caused OSS
3192 hangs after printing "All ost request buffers busy."
3195 Frequency : always on liblustre builds
3197 Description: Cleanup compiler warnings on liblustre
3200 Frequency : always on liblustre builds on XT3
3202 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3206 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3207 > 32000 subdirectories
3208 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3209 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3210 subdirectory count crosses 32000. This will aid e2fsck to
3211 correctly handle more than 32000 subdirectories.
3214 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3216 Description: System hang on startup
3217 Details : This bug allowed the liblustre (e.g. catamount) client to
3218 return to the app before handling all startup RPCs. This
3219 could leave the node unresponsive to lustre network traffic
3220 and manifested as a server ptllnd timeout.
3222 Severity : enhancement
3224 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3225 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3226 boolean to control whether to print peer debug info when a
3227 client's RPC times out.
3232 Description: make O_SYNC working on 2.6 kernels
3233 Details : 2.6 kernels use different method for mark pages for write,
3234 so need add a code to lustre for O_SYNC work.
3239 Description: Failure to close file and release space on NFS
3240 Details : Put inode details into lock acquired in ll_intent_file_open.
3241 Use mdc_intent_lock in ll_intent_open to properly
3242 detect all kind of errors unhandled by mdc_enqueue.
3247 Description: proc file read during shutdown sometimes raced obd removal,
3249 Details : Add lock to prevent obd access after proc file removal.
3252 Frequency : Only for files larger than 4GB on 32-bit clients.
3254 Description: improperly doing page alignment of locks
3255 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3256 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3261 Description: RPCs being resent when they shouldn't be
3262 Details : Some RPCs that should not be resent are being resent. This
3263 can cause inconsistencies in the RPC state machine. Do not
3264 resend such requests.
3267 Frequency : rare, only with NFS export
3269 Description: Crash on NFS re-export node
3270 Details : under very unusual load conditions an assertion is hit in
3271 ll_intent_file_open()
3274 Frequency : only if OST filesystem is corrupted
3276 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3277 Details : for a short time RPCs with bulk IO are in the replay list,
3278 but replay of bulk IOs is unimplemented. If the OST filesystem
3279 is corrupted due to disk cache incoherency and then replay is
3280 started it is possible to trip an assertion. Avoid putting
3281 committed RPCs into the replay list at all to avoid this issue.
3286 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3287 Details : Large single O_DIRECT read and write calls can fail to allocate
3288 a sufficiently large buffer to process the request. In case of
3289 allocation failure the allocation is retried with a smaller
3290 buffer and broken into smaller requests.
3292 Severity : enhancement
3294 Description: Add -o localflock option to simulate old noflock behaviour.
3295 Details : This will achieve local-only flock/fcntl locks coherentness.
3300 Description: versioning check is incomplete
3301 Details : Checking the version difference of client vs. server, report
3302 error if the gap is too big.
3307 Description: add support PG_writeback bit
3308 Details : add support for PG_writeback bit for Lustre, for more carefull
3309 work with page cache in 2.6 kernel. This also fix some deadlocks
3310 and remove hack for work O_SYNC with 2.6 kernel.
3312 Severity : enhancement
3314 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3315 Details : The uninit_groups feature works in conjunction with the kernel
3316 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3317 up the pass1 processing of e2fsck. This is a read-only feature
3318 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3319 mount filesystems that have had this feature enabled.
3321 Severity : enhancement
3323 Description: Improve multi-block allocation algorithm to avoid fragmentation
3324 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3325 allocation locality and avoid filesystem fragmentation.
3327 ------------------------------------------------------------------------------
3329 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3331 * Support for kernels:
3332 2.6.9-42.0.3.EL (RHEL 4)
3333 2.6.5-7.276 (SLES 9)
3334 2.4.21-47.0.1.EL (RHEL 3)
3335 2.6.12.6 vanilla (kernel.org)
3336 2.6.16.21-0.8 (SLES10)
3337 * Recommended e2fsprogs version: 1.39.cfs2-0
3339 * The backwards-compatible /proc/sys/portals symlink has been removed
3340 in this release. Before upgrading, please ensure that you change
3341 any configuration scripts or /etc/sysctl.conf files that access
3342 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3343 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3344 in advance of the upgrade on any system running Lustre 1.4.6 or
3345 newer, since /proc/sys/lnet was added in that version.
3346 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3350 Frequency : only when quota is used
3352 Description: avoid scanning export list for quota master
3353 Details : Change the algorithms to avoid scanning export list in order
3354 to improve the efficiency.
3357 Frequency : MDS failover only, very rarely
3359 Description: "went back in time" messages on mds failover
3360 Details : The greatest transno may be lost when the current operation
3361 finishes with an error (transno==0) and the client's last_rcvd
3362 record is over-written. Save the greatest transno in the
3363 mds_last_transno for this case.
3366 Frequency : always for specific kernels and striping counts
3368 Description: client may get "Matching packet too big" without ACL support
3369 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3370 when trying to access files in certain configurations. The
3371 clients should in fact be denied when mounting because they do
3372 not understand ACLs.
3375 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3377 Description: many clients connecting with IO in progress causes connect timeouts
3378 Details : Avoid synchronous journal commits to avoid delays caused by many
3379 clients connecting/disconnecting when bulk IO is in progress.
3380 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3381 OST_IO_PORTAL to avoid delays behind potentially many pending
3385 Frequency : occasionally with multiple writers to a single file
3387 Description: shared writes to file may result in wrong size reported by stat()
3388 Details : Allow growing of kms when extent lock is cancelled
3391 Frequency : always with random mmap IO to multi-striped file
3393 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3394 Details : Only if the hole is at the end of OST object so that kms is too
3395 small. Fix is to increase kms accordingly in ll_nopage.
3398 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3400 Description: writes to a missing object would leak memory on the OST
3401 Details : If there is an inconsistency between the MDS and OST filesystems,
3402 such that the MDS references an object that doesn't exist, writes
3403 to that object will leak memory due to incorrect cleanup in the
3404 error handling path, eventually running out of memory on the OST.
3409 Description: Creating too long symlink causes lustre errors
3410 Details : Check symlink and name lengths before sending requests to MDS.
3413 Frequency : only if flock is enabled (not on by default)
3415 Description: posix locks not released on fd closure on 2.6.9+
3416 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3417 kernels, this caused such locks not to be released on fd close and
3418 then assertions on fs unmount about still used locks.
3421 Frequency : MDS failover only, very rarely
3423 Description: clients may get ASSERTION(granted_lock != NULL)
3424 Details : When request was taking a long time, and a client was resending
3425 a getattr by name lock request. The were multiple lock requests
3426 with the same client lock handle and
3427 mds_getattr_name->fixup_handle_for_resent_request found one of the
3428 lock handles but later failed with ASSERTION(granted_lock != NULL).
3433 Description: handle->h_buffer_credits > 0, assertion failure
3434 Details : h_buffer_credits is zero after truncate, causing assertion
3435 failure. This patch extends the transaction or creates a new
3439 Frequency : NFS re-export or patchless client
3440 Bugzilla : 11179, 10796
3441 Description: Crash on NFS re-export node (__d_move)
3442 Details : We do not want to hash the dentry if we don't have a lock.
3443 But if this dentry is later used in d_move, we'd hit uninitialised
3444 list head d_hash, so we just do this to init d_hash field but
3445 leave dentry unhashed.
3448 Frequency : NFS re-export or patchless client
3450 Description: NFS exports has problem with symbolic link
3451 Details : lustre client didn't properly install dentry when re-exported
3452 to NFS or running patchless client.
3455 Frequency : NFS re-export or patchless client
3457 Description: Various nfs/patchless fixes.
3458 Details : fixes reuse disconected alias for lookup process - this fixes
3459 warning "find_exported_dentry: npd != pd",
3460 fix permission error with open files at nfs.
3461 fix apply umask when do revalidate.
3464 Frequency : occasional
3466 Description: Crash on NFS re-export node
3467 Details : calling clear_page() on the wrong pointer triggered oops in
3468 generic_mapping_read().
3471 Frequency : rarely, using O_DIRECT IO
3473 Description: unaligned directio crashes client with LASSERT
3474 Details : check for unaligned buffers before trying any requests.
3477 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3479 Description: stale data returned from RAID cache
3480 Details : If only a small amount of IO is done to the RAID device before
3481 reading it again it is possible to get stale data from the RAID
3482 cache instead of reading it from disk.
3485 Frequency : always for sles10 kernel
3487 Description: sles10 support
3488 Details : ll_follow_link: compile fixes and using of nd_set_link
3489 under newer kernels.
3492 Frequency : depends on arch, kernel and compiler version, always on sles10
3495 Description: recursive or deep enough symlinks cause stack overflow
3496 Details : getting rid of large stack-allocated variable in
3500 Frequency : depends on hardware
3502 Description: lustre write performance loss in the SLES10 kernel
3503 Details : the performance loss is caused by using of write barriers in the
3504 ext3 code. The SLES10 kernel turns barrier support on by
3505 default. The fix is to undo that change for ldiskfs.
3507 ------------------------------------------------------------------------------
3509 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3511 * Support for kernels:
3512 2.6.9-42.0.3EL (RHEL 4)
3513 2.6.5-7.276 (SLES 9)
3514 2.4.21-47.0.1.EL (RHEL 3)
3515 2.6.12.6 vanilla (kernel.org)
3519 Frequency : quota enabled and large files being deleted
3521 Description: releasing more than 4GB of quota at once hangs OST
3522 Details : If a user deletes more than 4GB of files on a single OST it
3523 will cause the OST to spin in an infinite loop. Release
3524 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3529 Description: statfs data retrieved from /proc may be stale or zero
3530 Details : When reading per-device statfs data from /proc, in the
3531 {kbytes,files}_{total,free,avail} files, it may appear
3532 as zero or be out of date.
3535 Frequency : systems with MD RAID1 external journal devices
3537 Description: lconf's call to blkid is confused by RAID1 journal devices
3538 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3539 of returning all block devices that match the journal UUID.
3542 Frequency : always, for aggregate stripe size over 4GB
3544 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3545 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3546 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3549 Frequency : always if "lfs find" used on a local file/directory
3551 Description: "lfs find" segfaults if used on a local file/directory
3552 Details : The case where a directory component was not specified wasn't
3553 handled correctly. Handle this properly.
3556 Frequency : always on ppc64
3558 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3559 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3560 perform endianness conversion.
3563 Frequency : rarely (truncate to non-zero file size after write under load)
3564 Bugzilla : 10730, 10687
3565 Description: Files padded with zeros to next 4K multiple
3566 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3567 it is possible that files that are truncated to a non-zero size
3568 immediately after being written are filled with zero bytes beyond
3569 the truncated size. No file data is lost.
3571 Severity : enhancement
3573 Description: Allow recovery/failover for liblustre clients.
3574 Details : liblustre clients were unaware of failover configurations until
3577 Severity : enhancement
3579 Description: user file locks should fail when not mounting with flock option
3580 Details : Set up an error-returning stub in ll_file_operations.lock field
3581 to prevent incorrect behaviour when client is mounted without
3582 flock option. Also, set up properly f_op->flock field for
3586 Frequency : always on ia64
3588 Description: "lfs df" loops on printing out MDS statfs information
3589 Details : The obd_ioctl_data was not initialized and in some systems
3590 this caused a failure during the ioctl that did not return
3591 an error. Initialize the struct and return an error on failure.
3594 Frequency : SLES 9 only
3596 Description: Error of copying files with lustre special EAs as root
3597 Details : Client side always return success for setxattr call for lustre
3598 special xattr (currently only "trusted.lov").
3601 Frequency : rarely on clusters with both ia64+i386 clients
3603 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3604 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3605 clients are doing concurrent writes to the same file it is
3606 possible that the ia64 clients may LASSERT because the OST
3607 extent locks are not PAGE_SIZE aligned. Ensure that grown
3608 locks are always aligned on the request boundary.
3611 Frequency : specific use, occasional
3613 Description: Overwriting in use executable truncates on-disk binary image
3614 Details : If one node attempts to overwrite an executable in use by
3615 another node, we now correctly return ETXTBSY instead of
3616 truncating the file.
3618 Severity : enhancement
3620 Description: Async OSC create to avoid the blocking unnecessarily.
3621 Details : If a OST has no remain object, system will block on the creating
3622 when need to create a new object on this OST. Now, ways use
3623 pre-created objects when available, instead of blocking on an
3624 empty osc while others are not empty. If we must block, we block
3625 for the shortest possible period of time.
3630 Description: chmod on Lustre root is propagated to other clients
3631 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3632 invalid by the follow_mount time.
3635 Frequency : liblustre clients only
3637 Description: Race in 'instant cancel' lock handling could lead to such locks
3638 never to be granted in case of SMP MDS
3639 Details : Do not destroy not yet granted but cbpending locks in
3643 Frequency : replay/resend of open
3645 Description: non null lock assetion failure in mds_intent_policy
3646 Details : Trying to replay/resend lockless open requests resulted in
3647 mds_open() returning 0 with no lock. Now it sets a flag if
3648 a lock is going to be returned.
