1 tbd Sun Microsystems, Inc.
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
17 Description: Update to OFED 1.5.1
18 Details : OFED 1.5.1 will be used in environments that don't use
23 Description: Remove lnet. prefix from lctl params display
24 Details : Also renamed lnet.memused to lnet_memused to
29 Description: Use conf_param -d to remove permanant settings.
33 Description: Add list_param -R
34 Details : lctl list_param now supports recursive mode.
37 Bugzilla : 21829, 22312
38 Description: Fix llobdstat.
39 Details : 1. Now updating the write_bytes counter
40 2. Port script fixes from 1.8.
44 Description: decrease the usage of memory on clients.
45 Details : 1. On clients, recycle dentries and inodes unused.
46 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
51 Description: llobdstat fix and enhancement
52 Details : add a counter to set a limit to how many samples will be returned
53 fix a wildcard in the path to limit to obdfilter stats only
57 Description: lfs check only allowed for root
58 Details : Code cleanups and added a regresion test of non-root use of
63 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
64 Details : Fix hr->hr_index overload.
68 Description: Fix FIEMAP functionality.
72 Description: FC11 patchless client support added
76 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
77 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
81 Description: New acc-sm script and various test-framework enhancments.
85 Description: Previously landed patch was reverted.
89 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
93 Description: Add ldiskfs maxdirsize mount option.
97 Description: Fix mkfs.lustre support for 16TB LUNs.
101 Description: Print mmp_check_interval
102 Details : print mmp_check_interval and make it possible to abort
103 mount operation in case it takes too long.
107 Description: Add cascading_rw to lustre mpi tests.
111 Description: Store SOM data in EA.
112 Details : Also several SOM related cleanups.
116 Description: Add refcount to md_open_data
120 Description: Use refcount on lov_request_set to prevent memory corruption.
124 Description: Statahead: use parent dir's i_mutex for synchronization.
128 Description: loadgen improvements.
132 Description: Enable peer health detection in ksocklnd and ko2iblnd.
136 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
140 Description: Fix race on obd_devcie:md_stats
144 Description: Fix SLES 11 client builds.
145 Details : Resolved uninitalized pointer gcc warnings.
149 Description: Allow non-root access for "lfs check".
150 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
154 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
156 Severity : enhancement
158 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
159 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
163 Description: Sanity test_27p fails.
164 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
165 Sanity tests 27 were cleaned up and improved.
169 Description: Unify req format on client/servers
170 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
171 buffers and swabbers approach..
176 Description: drop join file stuff.
178 Severity : enhancement
180 Description: Add 2.6.27 and SLES11 (patchless client) support
184 Description: Fix in ptlrpc_expire_one_request() to print the signed time
187 Severity : enhancement
189 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
192 Severity : enhancement
194 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
196 Severity : enhancement
198 Description: correctly handle big reply message.
199 Details : send LNet event if reply is bigger then buffer and adjust this buffer
204 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
206 Severity : enhancement
208 Description: Add support for OEL5.
212 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
216 Description: client could not reconnect to OST because of an active request.
217 Details : The client now retries to reconnect to the same server, if a
218 connect request failed with EBUSY or -EAGAIN. Test case 35b
219 conf-sanity.sh has been created.
223 Description: Don't allow make backward step on assignin osc next id.
224 Details : race between allocation next id and ll_sync thread can be cause
225 of set wrong osc next id and can be kill valid ost objects.
229 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
233 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
237 Description: Error handling in osc_statfs_interpret() has been improved.
238 Details : Check in osc_statfs_interpret() for EBADR.
242 Description: Do not send statfs() requests to OSTs disabled by administrator.
243 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
245 Severity : enhancement
247 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
249 Severity : enhancement
251 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
253 Severity : enhancement
255 Description: Enhancement for lfs(1) command to use numeric uid/gid.
259 Description: Improved cleanup procedure for conf-sanity test_46a.
263 Description: Do not start bulk transfer, if client tried to reconnect.
264 Details : Do not start bulk transfer, if client tried to reconnect,
265 but failed due to active rpcs. Check for exp_abort_active_req
266 has been implemented.
270 Description: Prevent inconsistences between linux and lustre mount structures.
271 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
272 Make the sleep interruptible.
274 Severity : enhancement
276 Description: provide server to client comms path
277 Details : server to client communications path via new LDLM_SET_INFO rpc
281 Description: truncate starts GFP_FS allocation under transaction causing deadlock
282 Details : ldiskfs_truncate calls grab_cache_page which may start page
283 allocation under an open transaction. This may lead to
284 calling prune_icache with consequent lustre reentrance.
288 Description: send correctly lsm on open replay
289 Details : MDS is trust to LSM size on replay open, but client can set wrong size
294 Description: enable client interface failover
295 Details : When a child reconnects from another NID, properly update export
296 nid hash position and ldlm reverse import.
298 Severity : enhancement
300 Description: Communicate OST degraded/readonly state via statfs to MDS
301 Details : Flags in the statfs returned from OSTs indicate whether the
302 OST is in a degraded RAID state, or if the filesystem has
303 turned read-only after a filesystem error is detected.
308 Description: don't panic if EPROTO was hit when reading symlink
309 Details : correctly handling request reference in error cases.
311 Severity : enhancement
313 Description: Add LustreNetLink, a kernel-userspace communcation path.
315 Severity : enhancement
317 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
322 Description: access to llog context before init.
323 Details : move handling CATALOGS file at osc layer and forbid access to llog
327 Frequency : rare, only when client is writing to OST on local node
329 Description: Avoid deadlock for local client writes
330 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
331 memory freeing context. This allows OST threads to set the
332 PF_MEMALLOC flag on task structures in order to allocate memory
333 from reserved pools and complete IO.
334 Use GFP_HIGHUSER for OST allocations for non-local client writes,
335 so that the OST threads generate memory pressure and allow
336 inactive pages to be reclaimed.
338 Severity : enhancement
340 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
345 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
346 Details : move ldlm namespace creation in setup phase to avoid grab
347 _lprocfs_lock with cli_sem held.
351 Description: Temporarily disable grant shrink.
352 Details : Disable the feature for debugging.
355 Frequency : only when formatting test filesystems
357 Description: Unable to run several mkfs.lustre on loop devices at the same time
358 Details : mkfs.lustre returns error 256 on the concurrent loop devices
359 formatting. The solution is to proper handle the error.
361 Severity : enhancement
363 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
365 Severity : enhancement
367 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
372 Description: resolve race between obd_disconnect and class_disconnect_exports
373 Details : if obd_disconnect will be called to already disconnected export he
374 forget release one reference and osc module can't unloaded.
376 Severity : enhancement
378 Description: Allow tuning service thread via /proc
379 Details : For each service a new
380 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
381 created that can be used to set min/max thread counts, and get the
382 current number of running threads.
387 Descriptoin: don't return error if have partially created objects for file.