3650 Severity : enhancement
3652 Description: Checksum enhancements
3653 Details : New checksum enhancements allow for resending RPCs that failed
3656 Severity : enhancement
3658 Description: Tunables on number of dirty pages in cacche
3659 Details : Allow to set limit on number of dirty pages cached.
3664 Description: client crash on unmount - lock still has references
3665 Details : In some error handling cases it was possible to leak a lock
3666 reference on a client while accessing a file. This was not
3667 harmful to the client during operation, but would cause the
3668 client to crash when the filesystem is unmounted.
3671 Frequency : specific case, rare
3673 Description: ETXTBSY on mds though file not in use by client
3674 Details : ETXTBSY is no longer incorrectly returned when attempting to
3675 chmod or chown a directory that the user previously tried to
3676 execute or a currently-executing binary.
3679 Frequency : extremely rare except on liblustre-based clients
3681 Description: Lustre space not freed when files are deleted
3682 Details : Clean up open-unlinked files after client eviction. Previously
3683 the unlink was skipped and the files remained as orphans.
3688 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3689 Details : In some cases it was possible to send a blocking callback to
3690 a client doing a glimpse, even though that client didn't get
3691 a lock granted. When the glimpse lock is cancelled on the OST
3692 the freed lock is left on the waiting list and corrupted the list.
3695 Frequency : all core dumps
3697 Description: Broke core dumps to lustre
3698 Details : Negative dentry may be unhashed if parent does not have UPDATE
3699 lock, but some callers, e.g. do_coredump, expect dentry to be
3700 hashed after successful create, hash it in ll_create_it.
3702 ------------------------------------------------------------------------------
3704 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3706 * Support for kernels:
3707 2.6.9-42.0.2.EL (RHEL 4)
3708 2.6.5-7.276 (SLES 9)
3709 2.4.21-40.EL (RHEL 3)
3710 2.6.12.6 vanilla (kernel.org)
3714 Frequency : always on RHEL 3
3716 Description: Number of open files grows over time
3717 Details : The number of open files grows over time, whether or not
3718 Lustre is started. This was due to a filp leak introduced
3719 by one of our kernel patches.
3721 ------------------------------------------------------------------------------
3723 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3725 * Support for kernels:
3726 2.6.9-42.EL (RHEL 4)
3727 2.6.5-7.276 (SLES 9)
3728 2.4.21-40.EL (RHEL 3)
3729 2.6.12.6 vanilla (kernel.org)
3734 Bugzilla : 5719, 9635, 9792, 9684
3735 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3736 Details : If a server is under heavy load and cannot reply to new
3737 connection requests before the client resends the (re)connect,
3738 the connection handling code can behave badly if two service
3739 threads are concurrently handing separate (re)connections from
3740 the same client. Add better locking to the connection handling
3741 code, and ensure that only a single connection will be processed
3742 for a given client UUID, even if the lock is dropped.
3744 Severity : enhancement
3746 Description: add TCP zero-copy support to kernel
3747 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3748 sends if the hardware supports scatter-gather and checksumming.
3749 This allows socklnd to do client-write and server-read more
3750 efficiently and reduce CPU utilization from skbuf copying.
3753 Frequency : only if NFS exporting from client
3755 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3756 Details : If setting attributes on a file created under NFS that had
3757 never been opened it would be possible to oops the client
3758 if the file had no objects.
3761 Frequency : always for liblustre
3763 Description: liblustre client does MDS+OSTs setattr RPC for each write
3764 Details : When doing a write from a liblustre client, the client
3765 incorrectly issued an RPC to the MDS and each OST the file was
3766 striped over in order to update the timestamps. When writing
3767 with small chunks and many clients this could overwhelm the MDS
3768 with RPCs. In all cases it would slow down the write because
3769 these RPCs are unnecessary.
3771 Severity : enhancement
3773 Description: allow number of MDS service threads to be changed at module load
3774 Details : It is now possible to change the number of MDS service threads
3775 running. Adding "options mds mds_num_threads={N}" to the MDS's
3776 /etc/modprobe.conf will set the number of threads for the next
3777 time Lustre is restarted (assuming the "mds" module is also
3778 reloaded at that time). The default number of threads will
3779 stay the same, 32 for most systems.
3784 Description: OST crash if filesystem is unformatted or corrupt
3785 Details : If an OST is started on a device that has never been formatted
3786 or if the filesystem is corrupt and cannot even mount then the
3787 error handling cleanup routines would dereference a NULL pointer.
3792 Description: NULL pointer deref in llap_from_page.
3793 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3794 incorrect) mapping if the page was truncated/reclaimed while it was
3795 searched for. Check for this condition and skip such pages when
3796 doing readahead. Introduce extra check to llap_from_page() to
3797 verify page->mapping->host is non-NULL (so page is not anonymous).
3800 Frequency : Sometimes when using sys_sendfile
3802 Description: "page not covered by a lock" warnings from ll_readpage
3803 Details : sendfile called ll_readpage without right page locks present.
3804 Now we introduced ll_file_sendfile that does necessary locking
3805 around call to generic_file_sendfile() much like we do in
3809 Frequency : with certain MDS communication failures at client mount time
3811 Description: NULL pointer deref after failed client mount
3812 Details : a client connection request may delayed by the network layer
3813 and not be sent until after the PTLRPC layer has timed out the
3814 request. If the client fails the mount immediately it will try
3815 to clean up before the network times out the request. Add a
3816 reference from the request import to the obd device and delay
3817 the cleanup until the network drops the request.
3820 Frequency : occasionally during client (re)connect
3822 Description: assertion failure during client (re)connect
3823 Details : processing a client connection request may be delayed by the
3824 client or server longer than the client connect timeout. This
3825 causes the client to resend the connection request. If the
3826 original connection request is replied in this interval, the
3827 client may trip an assertion failure in ptlrpc_connect_interpret()
3828 which thought it would be the only running connect process.
3831 Frequency : only with obd_echo servers and clients that are rebooted
3833 Description: kernel BUG accessing uninitialized data structure
3834 Details : When running an obd_echo server it did not start the ping_evictor
3835 thread, and when a client was evicted an uninitialized data
3836 structure was accessed. Start the ping_evictor in the RPC
3837 service startup instead of the OBD startup.
3839 Severity : enhancement
3840 Bugzilla : 10193 (patchless)
3841 Description: Remove dependency on various unexported kernel interfaces.
3842 Details : No longer need reparent_to_init, exit_mm, exit_files,
3843 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3846 Frequency : rare (only users of deprecated and unsupported LDAP config)
3848 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3849 Details : LDAP apparently contains 'lustreName' attributes instead of
3850 'name'. A simple remapping of the name is sufficient.
3853 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3855 Description: waiting_locks_callback trips kernel BUG if client is evicted
3856 Details : Running with the dump_on_timeout debug flag turned on makes
3857 it possible that the waiting_locks_callback() can try to dump
3858 the Lustre kernel debug logs from an interrupt handler. Defer
3859 this log dumping to the expired_lock_main() thread.
3861 Severity : enhancement
3863 Description: Support NFS exporting on 2.6 kernels.
3864 Details : Implement non-rawops metadata methods for NFS server to use without
3865 changing NFS server code.
3868 Frequency : very rare (synthetic metadata workload only)
3870 Description: two racing renames might cause an MDS thread to deadlock
3871 Details : Running the "racer" program may cause one MDS thread to rename
3872 a file from being the source of a rename to being the target of
3873 a rename at exactly the same time that another thread is doing
3874 so, and the second thread has already enqueued these locks after
3875 doing a lookup of the target and is trying to relock them in
3876 order. Ensure that we don't try to re-lock the same resource.
3879 Frequency : only very large systems with liblustre clients
3881 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3882 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3883 all OSTs in parallel. This allows the eviction of stale liblustre
3884 clients to proceed much faster than if they were done in series,
3885 and also offers similar improvements for other set_info RPCs.
3890 Description: excessive CPU usage during initial read phase on client
3891 Details : During the initial read phase on a client, it would agressively
3892 retry readahead on the file, consuming too much CPU and impacting
3893 performance (since 1.4.5.8). Improve the readahead algorithm
3894 to avoid this, and also improve some other common cases (read
3895 of small files in particular, where "small" is files smaller than
3896 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3901 Description: MDS crash when receiving packet with unknown intent.
3902 Details : Do not LBUG in unknown intent case, just return -EFAULT
3904 Severity : enhancement
3905 Bugzilla : 9293, 9385
3906 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3907 Details : Do not serialize getattr (non-intent version) and statfs.
3910 Frequency : occasional, when OST network is overloaded/intermittent
3912 Description: client evicted by OST after bulk IO timeout
3913 Details : If a client sends a bulk IO request (read or write) the OST
3914 may evict the client if it is unresposive to its data GET/PUT
3915 request. This is incorrect if the network is overloaded (takes
3916 too long to transfer the RPC data) or dropped the OST GET/PUT
3917 request. There is no need to evict the client at all, since
3918 the pinger and/or lock callbacks will handle this, and the
3919 client can restart the bulk request.
3922 Frequency : Always when mmapping file with no objects
3924 Description: client crashes when mmapping file with no objects
3925 Details : Check that we actually have objects in a file before doing any
3926 operations on objects in ll_vm_open, ll_vm_close and
3932 Description: Request leak when working with deleted CWD
3933 Details : Introduce advanced request refcount tracking for requests
3934 referenced from lustre intent.
3936 Severity : Enhancement
3938 Description: Cache open file handles on client.
3939 Details : MDS now will return special lock along with openhandle, if
3940 requested and client is allowed to hold openhandle, even if unused,
3941 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3942 closing files for every read/write openration.
3944 Severity : Enhancement
3946 Description: Cache open negative dentries on client when possible.
3947 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3948 negative dentries on lock revocation.
3953 Description: Remounting a client read-only wasn't possible with a zconf mount
3954 Details : It wasn't possible to remount a client read-only with llmount.
3956 Severity : enhancement
3957 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3958 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3959 MPICH is included to add Lustre-specific ADIO interfaces.
3960 This is based closely on the UFS ADIO layer and only differs
3961 in file creation, in order to allow the OST striping to be set.
3962 This is user-contributed code and not supported by CFS.
3967 Description: extended inode attributes (immutable, append-only) work improperly
3968 when 2.4 and 2.6 kernels are used on client/server or vice versa
3969 Details : Introduce kernel-independent values for these flags.
3971 Severity : enhancement
3974 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3975 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3976 granularity. Now, Fractional MB granularity is be supported,
3977 this is very useful for low memory system.
3979 Severity : enhancement
3981 Description: Getattr by fid
3982 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3983 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3984 have locks and avoids getattr rpc requests for it on every stat.
3987 Frequency : Always, for filesystems larger than 2TB
3989 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3990 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3991 filesystems larger than 2TB to be mounted with mballoc enabled.
3994 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3996 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3997 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3998 filesystems larger than 2TB is mounted then the kernel will
3999 silently corrupt the start of the filesystem. CONFIG_LBD is
4000 enabled for all CFS-supported kernels, but the possibility of
4001 this happening with a modified kernel config exists.
4003 Severity : enhancement
4005 Description: add client O_DIRECT support for 2.6 kernels
4006 Details : It is now possible to do O_DIRECT reads and writes to files
4007 in the Lustre client mountpoint on 2.6 kernel clients.
4009 Severity : enhancement
4011 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4012 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4013 parallel, not waiting for response from every OST before sending
4014 a rpc to the next OST.
4019 Description: setattr vs write race when updating file timestamps
4020 Details : Client processes that update a file timestamp into the past
4021 right after writing to the file (e.g. tar) it is possible that
4022 the updated file modification time can be reset to the current
4023 time due to a race between processing the setattr and write RPC.
4025 Severity : enhancement
4027 Description: Bring 'lfs find' closer in line with regular Linux find.
4028 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4029 -print0 options and obtains all the needed info through the lustre
4032 Severity : enhancement
4034 Description: support up to 1024 configured devices on one node
4035 Details : change obd_dev array from statically allocated to dynamically
4036 allocated structs as they are first used to reduce memory usage
4041 Description: Flush dirty partially truncated pages during truncate
4042 Details : Immediatelly flush partially truncated pages in filter_setattr,
4043 this way we completely avoid having any pages in page cache on OST
4044 and can retire ugly workarounds during writes to flush such pages.
4049 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4050 Details : protect inode from truncation within vfs_unlink() context
4051 just take a reference before calling vfs_unlink() and release it
4052 when parent's i_sem is free.
4057 Description: last_id value checked outside lock on OST caused LASSERT failure
4058 Details : If there were multiple MDS->OST object precreate requests in
4059 flight, it was possible that the OST's last object id was checked
4060 outside a lock and incorrectly tripped an assertion. Move checks
4061 inside locks, and discard old precreate requests.
4064 Frequency : always, if extents are used on OSTs
4066 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4067 Details : OSTs using the extents format would not zero the high 16 bits of
4068 the index physical block number. This is not a problem for any
4069 OST filesystems smaller than 16TB, and no kernels support ext3
4070 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4071 new/modified files) and can be fixed for existing filesystems
4072 with e2fsprogs-1.39-cfs1.
4077 Description: import connection selection may be incorrect if timer wraps
4078 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4079 import connections to be ignored if the 32-bit jiffies counter
4080 wraps. Use a 64-bit jiffies counter.
4083 Frequency : very large clusters immediately after boot
4085 Description: LNET request buffers exhausted under heavy short-term load
4086 Details : If a large number of client requests are generated on a service
4087 that has previously never seen so many requests it is possible
4088 that the request buffer growth cannot keep up with the spike in
4089 demand. Instead of dropping incoming requests, they are held in
4090 the LND until the RPC service can accept more requests.
4093 Frequency : Sometimes during replay
4095 Description: Assertion failure in ll_local_open after replay.
4096 Details : If replay happened on an open request reply before we were able
4097 to set replay handler, reply will become not swabbed tripping the
4098 assertion in ll_local_open. Now we set the handler right after
4099 recognising of open request
4102 Frequency : very rare
4104 Description: kernel reports "badness in vsnprintf"
4105 Details : Reading from the "recovery_status" /proc file in small chunks
4106 may cause a negative length in lprocfs_obd_rd_recovery_status()
4107 call to vsnprintf() (which is otherwise harmless). Exit early
4108 if there is no more space in the output buffer.
4110 Severity : enhancement
4112 Description: clear OBD RPC statistics by writing to them
4113 Details : It is now possible to clear the OBD RPC statistics by writing
4114 to the "stats" file.
4119 Description: Client mtime is not the same on different clients after utimes
4120 Details : In some cases, the client was using the utimes() syscall on
4121 a file cached on another node. The clients now validate the
4122 ctime from the MDS + OSTs to determine which one is right.
4127 Description: Inability to activate failout mode
4128 Details : lconf script incorrectly assumed that in pythong string's numeric
4129 value is used in comparisons.
4132 Frequency : always with multiple stripes per file
4134 Description: Inefficient object allocation for mutli-stripe files
4135 Details : When selecting which OSTs to stripe files over, for files with
4136 a stripe count that divides evenly into the number of OSTs,
4137 the MDS is always picking the same starting OST for each file.
4138 Return the OST selection heuristic to the original design.
4143 Description: mount failures may take full timeout to return an error
4144 Details : Under some heavy load conditions it is possible that a
4145 failed mount can wait for the full obd_timeout interval,
4146 possibly several minutes, before reporting an error.
4147 Instead return an error as soon as the status is known.
4149 Frequency : quota enabled and large files being deleted
4151 Description: releasing more than 4GB of quota at once hangs OST
4152 Details : If a user deletes more than 4GB of files on a single OST it
4153 will cause the OST to spin in an infinite loop. Release
4154 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4159 Description: statfs data retrieved from /proc may be stale or zero
4160 Details : When reading per-device statfs data from /proc, in the
4161 {kbytes,files}_{total,free,avail} files, it may appear
4162 as zero or be out of date.
4165 Frequency : systems with MD RAID1 external journal devices
4167 Description: lconf's call to blkid is confused by RAID1 journal devices
4168 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4169 of returning all block devices that match the journal UUID.
4172 Frequency : always, for aggregate stripe size over 4GB
4174 Description: assertion fails when trying to use 4GB stripe size
4175 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4176 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4179 Frequency : always on ppc64
4181 Description: the first write on an ext3 filesystem with mballoc got stuck
4182 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4183 perform endianness conversion.
4185 ------------------------------------------------------------------------------
4187 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4189 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4190 INTEROPERATE with older versions automatically. Please read the
4191 user documentation before upgrading any part of a live system.
4192 * WARNING: Lustre networking configuration changes are required with
4193 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4196 * Support for kernels:
4197 2.6.9-22.0.2.EL (RHEL 4)
4198 2.6.5-7.244 (SLES 9)
4199 2.6.12.6 vanilla (kernel.org)
4202 Severity : enhancement
4203 Bugzilla : 7981/8208
4204 Description: Introduced Lustre Networking (LNET)
4205 Details : LNET is new networking infrastructure for Lustre, it includes
4206 a reorganized network configuration mode (see the user
4207 documentation for full details) as well as support for routing
4208 between different network fabrics. Lustre Networking Devices
4209 (LNDs) for the supported network fabrics have also been
4210 created for this new infrastructure.
4212 Severity : enhancement
4213 Description: Introduced Access control lists
4214 Details : clients can set ACLs on files and directories in order to have
4215 more fine-grained permissions than the standard Unix UGO+RWX.
4216 The MDS must be started with the "-o acl" mount option.
4218 Severity : enhancement
4219 Description: Introduced filesystem quotas
4220 Details : Administrators may now establish per-user quotas on the
4223 Severity : enhancement
4225 Description: Configuration change for the XT3
4226 The PTLLND is now used to run Lustre over Portals on the XT3
4227 The configure option(s) --with-cray-portals are no longer used.
4228 Rather --with-portals=<path-to-portals-includes> is used to
4229 enable building on the XT3. In addition to enable XT3 specific
4230 features the option --enable-cray-xt3 must be used.
4235 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4236 Details : A race between allocating a new llcd and re-getting the llcd_lock
4237 allowed another thread to grab newly-allocated llcd.
4239 Severity : enhancement
4241 Description: 2.6 OST async journal commit and locking fix to improve performance
4242 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4243 2.6 kernels are now async as they already were in 2.4 kernels so
4244 that they can commit concurrently with the network bulk transfer.
4245 For block-allocated files the filter allocation semaphore is held
4246 to avoid filesystem fragmentation during allocation. BKL lock
4247 removed for 2.6 xattr operations where it is no longer needed.
4252 Description: lconf incorrectly determined whether two IP networks could talk
4253 Details : In some more complicated routing and multiple-network
4254 configurations, lconf will avoid trying to make a network
4255 connection to a disjoint part of the IP space. It was doing the
4256 math incorrectly for one set of cases.
4261 Description: Fix for potential infinite loop processing records in an llog.
4262 Details : If an llog record is corrupted/zeroed, it is possible to loop
4263 forever in llog_process(). Validate the llog record length
4264 and skip the remainder of the block on error.
4267 Frequency : occasional (liblustre only)
4269 Description: liblustre could not open files whose last component is a symlink
4270 Details : sysio_path_walk() would incorrectly pass the open intent to
4271 intermediate path components.
4274 Frequency : rare (liblustre only with non-standard tuning)
4275 Bugzilla : 7201 (7350)
4276 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4277 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4278 after it has been released, only do this for referenced locks
4280 Severity : enhancement
4282 Description: specifying an (invalid) directory default stripe_size of -1
4283 would reset the directory default striping
4284 Details : stripe_size -1 was used internally to signal directory stripe
4285 removal, now use "all default" to signal dir stripe removal
4286 as a directory striping of "all default" is not useful
4289 Frequency : common for large clusters running liblustre clients
4291 Description: doing an ls when liblustre clients are running is slow
4292 Details : sending a glimpse AST to a liblustre client waits for every AST
4293 to time out, as liblustre clients will not respond. Since they
4294 cannot cache data we refresh the OST lock LVB from disk instead.
4296 Severity : enhancement
4298 Description: doing an ls at the same time as file IO can be slow
4299 Details : enqueue and other "small" requests can be blocked behind many
4300 large IO requests. Create a new OST IO portal for non-IO
4301 requests so they can be processed faster.
4304 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4306 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4307 created file with mode 000
4308 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4309 allow an ftruncate() as a non-root user to a file with mode 000.
4310 Lustre now respects this flag to disable mode checks when
4311 truncating a file owned by the user
4314 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4317 Description: Revoking locks from clients that went dead or catatonic might take
4319 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4320 cancellation of such locks instant on servers without waiting for
4321 any reply from clients. Clients drops these locks when cancel
4322 notification from server is received without replying.
4325 Frequency : liblustre-only, when liblustre client dies or becomes busy
4327 Description: Doing ls on Linux clients can take a long time with active
4329 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4330 granting such locks to it in the first place if possible. Locks
4331 are taken by proxy on the OST during the read or write and
4332 dropped immediately afterward. Add connect flags handling, do
4333 not grant locks to liblustre clients for glimpse ASTs.
4335 Severity : enhancement
4337 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4338 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4339 sequential reads and behaves badly with applications doing random
4340 reads. Improve it by reading ahead at least read region, and
4341 avoiding excessive large RPC for small reads.
4343 Severity : enhancement
4345 Description: Creating more than 1000 files for a single job may cause a load
4346 imbalance on the OSTs if there are also a large number of OSTs.
4347 Details : qos_prep_create() uses an OST index reseed value that is an
4348 even multiple of the number of available OSTs so that if the
4349 reseed happens in the middle of the object allocation it will
4350 still utilize the OSTs as uniformly as possible.
4355 Description: OST or MDS may oops in ping_evictor_main()
4356 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4357 but doesn't restart at beginning of obd_exports_timed list
4360 Severity : enhancement
4362 Description: improve by-nid export eviction on the MDS and OST
4363 Details : allow multiple exports with the same NID to be evicted at one
4364 time without re-searching the exports list.
4367 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4369 Description: MDS may oops in groups_free()
4370 Details : in rare race conditions a newly allocated group_info struct is
4371 freed again, and this can be NULL. The 2.4 compatibility code
4372 for groups_free() checked for a NULL pointer, but 2.6 did not.
4375 Frequency : common for liblustre clients doing little filesystem IO
4376 Bugzilla : 9352, 7313
4377 Description: server may evict liblustre clients accessing contended locks
4378 Details : if a client is granted a lock or receives a completion AST
4379 with a blocking AST already set it would not reply to the AST
4380 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4383 Frequency : lfs setstripe, only systems with more than 160 OSTs
4385 Description: unable to set striping with a starting offset beyond OST 160
4386 Details : llapi_create_file() incorrectly limited the starting stripe
4387 index to the maximum single-file stripe count.
4390 Frequency : LDAP users only
4392 Description: lconf did not handle in-kernel recovery with LDAP properly
4393 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4395 Severity : enhancement
4397 Description: bind OST threads to NUMA nodes to improve performance
4398 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4399 node and do their allocations there to localize memory access
4401 Severity : enhancement
4403 Description: llmount can determine client NID directly from Myrinet (GM)
4404 Details : the client NID code from gmnalnid was moved directly into
4405 llmount, removing the need to use this or specifying the
4406 client NID explicitly when mounting GM clients with zeroconf
4409 Frequency : if client is started with down MDS
4411 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4412 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4413 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4414 ptlrpc_import_delayed_req() because we didn't check if the
4415 request was interrupted, and we also didn't break out of the
4416 event loop if there was no timeout
4421 Description: data loss during non-page-aligned writes to a single file from
4422 both multiple nodes and multiple threads on one node at same time
4423 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4424 inconsistency led to false short-reads, that were cached and later
4425 used by ->prepare_write() to fill in partially written page,
4426 leading to data loss.
4429 Frequency : always, if lconf --abort_recovery used
4431 Description: lconf --abort_recovery fails with 'Operation not supported'
4432 Details : lconf was attempting to abort recovery on the MDT device and not
4435 Severity : enhancement
4437 Description: remove cleanup logs
4438 Details : replace lconf-generated cleanup logs with lustre internal
4439 cleanup routines. Eliminates the need for client-cleanup and
4442 Severity : enhancement
4444 Description: add support for EAs (user and system) on lustre filesystems
4445 Details : it is now possible to store extended attributes in the Lustre
4446 client filesystem, and with the user_xattr mount option it
4447 is possible to allow users to store EAs on their files also
4449 Severity : enhancement
4451 Description: Add possibility (config option) to show minimal available OST free
4453 Details : When compiled with --enable-mindf configure option, statfs(2)
4454 (and so, df) will return least minimal free space available from
4455 all OSTs as amount of free space on FS, instead of summary of
4456 free spaces of all OSTs.
4458 Severity : enhancement
4460 Description: do not expand extent locks acquired on OST-side
4461 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4462 by server: they are not cached anyway.
4465 Frequency : when mmap is used/binaries executed from Lustre
4467 Description: Unmmap pages before throwing them away from read cache.
4468 Details : llap_shrink cache now attempts to unmap pages before discarding
4469 them (if unmapping failed - do not discard). SLES9 kernel has
4470 extra checks that trigger if this unmapping is not done first.
4475 Description: lconf didn't resolve symlinks before checking to see whether a
4476 given mountpoint was already in use
4479 Frequency : when migrating failover services
4480 Bugzilla : 6395, 9514
4481 Description: When migrating a subset of services from a node (e.g. failback
4482 from a failover service node) the remaining services would
4483 time out and evict clients.
4484 Details : lconf --force (implied by --failover) sets the global obd_timeout
4485 to 5 seconds in order to quickly disconnect, but this caused
4486 other RPCs to time out too quickly. Do not change the global
4487 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4489 Severity : enhancement
4490 Frequency : if MDS is started with down OST
4491 Bugzilla : 9439,5706
4492 Description: Allow startup/shutdown of an MDS without depending on the
4493 availability of the OSTs.
4494 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4495 Add appropriate locking and lov-osc refcounts for safe
4496 cleaning. Add osc abort_inflight calls in case the
4497 synchronize never started.
4500 Frequency : occasional (Cray XT3 only)
4502 Description: root not authorized to access files in CRAY_PORTALS environment
4503 Details : The client process capabilities were not honoured on the MDS in
4504 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4505 been accessed by an authorized user then root was able to access
4506 the file on the local client also. The root user capabilities
4507 are now allowed on the MDS, as this environment has secure UID.
4510 Frequency : occasional
4512 Description: ldiskfs "too long searching" message happens too often
4513 Details : A debugging message (otherwise harmless) prints too often on
4514 the OST console. This has been reduced to only happen when
4515 there are fragmentation problems on the filesystem.
4520 Description: Division by zero in statfs when all OSCs are inactive
4521 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4522 lov_statfs divides by value returned by lov_get_stripecnt().
4526 Bugzilla : 9489, 3273
4527 Description: First write from each client to each OST was only 4kB in size,
4528 to initialize client writeback cache, which caused sub-optimal
4529 RPCs and poor layout on disk for the first writen file.
4530 Details : Clients now request an initial cache grant at (re)connect time
4531 and so that they can start streaming writes to the cache right
4532 away and always do full-sized RPCs if there is enough data.
4533 If the OST is rebooted the client also re-establishes its grant
4534 so that client cached writes will be honoured under the grant.
4539 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4540 Details : Now I/O RPCs go to different portal number and (presumably) fast
4541 lock requests (and glimses) and other RPCs get their own service
4542 threads pool that should be able to service those RPCs
4545 Severity : enhancement
4547 Description: Ability to exchange lustre version between client and servers and
4548 issue warnings at client side if client is too old. Also for
4549 liblustre clients there is ability to refuse connection of too old
4551 Details : New 'version' field is added to connect data structure that is
4552 filled with version info. That info is later checked by server and
4556 Frequency : rare, liblustre only.
4557 Bugzilla : 9296, 9581
4558 Description: Two simultaneous writes from liblustre at offset within same page
4559 might proceed at the same time overwriting eachother with stale
4561 Details : I/O lock withing llu_file_prwv was released too early, before data
4562 actually was hitting the wire. Extended lock-holding time until
4563 server acknowledges receiving data.
4566 Frequency : extremely rare. Never observed in practice.
4568 Description: avoid generating lustre_handle cookie of 0.
4569 Details : class_handle_hash() generates handle cookies by incrementing
4570 global counter, and can hit 0 occasionaly (this is unlikely, but
4571 not impossible, because initial value of cookie counter is
4572 selected randonly). Value of 0 is used as a sentinel meaning
4573 "unassigned handle" --- avoid it. Also coalesce two critical
4574 sections in this function into one.
4576 Severity : enhancement
4578 Description: allow liblustre clients to delegate truncate locking to OST
4579 Details : To avoid overhead of locking, liblustre client instructs OST to
4580 take extent lock in ost_punch() on client's behalf. New connection
4581 flag is added to handle backward compatibility.
4583 Severity : enhancement
4584 Bugzilla : 4928, 7341, 9758
4585 Description: allow number of OST service threads to be specified
4586 Details : a module parameter allows the number of OST service threads
4587 to be specified via "options ost ost_num_threads={N}" in the
4588 OSS's /etc/modules.conf or /etc/modprobe.conf.
4592 Bugzilla : 6146, 9635, 9895
4593 Description: servers crash with bad pointer in target_handle_connect()
4594 Details : In rare cases when a client is reconnecting it was possible that
4595 the connection request was the last reference for that export.
4596 We would temporarily drop the export reference and get a new
4597 one, but this may have been the last reference and the export
4598 was just destroyed. Get new reference before dropping old one.
4600 Severity : enhancement
4601 Frequency : if client is started with failover MDS
4603 Description: Allow multiple MDS hostnames in the mount command
4604 Details : Try to read the configuration from all specified MDS
4605 hostnames during a client mount in case the "primary"
4608 Severity : enhancement
4610 Description: Stop sending data to evicted clients as soon as possible.
4611 Details : Check if the client we are about to send or are sending data to
4612 was evicted already. (Check is done every second of waiting,
4613 for which l_wait_event interface was extended to allow checking
4614 of exit condition at specified intervals).
4617 Frequency : rare, normally only when NFS exporting is done from client
4619 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4621 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4624 Frequency : always, if srand() is called before liblustre initialization
4626 Description: Liblustre uses system PRNG disturbing its usage by user application
4627 Details : Introduce internal to lustre fast and high-quality PRNG for
4628 lustre usage and make liblustre and some other places in generic
4629 lustre code to use it.
4631 Severity : enhancement
4632 Bugzilla : 9477, 9557, 9870
4633 Description: Verify that the MDS configuration logs are updated when xml is
4634 Details : Check if the .xml configuration logs are newer than the config
4635 logs stored on the MDS and report an error if this is the case.
4636 Request --write-conf, or allow starting with --old_conf.
4638 Severity : enhancement
4640 Description: Handle symlinks in the path when checking if Lustre is mounted.
4641 Details : Resolve intermediate symlinks when checking if a client has
4642 mounted a filesystem to avoid duplicate client mounts.
4647 Description: lconf can hit an error exception but still return success.
4648 Details : The lconf command catches the Command error exception at the top
4649 level script context and will exit with the associated exit
4650 status, but doesn't ensure that this exit status is non-zero.
4655 Description: failure of ptlrpc thread startup can cause oops
4656 Details : Starting a ptlrpc service thread can fail if there are a large
4657 number of threads or the server memory is very fragmented.
4658 Handle this without oopsing.
4661 Frequency : always, only if liblustre and non-default acceptor port was used
4663 Description: liblustre cannot connect to servers with non-default acceptor port
4664 Details : tcpnal_set_default_params() was not called and was therefore
4665 ignoring the environment varaible TCPNAL_PORT, as well as other
4666 TCPNAL_ environment variables
4671 Description: two objects could be created on the same OST for a single file
4672 Details : If an OST is down, in some cases it was possible to create two
4673 objects on a single OST for a single file. No problems other
4674 than potential performance impact and spurious error messages.
4678 Bugzilla : 5681, 9562
4679 Description: Client may oops in ll_unhash_aliases
4680 Details : Client dcache may become inconsistent in race condition.
4681 In some cases "getcwd" can fail if the current directory is
4687 Description: Inode refcounting problems in NFS export code
4688 Details : link_raw functions used to call d_instantiate without obtaining
4689 extra inode reference first.
4693 Bugzilla : 9942, 9903
4694 Description: Referencing freed requests leading to crash, memleaks with NFS.
4695 Details : We used to require that call to ll_revalidate_it was always
4696 followed by ll_lookup_it. Also with revalidate_special() it is
4697 possible to call ll_revalidate_it() twice for the same dentry
4698 even if first occurence returned success. This fix changes semantic
4699 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4700 reference on a request referred from the intent.
4701 ll_intent_release() then releases such a request.
4704 Frequency : rare, normally benchmark loads only
4706 Description: unlinked inodes were kept in memory on the client
4707 Details : If a client is repeatedly creating and unlinking files it
4708 can accumulate a lot of stale inodes in the inode slab cache.
4709 If there is no other client load running this can cause the
4710 client node to run out of memory. Instead flush old inodes
4711 from client cache that have the same inode number as a new inode.
4714 Frequency : SLES9 2.6.5 kernel and long filenames only
4715 Bugzilla : 9969, 10379
4716 Description: utime reports stale NFS file handle
4717 Details : SLES9 uses out-of-dentry names in some cases, which confused
4718 the lustre dentry revalidation. Change it to always use the
4722 Frequency : rare, unless heavy write-truncate concurrency is continuous
4723 Bugzilla : 4180, 6984, 7171, 9963, 9331
4724 Description: OST becomes very slow and/or deadlocked during object unlink
4725 Details : filter_destroy() was holding onto the parent directory lock
4726 while truncating+unlinking objects. For very large objects this
4727 may block other threads for a long time and slow overall OST
4728 responsiveness. It may also be possible to get a lock ordering
4729 deadlock in this case, or run out of journal credits because of
4730 the combined truncate+unlink. Solution is to do object truncate
4731 first in one transaction without parent lock, and then do the
4732 final unlink in a new transaction with the parent lock. This
4733 reduces the lock hold time dramatically.
4736 Frequency : rare, 2.4 kernels only
4738 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4739 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4740 before unmounting their filesystem in order to avoid deadlock.
4741 The kernel_locked() function in 2.4 kernels only checks whether
4742 the kernel lock is held, not whether it is this process that is
4743 holding it as 2.6 kernels do.
4748 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4749 Details : The client ptlrpc code may be trying to reconnect to a down
4750 server before a previous connection attempt has timed out.
4751 Increase the reconnect interval to be longer than the connection
4752 timeout interval to avoid sending duplicate connections to
4756 Frequency : echo_client brw_test command
4758 Description: fix echo_client to work with OST preallocated code
4759 Details : OST preallocation code (5137) didn't take echo_client IO path
4760 into account: echo_client calls filter methods outside of any
4761 OST thread and, hence, there is no per-thread preallocated
4762 pages and buffers to use. Solution: hijack pga pages for IO. As
4763 a byproduct, this avoids unnecessary data copying.
4767 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4768 Description: Client can oops in mdc_commit_close() after open replay
4769 Details : It was possible for the MDS to return an open request with no
4770 transaction number in mds_finish_transno() if the client was
4771 evicted, but without actually returning an error. Clients
4772 would later try to replay that open and may trip an assertion
4773 Simplify the client close codepath, and always return an error
4774 from the MDS in case the open is not successful.
4777 Frequency : rare, 2.6 OSTs only
4779 Description: OST may deadlock under high load on fragmented files
4780 Details : If there was a heavy load and highly-fragmented OST filesystems
4781 it was possible to have all the OST threads deadlock waiting on
4782 allocation of biovecs, because the biovecs were not released
4783 until the entire RPC IO was completed. Instead, release biovecs
4784 as soon as they are complete to ensure forward IO progress.
4786 Severity : enhancement
4788 Description: Support for specifying external journal device at mount
4789 Details : If an OST or MDS device is formatted with an external journal
4790 device, this device major/minor is stored in the ext3 superblock
4791 and may not be valid for failover. Allow detecting and
4792 specifying the external journal at mount time.
4797 Description: Mounting an MDS with pending unlinked files may cause oops
4798 Details : target_finish_recovery() calls mds_postrecov() which returned
4799 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4800 considers this an error and immediately begins cleaning up the
4801 lov, just after starting the mds_lov process
4803 Severity : enhancement
4805 Description: Implement 'lfs df' to report actual free space on per-OST basis
4806 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4807 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4808 usage of objects; '-h' to report in human readable format.
4810 ------------------------------------------------------------------------------
4812 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4819 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4820 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4821 at startup during orphan unlinking. Instead, skip buddy update
4822 before setup, as it will be regenerated after recovery is complete.
4827 Description: If an OST is inactive, its locks might reference stale inodes.
4828 Details : lov_change_cbdata() must iterate over all namespaces, even if
4829 they are inactive to clear inode references from the lock.
4831 Severity : enhancement
4832 Frequency : occasional, if non-standard max_dirty_mb used
4834 Description: Client will block write RPCs if not enough grant
4835 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4836 then the client will block writes while waiting for another RPC
4837 to complete instead of consuming its dirty limit. With change
4838 we get improved performance when max_dirty_mb is small.
4840 Severity : enhancement
4841 Bugzilla : 3389, 6253
4842 Description: Add support for supplementary groups on the MDS.
4843 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4844 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4845 lookups for user supplementary groups into a cache.
4849 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4850 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4851 for liblustre O_EXCL is enforced by MDS.
4855 Description: readdir never returns NULL in liblustre.
4856 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4857 dentry in struct dirent.
4861 Frequency : liblustre only, depends on application IO pattern
4862 Description: liblustre clients evicted if not contacting servers
4863 Details : Don't put liblustre clients into the ping_evictor list, so
4864 they will not be evicted by the pinger ever.
4866 Severity : enhancement
4868 Description: Add ability to evict clients by NID from MDS.
4869 Details : By echoing "nid:$NID" string into
4870 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4871 $NID would be instantly evicted from this MDS and from all active
4872 OSTs connected to it.
4876 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4877 Details : lookup_it_finish() used to query file size from OSTs that was not
4882 Description: service threads change working directory to that of init
4883 Details : Starting lustre service threads may pin the working directory
4884 of the parent thread, making that filesystem busy. Threads
4885 now change to the working directory of init to avoid this.
4889 Frequency : during shutdown only
4890 Description: shutdown with a failed MDS or OST can cause unmount to hang
4891 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4896 Frequency : 2.6 only
4897 Description: chmod/chown may include an extra supplementary group
4898 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4899 supplementary group and if none is specified this is used.
4902 Bugzilla : 5479 (6816)
4904 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4905 Details : If lookup is in progress on a file that is unlinked we might try
4906 to revalidate the inode and fail in revalidate after lookup is
4907 complete and ll_file_open() enqueues the open again but
4908 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4911 Frequency : always, if lconf --abort_recovery used
4913 Description: lconf --abort_recovery fails with 'Operation not supported'
4914 Details : lconf was attempting to abort recovery on the MDT device and not
4917 ------------------------------------------------------------------------------
4919 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4924 Frequency : rare (only unsupported configurations with a node running as an
4926 Bugzilla : 6514, 5137
4927 Description: Mounting a Lustre file system on a node running as an OST could
4929 Details : OSTs now preallocates memory needed to write out data at
4930 startup, instead of when needed, to avoid having to
4931 allocate memory in possibly low memory situations.
4932 Specifically, if the file system is mounted on on OST,
4933 memory pressure could force it to try to write out data,
4934 which it needed to allocate memory to do. Due to the low
4935 memory, it would be unable to do so and the node would
4936 become unresponsive.
4938 Severity : enhancement
4940 Description: Addition of lconf --service command line option
4941 Details : lconf now accepts a '--service <arg>' option, which is
4942 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4944 Severity : enhancement
4946 Description: Failover mode is now the default for OSTs.
4947 Details : By default, OSTs will now run in failover mode. To return to
4948 the old behaviour, add '--failout' to the lmc line for OSTs.
4950 Severity : enhancement
4952 Description: Health checks are now provided for MDS and OSTs
4953 Details : Additional detailed health check information on MSD and OSTs
4954 is now provided through the procfs health_check value.
4957 Frequency : occasional, depends on IO load
4959 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4960 after numerous create/delete cycles
4961 Details : The ext3 inode allocation policy would not allocate new inodes
4962 very well on the OSTs because there are no new directories
4963 being created. Instead we look for groups with free space if
4964 the parent directories are nearly full.
4969 Description: Network or server problems during mount may cause partially
4970 mounted clients instead of returning an error.
4971 Details : The config llog parsing code may overwrite the error return
4972 code during mount error handling, returning success instead
4978 Description: MDS can fail to allocate large reply buffers
4979 Details : After long uptimes the MDS can fail to allocate large reply
4980 buffers (e.g. zconf client mount config records) due to memory
4981 fragmentation or consumption by the buffer cache. Preallocate
4982 some large reply buffers so that these replies can be sent even
4983 under memory pressure.
4987 Frequency : rare (liblustre)
4988 Description: fsx running with liblustre complained that using truncate() to
4989 extend the file doesn't work. This patch corrects that issue.
4990 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4991 ATTR_SIZE and lsm use in llu_setattr_raw.
4995 Frequency : rare, only 2.6 kernels
4996 Description: Unusual file access patterns on the MDS may result in inode
4997 data being lost in very rare circumstances.
4998 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4999 code in the 2.6 kernel caused the inode and/or EA data not to
5000 be read from disk, causing single-file corruption.
5004 Frequency : rare, only 2.6 filesystems using extents
5005 Description: Heavy concurrent write and delete load may cause data corruption.
5006 Details : It was possible under high-load situations to have an extent
5007 metadata block in the block device cache from a just-unlinked
5008 file overwrite a newly-allocated data block. We now unmap any
5009 metadata buffers that alias just-allocated data blocks.
5013 Frequency : filesystems with default stripe_count larger than 77
5014 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5015 Details : lconf specifies an inode size of 4096 bytes when the default
5016 stripe_count is larger than 77. This conflicts with the default
5017 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5018 this case to avoid pinning too much memory for large EAs.
5020 ------------------------------------------------------------------------------
5022 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5027 Frequency : rare (extremely heavy IO load with hundreds of clients)
5029 Description: Client is evicted, gets IO error writing to file
5030 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5031 caused clients to be evicted for AST timeouts. The fixes in
5032 bug 5192 mean we no longer need to have such short AST timeouts
5033 so ldlm_timeout has been increased.
5036 Frequency : occasional during --force or --failover shutdown under load
5037 Bugzilla : 5949, 4834
5038 Description: Server oops/LBUG if stopped with --force or --failover under load
5039 Details : a collection of import/export refcount and cleanup ordering
5040 issues fixed for safer force cleanup
5043 Frequency : only filesystems larger than 120 OSTs
5044 Bugzilla : 5990, 6223
5045 Description: lfs getstripe would oops on a very large filesystem
5046 Details : lov_getconfig used kfree on vmalloc'd memory
5049 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5050 Bugzilla : 6242, 6243
5051 Description: reading from files that had been truncated to a non-zero size
5052 but never opened returned no data
5053 Details : ll_file_read() reads zeros from no-object files to EOF
5058 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5059 Details : There's little harm in aborting MDS/OSS recovery and letting it
5060 try again, so I removed the LASSERT and return an error instead.
5062 Severity : enhancement
5064 Description: New debugging infrastructure for tracking down data corruption
5065 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5066 (b) cover more of the client-side code path, and (c) try to narrow
5067 down where problems occurred
5071 Bugzilla : 3819, 4364, 4397, 6313
5072 Description: Racing close and eviction MDS could cause assertion in mds_close
5073 Details : It was possible to get multiple mfd references during close and
5074 client eviction, leading to one thread referencing a freed mfd.
5076 Severity: : enhancement
5077 Bugzilla : 3262, 6359
5078 Description: Attempts to reconnect to servers are now more aggressive.
5079 Details : This builds on the enhanced upcall-less recovery that was added
5080 in 1.4.2. When trying to reconnect to servers, clients will
5081 now try each server in the failover group every 10 seconds. By
5082 default, clients would previously try one server every 25 seconds.
5087 Description: After recovery, certain operations trigger a failed
5088 assertion on a client.
5089 Details : Failing over an mds, using lconf -d --failover, while a
5090 client was doing a readdir() call would cause the client to
5091 LBUG after recovery completed and the readdir() was resent.
5093 Severity : enhancement
5095 Description: Default groups are now added by lconf
5096 Details : You can now run lconf --group <servicename> without having to
5097 manually add groups with lmc.
5100 Frequency : occasional
5102 Description: Nodes with an elan id of 0 trigger a failed assertion
5105 Frequency : always when accessing e.g. tty/console device nodes
5107 Description: tty and some other devices nodes cannot be used on lustre
5108 Details : file's private_data field is used by device data and lustre
5109 values in there got lost. New field was added to struct file to
5110 store fs-specific private data.
5113 Frequency : when exporting Lustre via NFS
5115 Description: NFSD failed occasionally when looking up a path component
5116 Details : NFSD is looking up ".." which was broken in ext3 directories
5117 that had grown large enough to become hashed.
5120 Frequency : Clusters with multiple interfaces not on the same subnet
5122 Description: Nodes will repeatedly try to reconnect to an interface which it
5123 cannot reach and report an error to the log.
5124 Details : Extra peer list entries will be created by lconf with some peers
5125 unreachable. lconf now validates the peer before adding it.
5128 Frequency : Only if a default stripe is set on the filesystem root.
5130 Description: Setting a default stripe on the filesystem root prevented the
5131 filesystem from being remounted.
5132 Details : The client was sending extra request flags in the root getattr
5133 request and did not allocate a reply buffer for the dir EA.
5136 Frequency : occasional, higher if lots of files are accessed by one client
5137 Bugzilla : 6159, 6097
5138 Description: Client trips assertion regarding lsm mismatch/magic
5139 Details : While revalidating inodes the VFS looks up inodes with ifind()
5140 and in rare cases can find an inode that is being freed.
5141 The ll_test_inode() code will free the lsm during ifind()
5142 when it finds an existing inode and then the VFS later attaches
5143 this free lsm to a new inode.
5147 Bugzilla : 6422, 7030
5148 Description: MDS deadlock between mkdir and client eviction
5149 Details : Creating a new file via mkdir or mknod (starting a transaction
5150 and getting the ns lock) can deadlock with client eviction
5151 (gets ns lock and trying to finish a synchronous transaction).
5154 Frequency : occasional
5155 Description: While starting a server, the fsfilt_ext3 module could not be
5157 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5158 kernels. Previously, lconf would still use the ext3 name
5159 when trying to load modules. Now, it will correctly use
5160 ext3 on 2.4 and ldiskfs on 2.6.
5162 Severity : enhancement
5163 Description: The default stripe count has been changed to 1
5164 Details : The interpretation of the default stripe count (0, to lfs
5165 or lmc) has been changed to mean striping across a single
5166 OST, rather than all available. For general usage we have
5167 found a stripe count of 1 or 2 works best.
5169 Severity : enhancement
5170 Description: Add support for compiling against Cray portals.
5171 Details : Conditional compiling for some areas that are different
5175 Frequency : occasional
5176 Bugzilla : 6409, 6834
5177 Description: Creating files with an explicit stripe count may lead to
5178 a failed assertion on the MDS
5179 Details : If some OSTs are full or unavailable, creating files may
5180 trigger a failed assertion on the MDS. Now, Lustre will
5181 try to use other servers or return an error to the
5185 Frequency : occasional
5187 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5188 caused lock timeout during readahead (since 1.4.2).
5189 Details : Processes doing readahead might match a lock that hasn't been
5190 granted yet if there are overlapping and conflicting lock
5191 requests. The readahead process waits on ungranted lock
5192 (original lock is CBPENDING), while OST waits for that process
5193 to cancel CBPENDING read lock and eventually evicts client.
5195 Severity : enhancement
5197 Description: Initial enabling of flock support for clients
5198 Details : Implements fcntl advisory locking and file status functions.
5199 This feature is provided as an optional mount flag (default
5200 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5201 locking are implemented yet, and those that are are not guaranteed
5202 to be completely correct in production environments.
5203 mount -t lustre -o [flock|noflock] ...
5206 Frequency : occasional
5208 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5209 assertion in the ext3 JBD (journaling) layer.
5210 Details : The b_committed_data struct is protected by the big kernel lock
5211 in 2.4 kernels, serializing journal_commit_transaction() and
5212 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5213 access to this struct. In 2.6 kernels there is finer grained
5214 locking to improve SMP performance of the JBD layer.
5218 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5223 Description: merge and cleanup kernel patches.
5226 -----------------------------------------------------------------------------
5228 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5230 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5231 versions of Lustre. Please update all servers and clients to
5232 version 1.4.2 or later at the same time. You must also run
5233 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5234 to update the configuration logs.
5236 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5237 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5238 - swab llog records of type '0' so we get proper header size/idx (5861)
5239 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5240 - fix rename of one directory over another leaking an inode (5953)
5241 - avoid SetPageDirty on 2.6 (5981)
5242 - don't re-add just-being-destroyed locks to the waiting list (5653)
5243 - when creating new directories, inherit the parent's custom
5244 striping settings if present parent (3048)
5245 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5246 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5247 - don't hold client locks on temporary worklist from l_lru (5666)
5248 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5249 - automatically evict dead clients (5921)
5250 - Update file size properly in create+truncate+fstat case (6196)
5251 - Do not unhash mountpoint dentries, do not allow removal of
5253 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5254 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5255 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5256 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5257 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5258 - Correctly handle reads of files with no objects (6243)
5259 - lctl recover will also mark a device active if deactivate used (5933)
5261 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5262 - allow --write-conf on an MDS with different nettype than client (5619)
5263 - don't write config llogs to MDS for mounts not from that MDS (5617)
5264 - lconf should create multiple TCP connections from a client (5201)
5265 - init scripts are now turned off by default; run chkconfig --on
5266 lustre and chkconfig --on lustrefs to use them
5267 - upcalls are no longer needed for clients to recover to failover
5269 - add --abort-recovery option to lconf to abort recovery on device
5271 - add support for an arbitrary number of OSTs (3026)
5272 - Quota support protocol changes.
5273 - forward compatibility changes to wire structs (6007)
5274 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5275 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5276 - improved "lustre status" script
5277 - initialize blocksize for non-regular files (6062)
5278 - added --disable-server and --disable-client configure options (5782)
5279 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5280 - Vanilla 2.4.29 support
5281 - increase maximum number of obd devices to 520 (6242)
5282 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5283 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5284 SLES 9 2.6.5 kernels
5285 - limit stripes per file to 160 (the maximum EA size) (6093)
5287 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5290 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5291 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5292 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5293 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5294 - small ext3 extents cleanups and fixes (5733)
5295 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5296 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5297 - avoid needless client->OST connect, fix handle mismatch (5317)
5298 - fix DLM error path that led to out-of-sync client, long delays (5779)
5299 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5300 - fix several locking issues related to i_size (5492,5624,5654,5672)
5301 - don't move pending lock onto export if it is already evicted (5683)
5302 - fix kernel oops when creating .foo in unlinked directory (5548)
5303 - fix deadlock in obdfilter statistics vs. object create (5811)
5304 - use time_{before,after} to avoid timer jiffies wrap (5882)
5305 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5306 - Do not leak request if server was not able to process it (5154)
5307 - If mds_open unable to find parent dir, make that negative lookup(5154)
5308 - don't create new directories with extent-mapping (5909, 5936)
5310 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5311 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5312 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5313 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5314 - support for RHEL4 kernel on i686 (5773)
5315 - provide error messages when incompatible logs are encountered (5898)
5317 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5318 * version 1.4.0.10 (1.4.1 release candidate 1)
5320 - don't keep a lock reference when lock is not granted (4238)
5321 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5322 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5323 - drop import inflight refcount on signal_completed_replay error (5255)
5324 - unlock page after async write error during send (3677)
5325 - handle missing objects in filter_preprw_read properly (5265)
5326 - no transno return for symlink open, don't save no-trasno open (3440)
5327 - don't try to complete elan receive that already failed (4012)
5328 - free RPC server reply state on error (5406)
5329 - clean up thread from ptlrpc_start_thread() on error (5160)
5330 - readahead could read extra page into cache that wasn't ejected (5388)
5331 - prevent races in class_attach/setup/cleanup/detach (5260)
5332 - don't dereference de->d_inode after l_dput of de (5458)
5333 - use "int" for stripe value returned from lock_to_stripe (5544)
5334 - mballoc allocation and error-checking fixes in 2.6 (5504)
5335 - block device patches to fix I/O request sizes in 2.6 (5482)
5336 - look up hostnames for IB nals (5602)
5337 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5338 - don't start multiple acceptors for the same port (5277)
5339 - fix incorrect LASSERT in mds_getattr_name (5635)
5340 - export a proc file for general "ping" checking (5628)
5341 - fix "lfs check" to not block when the MDS is down (5628)
5343 - service request history (4965)
5344 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5345 - create an "evict_client" /proc entry on OSTs, like the MDS has
5346 - fix mount usage message, return errors per mount(8) (5168)
5347 - change grep [] to grep "[]" in tests so they work in more UMLs
5348 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5349 - remove ancient LOV_MAGIC_V0 EA support (5047)
5350 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5351 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5352 - allow readahead allocations to fail when low on memory (5383)
5353 - mmap locking landed again, after considerable improvement (2828)
5354 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5356 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5359 - send OST transaction number in read/write reply to free req (4966)
5360 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5361 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5362 - the watchdog thread now runs as interruptible (5246)
5363 - flock/lockf fixes (but it's still disabled, pending 5135)
5364 - don't use EXT3 constants in llite code (5094)
5365 - memory shortage at startup could cause assertion (5176)
5367 - reorganization of lov code
5368 - single portals codebase
5370 - add extents/mballoc support (5025)
5371 - direct I/O reads in the obdfilter (4048)
5372 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5374 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5377 - send OST transaction number in read/write reply to free req (4966)
5378 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5379 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5380 - the watchdog thread now runs as interruptible (5246)
5381 - handle missing objects in filter_preprw_read properly (5265)
5382 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5383 - drop import inflight refcount on signal_completed_replay error (5255)
5384 - unlock page after async write error during send (3677)
5385 - return original error code on reconstructed replies (3761)
5386 - no transno return for symlink open, don't save no-trasno open (3440)
5388 - add pid to ldlm debugging output (4922)
5389 - bump the watchdog timeouts -- we can't handle 30sec yet
5390 - extra debugging for orphan dentry/inode bug (5259)
5392 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5395 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5396 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5397 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5398 - fix NULL dereference in /proc/sys/portals/routes (4827)
5399 - allow failed mdc_close() operations to be interrupted (4561)
5400 - stop precreate on OST before MDS would time out on it (4778)
5401 - don't send partial-page writes before EOF from client (4410)
5402 - discard client grant for sub-page writes on large-page clients (4520)
5403 - don't free dentries not owned by NFS code, check generation (4806)
5404 - fix lsm leak if mds_create_objects() fails (4801)
5405 - limit debug_daemon file size, always print CERROR messages (4789)
5406 - use transno after validating reply (3892)
5407 - process timed out requests if import state changes (3754)
5408 - update mtime on OST during writes, return in glimpse (4829)
5409 - add mkfsoptions to LDAP (4679)
5410 - use ->max_readahead method instead of zapping global ra (5039)
5411 - don't interrupt __l_wait_event() during strace
5413 - add software watchdogs to catch hung threads quickly (4941)
5414 - make lustrefs init script start after nfs is mounted
5415 - fix CWARN/ERROR duplication (4930)
5416 - return async write errors to application if possible (2248)
5417 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5418 - print NAL number in %x format (4645)
5419 - update barely-supported suse-2.4.21-171 series (4842)
5420 - support for sles 9 %post scripts
5421 - support for building 2.6 kernel-source packages
5422 - support for sles km_* packages
5424 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5427 - ignore -ENOENT errors in osc_destroy (3639)
5428 - notify osc create thread that OSC is being cleaned up (4600)
5429 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5430 - reconstruct ost_handle() like mds_handle() (4657)
5431 - create a new thread to do import eviction to avoid deadlock (3969)
5432 - let lconf resolve symlinked-to devices (4629)
5433 - don't unlink "objects" from directory with default EA (4554)
5434 - hold socknal file ref over connect in case target is down (4394)
5435 - allow more than 32000 subdirectories in a single directory (3244)
5436 - fix blocks count for O_DIRECT writes (3751)
5437 - OST returns ENOSPC from object create when no space left (4539)
5438 - don't send truncate RPC if file size isn't changing (4410)
5439 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5440 - bind to privileged port in socknal and tcpnal (3689)
5442 - rate limit CERROR/CWARN console message to avoid overload (4519)
5443 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5444 - basic mmap support (3918)
5445 - kernel patch series update from b1_4 (4711)
5447 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5450 - avoid crash during MDS cleanup with OST shut down (2775)
5451 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5452 - don't use bad inodes on the MDS (3744)
5453 - dynamic object preallocation to improve recovery speed (4236)
5454 - don't hold spinlock over lock dumping or change debug flags (4401)
5455 - don't zero obd_dev when it is force cleaned (3651)
5456 - print grants to console if they go negative (4431)
5457 - "lctl deactivate" will stop automatic recovery attempts (3406)
5458 - look for existing locks in ldlm_handle_enqueue() (3764)
5459 - don't resolve lock handle twice in recovery avoiding race (4401)
5460 - revalidate should check working dir is a directory (4134)
5462 - don't always mark "slow" obdfilter messages as errors (4418)
5464 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5467 - don't close LustreDB during write_conf until it is done (3860)
5468 - fix typo in lconf for_each_profile (3821)
5469 - allow dumping logs from multiple threads at one time (3820)
5470 - don't allow multiple threads in OSC recovery (3812)
5471 - fix debug_size parameters (3864)
5472 - fix mds_postrecov to initialize import for llog ctxt (3121)
5473 - replace config semaphore with spinlock (3306)
5474 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5475 - don't allow enqueue to complete on a destroyed export (3822)
5476 - down write_lock before checking llog header bitmap (3825)
5477 - recover from lock replay timeout (3764)
5478 - up llog sem before sending rpc (3652)
5479 - reduce ns lock hold times when setting kms (3267)
5480 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5481 - fix NULL deref and obd_dev leak on setup error (3312)
5482 - replace some LBUG about llog ops with error handling (3841)
5483 - don't match INVALID dentries from d_lookup and spin (3784)
5484 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5485 - fix invalid assertion in ptlrpc_set_wait (3880)
5487 - add libwrap support for the TCP acceptor (3996)
5488 - add /proc/sys/portals/routes for non-root route listing (3994)
5489 - allow setting MDS UUID in .xml (2580)
5490 - print the stack of a process that LBUGs (4228)
5492 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5495 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5496 - make sure to unset replay flag from failed open requests (3440)
5497 - if default stripe count is 0, use OST count for inode size (3636)
5498 - update parent mtime/ctime on client for create/unlink (2611)
5499 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5500 - free recovery state on server during a forced cleanup (3571)
5501 - unregister_reply for resent reqs (3063)
5502 - loop back devices mounting and status check on 2.6 (3563)
5503 - fix resource-creation race that can provoke i_size == 0 (3513)
5504 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5505 - more debugging for page-accounting assertion (3746)
5506 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5507 - avoid deadlock after precreation failure (3758)
5508 - fix race and lock order deadlock in orphan handling (3450, 3750)
5509 - add validity checks when grabbing inodes from l_ast_data (3599)
5511 - add /proc/.../recovery_status to obdfilter (3428)
5512 - lightweight CDEBUG infrastructure, debug daemon (3668)
5513 - change default OSC RPC parameters to be better on small clusters
5514 - turn off OST read cache for files smaller than 32MB
5515 - install man pages and include them in rpms (3100)
5516 - add new init script for (un)mounting lustre filesystems (2593)
5517 - run chkconfig in %post for init scripts (3701)
5518 - drop scimac NAL (unmaintained)
5520 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5523 - clean kiobufs before and after use (3485)
5524 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5525 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5526 - change init script to fail more clearly if not run as root (1528)
5527 - allow clients to reconnect during replay (1742)
5528 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5529 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5530 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5531 - fix cond_resched (3554)
5532 - only evict unfinished clients after recovery (3515)
5533 - allow bulk resend, prevent data loss (3570)
5534 - dynamic ptlrpc request buffer allocation (2102)
5535 - don't allow unlinking open directory if it isn't empty (2904)
5536 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5537 - remove extraneous obd dereference causing LASSERT failure (3334)
5538 - don't use get_cycles() when creating temp. files on the mds (3156)
5539 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5540 - handle EEXIST for set-stripe, set proper directory name (3336)
5542 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5543 - fix ksocknal_fmb_callback() error messages (2918)
5545 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5548 - don't copy lvb into (possibly NULL) reply on error (2983)
5549 - don't deref dentry after dput, don't free lvb on error (2922)
5550 - use the kms to determine writeback rpc length (2947)
5551 - increment oti_logcookies when osc is inactive (2948)
5552 - update client's i_blocks count via lvb messages (2543)
5553 - handle intent open/close of special files properly (1557)
5554 - mount MDS with errors=remount-ro, like obdfilter (2009)
5555 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5556 - don't use cancelling-locks' kms values (2947)
5557 - use highest lock extent for kms, not last one (2925)
5558 - don't dereference ERR_PTR() dentry in error handling path (3107)
5559 - fix thread race in portals_debug_dumplog() (3122)
5560 - create lprocfs device entries at setup instead of at attach (1519)
5561 - common AST error handler, don't evict client on completion race (3145)
5562 - zero nameidata in detach_mnt in 2.6 (3118)
5563 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5564 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5565 - initialize RPC timeout timer earlier for 2.6 (3219)
5566 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5567 - print nal/nid for unknown nid (3258)
5568 - additional checks for oscc recovery before doing precreate (3284)
5569 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5570 - don't crash in mdc_close for bad permissions on open (3285)
5571 - zero i_rdev for non-device files (3147)
5572 - clear page->private before handing to FS, better assertion (3119)
5573 - tune the read pipeline (3236)
5574 - fix incorrect decref of invalidated dentry (2350)
5575 - provide read-ahead stats and refine rpc in flight stats (3328)
5576 - don't hold journal transaction open across create RPC (3313)
5577 - update atime on MDS at close time (3265)
5578 - close LDAP connection when recovering to avoid server load (3315)
5579 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5580 - don't leak open file on MDS after open resend (3325)
5581 - serialize filter_precreate and filter_destroy_precreated (3329)
5582 - loop device shouldn't call sync_dev() for nul device (3092)
5583 - clear page cache after eviction (2766)
5584 - resynchronize MDS->OST in background (2824)
5585 - refuse to mount the same filesystem twice on same mountpoint (3394)
5586 - allow llmount to create routes for mounting behind routers (3320)
5587 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5588 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5589 - fix rare problem with rename on htree directories (3417)
5591 - allow default OST striping configuration per directory (1414)
5592 - fix compilation for qswnal for 2.6 kernels (3125)
5593 - increase maximum number of MDS request buffers for large systems
5594 - change liblustreapi to be useful for external progs like lfsck (3098)
5595 - increase local configuration timeout for slow disks (3353)
5596 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5598 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5601 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5602 - don't overwrite extent policy data in reply if lock was blocked (2901)
5603 - drop filter export grants atomically with removal from device (2663)
5604 - del obd_self_export from work_list in class_disconnect_exports (2908)
5605 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5606 - swab reply message in mdc_close, other PPC fixes (2464)
5607 - fix destroying of named logs (2325)
5608 - overwrite old logs when running lconf --write_conf (2264)
5609 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5610 - fix race in target_handle_connect (2898)
5611 - mds_reint_create() should take same inode create lock (2926)
5612 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5613 - don't close files for self_export to avoid uninitialized obd (2936)
5614 - allow MDS with the same name as client node (2939)
5615 - hold dentry reference for closed log files for unlink (2325)
5616 - reserve space for all logs during transactions (2059)
5617 - don't evict page beyond end of stripe extent (2925)
5618 - don't oops on a deleted current working directory (2399)
5619 - handle hard links to targets without a parent properly (2517)
5620 - don't dereference NULL lock when racing during eviction (2867)
5621 - don't grow lock extents when lots of conflicting locks (2919)
5623 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5626 - account for cache space usage on clients to avoid data loss (974)
5627 - lfsck support in lustre kernel code (2349)
5628 - reduce journal credits needed for BRW writes (2370)
5629 - orphan handling to avoid losing space on client/server crashes
5630 - ptlrpcd can be blocked, stopping ALL progress (2477)
5631 - use lock value blocks to assist in proper KMS, faster stat (1021)
5632 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5633 - recovery for initial connections (2355)
5634 - fixes for mds_cleanup_orphans (1934)
5635 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5636 - block all file creations until orphan recovery completes (1901)
5637 - client remove rq_connection from request struct (2423)
5638 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5639 - recovery timer race (2670)
5640 - mdc_close recovey bug (2532)
5641 - ptlrpc cleanup bug (2710)
5642 - mds timeout on local locks (2588)
5643 - namespace lock held during RPCs (2431)
5644 - handle interrupted sync write properly (2503)
5645 - don't try to handle a message that hasn't been replied to (2699)
5646 - client assert failure during cleanup after abort recovery (2701)
5647 - leak mdc device after failed mount (2712)
5648 - ptlrpc_check_set allows timedout requests to complete (2714)
5649 - wait for inflight reqs when ptlrpcd finishes (2710)
5650 - make sure unregistered services are removed from the srv_list
5651 - reset bulk XID's when resending them (caught by 1138 test)
5652 - unregister_bulk after timeout
5653 - fix lconf error (2694)
5654 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5655 - readahead locks pages, leaves pending causing memory pressure (2673)
5656 - increase OST request buffers to 4096 on large machines (2729)
5657 - fix up permission of existing directories in simple_mkdir (2661)
5658 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5659 - don't assign transno to errored transactions (2742)
5660 - don't delete objects on OST if given a bogus objid from MDS (2751)
5661 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5662 - if rq_no_resend, then timeout request after recovery (2432)
5663 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5664 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5665 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5666 - MDS recovery shouldn't delete objects that it has given out (2730)
5667 - if enqueue arrives after completion, don't clobber LVB (2819)
5668 - don't unlock pages twice when trigger_group_io returns error (2814)
5669 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5670 - don't write pages to disk if there was an error (1450)
5671 - don't ping imports that have recovery disabled (2676)
5672 - take buffered bytes into account when balancing socknal conn (2817)
5673 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5674 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5675 - remove little-endian swabbing of llog records (1987)
5676 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5677 - retry reposting request buffers if they fail (1191)
5678 - grow extent at grant time to avoid granting a revoked lock (2809)
5679 - lock revoke doesn't evict page if covered by a second lock (2765)
5680 - disable VM readahead to avoid reading outside lock extents (2805)
5682 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5683 - updated kernel patches for hp-2.4.20 kernel (2681)
5685 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5688 - fix truncated write corruption (2366)
5689 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5691 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5692 - ensure len(uuid) < 37 in lmc (1171)
5693 - fix ia64 OOPS in llog_test (2255)
5694 - zero end of page at obdfilter for partial page writes (2648)
5695 - don't leave stale dentries around after renames (bug 2428)
5696 - fix timeouts when evicting a client with a single lock held (2642)
5697 - set deadline for the initial HELLO message to drain (2634)
5698 - print out dotted-quad IP addresses in the socknal (2302)
5700 - additional debugging for MDS client eviction problem (2443)
5701 - fix mkfsoptions support for osts (2603, 2604)
5703 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5706 - add series for the vanilla 2.6.0 kernel
5707 - add series for the vanilla 2.4.24 kernel
5708 - add series for a cray x86/64 UL kernel drop
5709 - fix xattr patches for the vanilla 2.4.19 series
5711 - generate true UUIDs in lmc (1171)
5712 - have portals stack dumping break in UML (2466)
5713 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5714 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5715 - wake callback waiting threads on client eviction (2460)
5716 - Add --ptldebug and --subsystem to lmc (1719)
5717 - update assertion to allow safe interrupt allocation
5718 - set rq_no_resend for cancel requests (2432)
5719 - recalculate ptlrpcd timeout after resend (2494)
5720 - call vfs_rmdir when removing pending directories (2368)
5721 - fix renaming a file to itself (2429)
5722 - lmc creates a default one-stripe lov (2454)
5723 - expand procfs space to handle large clusters (2326)
5724 - increase UML stack to avoid overflow
5725 - update lconf's list of debug and subsystem masks
5726 - fix lfs find --obd (2510)
5727 - /proc tunable for disabling filter read caching (2591)
5728 - stop rpm packages from altering slapd.conf (2301)
5729 - disable nagle in the socknal under 0conf (2578)
5730 - choose mds inode size based on stripe count (2572)
5731 - fix kernel-source rpm problems (2516)
5733 - add --disable-doc to avoid pdf generation (2421)
5734 - update documentation, tests, type-os, comments
5735 - avoid format warnings on ia64
5736 - remove the TOE NAL
5737 - tiny code cleanups by removing unused fields
5739 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5742 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5743 - remove the most blatant lies from BUILDING file (2371)
5744 - change default debug level to reasonable production setting
5745 - reduce client side cache size to reduce cache flush time
5746 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5747 - make TCP zerocopy and pinger support enabled by default (2476)
5748 - sync writes completed after process exits caused crashes (2319)
5749 - maintain correct mount count on the MDS (2356)
5750 - backout 1557, because 2316 wasn't really fixed
5751 - better file I/O statistics gathering in /proc
5752 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5753 - another kernel patch to fix zero-copy TCP function export
5754 - don't take duplicate lock when processing re-sent getattr (2420)
5755 - lctl uses obd_self_export instead of creating new conn (2353)
5756 - MDS/OST recovery case which requires object creation asserted (2425)
5757 - move lfs from /usr/sbin to /usr/bin in packages
5758 - fix race between mds_client_add and mds_client_free (2417)
5759 - use kmalloc instead of slabs in portals (2430)
5760 - don't create duplicate records when a failover MDS is present (2442)
5761 - remove unnecessary mount age check (2332)
5762 - don't remove directory inodes from locks prematurely (2451)
5763 - don't break if MDS service name is the same as hostname (2103)
5764 - fix races in client write RPC generation when cache full (2482)
5766 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5769 - remove now-unused request->rq_obd (278)
5770 - if an allocation fails, print out how much memory we've used (1933)
5771 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5772 - add the "configurable stack size" patch to most series files (1256)
5773 - ability to write large log records, for 100+ OST configs (2306)
5774 - fix NULL deref when filter_prep fails (2314)
5775 - fix operator precedence error in filter_sync
5776 - dynamic allocation of socknal TX descriptors (2315)
5777 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5778 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5779 - ensure that CWARN messages in Portals always get to the syslog
5780 - __init/__exit are not for prototype decls (ldlm_init/exit)
5781 - x86-64 compile warning fixes
5782 - fix gateway LMC keyword conflict (2318)
5783 - fix MDS lock inversions in getattr/reint paths (1844)
5784 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5785 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5786 - DLM race condition prevented some lock evictions (2328)
5787 - ENOMEM detection and retry on socknal sends (2230)
5788 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5789 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5790 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5791 - better error messages when a client is rejected during recovery (1505)
5792 - avoid cancelling locks which were never granted, after failure (2330)
5793 - fix i_sem/journal inversion in mds_client_add (2333)
5794 - fix truncate/getattr lock cycle deadlock (2334)
5795 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5796 - fix two rare exit paths which could leak an l_lock() ref (2321)
5797 - fencepost error in MDS/OST orphan recovery (2226)
5798 - make log record alignment 8 bytes (1988)
5799 - lstripe now fails when requested offset > ost_count (2237)
5800 - ensure that all kernel series have a complete list.h (1607)
5801 - fix crashes in special-file operations (2316)
5802 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5804 - allow configurable automake binary, for testing new versions
5805 - small update to the lfs documentation
5807 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5809 * fix negative export reference count in fsfilt_sync (2312)
5811 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5812 * release candidate 0.9.1
5814 - orphans are moved into the PENDING directory for possible recovery
5815 - replayed opens now open by fid for orphan/rename safety (1042)
5816 - last close of an orphan inode generates a transno (683)
5817 - chdir() and mount() now pin the directory entry (1020)
5818 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5819 - discard very old requests without processing them (1502)
5820 - remove some common, well-understood CERRORs (1505)
5821 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5822 - clear "grant" flags in OST replies until OST grant code lands (1644)
5823 - fix read performance by not clobbering i_blksize on client (1598)
5824 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5825 - make LRU size a /proc tunable, clears locks when reduced (707)
5826 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5827 - server locks take references on exports now (1558)
5828 - build fixes for 2.4.20-rh trees (1663)
5829 - return an error from lov_create if all OSCs are inactive (1751)
5830 - fix import levels when a reconnect happens without a timeout (1597)
5831 - exit early from mds_open if we get a lookup error (1749)
5832 - partial page read at EOF wouldn't wait for disk before sending (1642)
5833 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5834 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5835 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5836 - write/truncate lock inversion (1639)
5837 - Don't auto-load obdclass, portals modules during cleanup (1495)
5838 - fix timestamps from jumping to "now" (1763)
5839 - extra journal assertions (1648)
5840 - add an extra multiunlink test (1771)
5841 - fix read_record/write_record API (1776)
5842 - fix leak of offset_extent, possible incorrect i_size later (1772)
5843 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5844 - Debugging for the kqswnal_get_idle_tx problems (1820)
5845 - Allow recovery to be attempted multiple times (1536)
5846 - Write out MDS last_rcvd file after it is first created (1600)
5847 - Fix tx_descriptor leak in failed transmit situations (1827)
5848 - ext3 journaling fixes for assertion failure after IO error (1871)
5849 - class_export_put() on freed export after completion AST error (1896)
5850 - Fix revalidate looping in VFS (1322)
5851 - Don't access a freed export during MDS_REINT timeout (1521)
5852 - Add open-unlink recovery support on the MDS (1673,1764)
5853 - Return an error if no MDS data was read from last_rcvd (1946)
5854 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5855 - Don't setup a disk device that doesn't match exported UUID (317)
5856 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5857 - avoid committing NULL handle in force close
5858 - local.sh is now a one-stripe LOV configuration
5859 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5860 - fix ext3 htree duplicate directory entry corruption (1516)
5861 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5862 - update to new LOV EA format (2097)
5863 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5865 2003-06-15 Phil Schwan <phil@clusterfs.com>
5868 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5869 - per-import/export recovery handling (958, 931, 959)
5870 - multiple last-rcvd slots, for serving multiple FSes (949)
5871 - connections are again shared between multiple imp/exports (963, 964)
5872 - "umount -f" would hang if any requests needed to be sent (393, 978)
5873 - avoid pinning large req buffer by copying for queued messages (989)
5874 - add "uuid" to "lctl device" command to help upcalls (991)
5875 - "open" RPCs with transnos would confuse recovery counters (1037)
5876 - do proper endian conversion of all wire messages (288, 340, 891)
5877 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5878 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5879 - fix ptlrpc_connection leak in target_handle_connect (1174)
5880 - fix import refcounting bug in OST and MDS cleanup (1134)
5881 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5882 - fix very unlikely obd_types race condition (501)
5883 - remove osc_open hack for echo_client (1187)
5884 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5885 - a failure in read_inode2 leads to deadlock (1139)
5886 - cancel ack-locks as soon as transaction is committed (1072)
5887 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5888 - make sure to commitrw after any preprw to avoid deadlock (1162)
5889 - failing to execute a file in a lustre FS would lock inode (1203)
5890 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5891 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5892 - obd_uuid2tgt would walk off the end of the list (1255)
5893 - on IA64 the osc would give portals incorrect bulk size (1258)
5894 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5895 - fix lock inversion caused by new llite matching code (1282)
5896 - limit the number of dirty pages on a client to 10MB (1286)
5897 - timed out locks were not being corrected cancelled (1289)
5898 - fix O_DIRECT above 4GB on IA-32 (1292)
5899 * major user-visible changes
5900 - fail out/fail over policy now controlled by the upcall (993)
5902 - add OBD_PING to check server availability and failure (954)
5903 - lustre messages are now sent in sending host order (288, 340, 891)
5904 - add eadatalen to MDS getattr reply (340)
5905 - OST read replies may contain second buffer, with per-page status (593)
5907 2003-03-11 Phil Schwan <phil@clusterfs.com>
5910 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5911 - failed open()s could cause deadlock; fixed (867, 869)
5912 - stop cancelling OST locks when files are closed (481)
5913 - overlapping XID spaces caused network corruption (851, 853)
5914 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5915 - setattr_raw added, to do single-RPC, server-side setattrs
5916 - lmc/lconf syntax change for OST UUIDs
5917 - fix crashy race condition between ptlrpc_free_req and osc_close
5918 - don't use request in mdc_enqueue if we hit a timeout (889)
5919 - don't set the inode i_size for regular files from the MDS (896)
5920 - handle out of order completion AST (842)
5921 - don't LBUG if a lock request times out after receiving AST (913)
5922 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5923 - if a bad lock AST arrives, send an error instead of dropping entirely
5924 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5925 - fix leak in bulk IO when only partially completed (899, 900, 926)
5926 - fix O_DIRECT for ia64 (55)
5927 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5928 - C-z after timeout could hang a process forever; fixed (977)
5930 - client-side I/O cache (678, 924, 929, 941, 970)
5932 - READPAGE and SETATTRs which don't take server-side locks get
5935 2003-02-11 Phil Schwan <phil@clusterfs.com>
5938 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5939 - Fix fsfilt_extN_readpage() to read a full page of directory
5940 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5941 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5942 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5943 - Most intent operations were removed, in favour of a new RPC mode
5944 that does a single RPC to the server and bypasses most of the VFS
5945 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5946 - Aggressively cancel local locks on DLM servers
5947 - mds_reint_unlink sends EA to the client if it's the last nlink.
5948 client uses that EA to unlink OST objects.
5949 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5950 - recursive symlinks were fixed (439)
5951 - fixed NULL deref in DEBUG_REQ
5952 - filter_update_lastobjid no longer calls sync, which annoyed extN
5953 - fixed multi-client small-writes to a single file problem (445)
5954 - fixed mtime updates during file writes (607)
5955 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5956 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5957 - fixed Linux OST size reporting problem (444, 656)
5958 - OST now updates object mtime with writes or setattr (607, 619)
5959 - client verifies file size before zeroing page past EOF (445)
5960 - OST now writes last allocated objid to disk with allocation (108)
5961 - LOV on echo now works (409)
5963 - mds_reint_unlink sends a new buffer, with the EA included. this
5964 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5965 set if a regular file was being unlinked, and it was the last link
5966 - use PtlGet from the target for bulk writes (315)
5967 - OST now updates object mtime with writes or setattr (607, 619)
5968 - LDLM now has a grant-time callback to revalidate locked items, if
5970 - Many MDS operations were reorganized to combat race conditions
5972 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5973 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5975 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5978 - Fully reactivate OST imports after reconnection (512, others)
5979 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5980 - More graceful error handling for truncating on dead OST (515)
5981 - Don't error out unless we're actually accessing dead stripes (474)
5982 - Fix garbage sizes when stripes are missing (410)
5983 - LRU counters were broken, causing constant lock purge (433, 432)
5984 - garbage on read from stripes with failed OSTs (441)
5985 - mark OSCs as active before reconnecting during recovery (438)
5986 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5987 - lfind did not preserve OST order in output (443)
5988 - symlinks cause hung clients, incorrect data (439)
5989 - stop dereferencing request after dropping refcount (457)
5990 - don't LASSERT(spin_is_locked) on non-SMP (455)
5991 - fixes for many rename() bugs
5992 - fstat didn't correctly synchronize attributes (399)
5993 - server must handle lock cancellation during blocking AST prep (487)
5994 - bulk descriptors were free()d too soon (511)
5995 - fix paths in lconf, which would load incorrect modules (451, 507)
5996 - fix confusing lconf 'host not found' error message (386)
5997 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5998 - fix race condition in mdc_blocking_ast() for inode access (526)
5999 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6000 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6001 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6002 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6003 - fix lstripe in several ways (kernel side) (527)
6004 - fix request leak in ldlm_cli_enqueue (262)
6005 - incorrect OSC was marked inactive after OST failure
6006 - call mds_fs_cleanup before unmounting filesystem (524)
6007 - fix races between taking ns_lock and ldlm_lock_change_resource
6008 - fix races updating LOV export open file list
6009 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6010 - fix recovery NULL deref in ldlm_cli_cancel_unused
6011 - fix some DLM races by using new hash table for lock handles (419)
6012 - permit the client to specify desired inodes, at replay
6013 - duplicate requests when we queue them for replay reintegration
6014 - fix last_rcvd offset calculation
6015 - sync after each recovered transaction, so we always make progress
6016 - never, not always, ERESTART requests without transnos
6017 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6019 - skip replay if the MDS says that the client is already connected
6020 - don't check for a recovery-enabled export to match lctl's UUID
6021 - don't INC_USE_COUNT for phantom exports
6022 - don't crash when cleaning up phantom exports (567)
6023 - don't double-finish or set replay data for errored mdc_open requests
6024 - abort requests when they time out, so we don't get old replies
6025 - send/receive replies for AST messages again
6026 - if the client says that it doesn't have the lock, cancel it on the
6028 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6029 mark it as destroyed, it will all work out when decref is called
6030 - fix module use counts (22, 581)
6032 - ASTs now expect a reply (server cancels lock on error reply)
6034 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6037 - fix many simultaneous client startup (392)
6038 - fix dentry->d_it clobbering
6039 - credentials weren't being shipped for readdir/getattr operations
6040 - remove invalid assertions triggered during some concurrent MD
6042 - proper Lustre versions added (336, 389)
6043 - fix memory leak for create error case (398)
6044 - fix LOV locking bug that would get cli/srv out of sync
6045 - fix echo client over LOV (409)
6046 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6047 - fix double-O_EXCL intent crash (424)
6048 - avoid sending multiple lock CANCELs (352)
6050 - MDS can do multi-client recovery (modulo bugs in new code)
6052 - many updates, edits, cleanups
6054 2002-11-18 Phil Schwan <phil@clusterfs.com>
6057 - fix null d_it dereference (346)
6058 - fix full OST/dbench hang (333)
6059 - fix permission problem with file removal (286)
6060 - fix removal of OSCs from LOV when they fail
6061 - fix NULL deref during bulk timeout (214)
6062 - fix problems related to multiple filesystems on one MDS (241)
6063 - fixed serious subtle metadata locking bugs
6064 - free locks on clients when inodes are removed due to memory
6066 - fix inode pointer in lock data (285)
6067 - partial support for multiple MDS on a single host (241)
6068 - data locks weren't cancelled at clear_inode time (290, 311)
6069 - intent locks could lead to unbounded lock growth (205)
6070 - added a maximum lock count, an LRU list, and a flusher
6071 - fix multiple rename (365)
6072 - properly abstracted the echo client
6073 - OSC locked 1 byte too many; fixed
6074 - rewrote brw callback code:
6075 - fixed recovery bugs related to LOVs (306)
6076 - fixed too-many-pages-in-one-write crash (191)
6077 - fixed (again) crash in sync_io_timeout (214)
6078 - probably fixed callback-related race (385)
6080 - Add capability to MDS protocol
6081 - LDLM cancellations and callbacks on different portals
6083 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6086 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6087 - defer open object destruction to close time (601981, 138)
6088 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6089 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6090 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6091 - don't call obd_setattr in ll_file_release for destroyed objects
6092 * protocol change to lustre_msg: move |version| and add |flags|
6093 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6094 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6095 * added replay of create, unlink, link and rename operations during
6096 MDS failover; recovery should be much more robust now
6097 * remove failed OSCs from LOVs (only lov_create uses this so far)
6098 * the lustre-HOWTO was brought (more) up to date (582544)
6100 2002-10-23 Phil Schwan <phil@clusterfs.com>
6103 - in-use dentries weren't being reused properly (617851)
6104 - prevent multiple LDLM setup (599178)
6105 - fix LOV size calculations for truncate (617853)
6106 - fix client handling of MDS intent errors (POSIX)
6107 - fix permission bug in lovstripe.c test (624321)
6108 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6109 - truncate past end of file could corrupt data
6110 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6111 - a race in recovery could return ETIMEDOUT to apps (623947)
6112 - building outside the source directory was fixed
6113 * the lustre-HOWTO was brought (more) up to date (582544)
6114 * major progress was made on recovery functionality
6116 2002-10-10 Phil Schwan <phil@clusterfs.com>
6119 - recovery deadlock fix
6120 - rm -rf causes LBUG fix (617817)
6121 - file open by multiple tasks fix (618962)
6122 - directory permissions bugs (602707 and 620007)
6123 - journal_stop fixed with locking (611313)
6124 - O_APPEND failures resolved (618273, perhaps 614459)
6125 - lconf PATH fix (619770)
6126 - IA64 build fix (621450)
6127 - RPC buffer sizes scale with amount of memory
6129 2002-10-01 Phil Schwan <phil@clusterfs.com>
6132 - locks would be cancelled without throwing away data pages,
6133 resulting in inconsistent data (605627)
6134 - inode attributes were not always being refreshed (605627, 612449)
6135 - lconf now continues to cleanup after lctl reports an error
6136 - MDS now enforces user permissions (602707)
6137 - lprocfs cleanup fixed, but not yet enabled (614157)
6138 - fixed infinite server hang, should a client not respond to an AST
6139 - avoid going into recovery if user calls readlink() with a buffer
6140 that's too small (613941)
6141 - AST RPCs no longer require replies (614867) -- this may be changed
6142 - don't crash server if client sends an IOV that's too big (611336)
6143 - fixed lock conversion deadlock (611892)
6144 - fixed the following of symlinks (614622)
6145 * recovery: the server can remove locks from a client that dies, other
6146 clients can make progress
6147 * more extN patch fixes
6148 * compile-time configurable ptlrpc buffer allocations
6150 - collaborative read cache document
6151 - Lustre Lite Performance CDR document-in-progress
6153 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6156 - fix typo in patch-2.4.18
6158 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6161 - clear ptlrpc request each time in handle_incoming_request()
6162 - unlink of files now destroys the object on the OST
6164 2002-09-19 Peter Braam <braam@clusterfs.com>
6166 * add hard link support
6167 * change obdfile creation method
6168 * kernel patch changed
6170 2002-09-19 Peter Braam <braam@clusterfs.com>
6173 - stack overflow bug in extN fixed
6175 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6177 * documentation updates
6178 - add man pages for config tools
6179 - update tests/README to describe testing with new config tools
6180 - finish metadata API descriptions
6181 * bug fixes and cleanups
6182 - statfs workaround for 16TB limit
6183 - LOV stripe allocation improved, can stripe on subset of OSTs
6184 - LOV file size/IO offset was wrong for files > 4GB in size
6185 - object EA data was being dropped, caused files to be unreadable
6186 - memory overflow with non-LOV OST caused memory corruption
6187 - fixed regression tests to work with new config tools, obdfilter
6188 - fixed bug when directory size became larger than 1 block
6189 - fixed bug (for single client case) when PWD was deleted
6190 - invalidate local directory pages when doing intent-based ops
6191 - avoid LDLM oops when lock callback contained bad data
6193 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6195 * documentation updates
6196 * bug fixes and cleanups
6197 - configuration tools
6200 - 64-bit compile warnings
6201 - 64-bit internal statfs data
6203 * test_brw on persistent OST devices
6205 * lprocfs (disabled)
6207 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6209 * documentation updates
6210 * bug fixes and cleanups
6211 * configuration tools
6213 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6216 * many small fixes to 0_5_4
6217 * io/network handling
6218 * thinkos in MDS operations
6220 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6223 * crucial basic fixes to 0.5.3
6224 * IOR, Iozone work over Elan
6227 2002-08-07 Phil Schwan <phil@clusterfs.com>
6228 * version 0_5_3, our first alpha
6229 * we use the new Portals iovs
6230 * documentation updates
6231 * bug fixes and cleanups
6232 * small changes in the DLM wire protocol
6234 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6235 * version 0_5_1 with some initial stability,
6236 * locking on MD and file I/O.
6237 * documentation updates
6238 * several bug fixes since 0.5.0
6239 * small changes in wire protocol
6241 2002-07-18 Phil Schwan <phil@clusterfs.com>
6243 * delivered as Lustre Light Alpha
6244 * fixed a crash after handling invalid MDS requests
6245 * fixed directory pages for architectures with non-4k pages sizes
6247 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6248 * release version v0_4_4
6249 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6251 * Fixes a number of interruption problems with OST operations.
6252 * Update documentation for portals header changes
6253 * Move all wire protocol structs/defines to lustre_idl.h
6254 * Fixes symlink length bug.
6255 * Add tcpdump to repository.
6257 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6258 * release version v0_4_3
6259 * Fixes statfs for inodes on extN.
6260 * Fixes bug in runtests which would delete /etc/hosts.
6261 * Use 64-bit object IDs wherever possible (not into VFS though)
6262 Remove ost_get_info, which is unused by lustre, and out of date.
6264 2002-07-03 Peter Braam <braam@clusterfs.com>
6265 * release version v0_4_2 Fixes a lookup error (type not passed)
6266 * move forward to head of Portals
6267 * move forward to latest Lustre kernel
6269 2002-06-25 Peter Braam <braam@clusterfs.com>
6270 * release version v0_4_1. Hopefully stable on single node use.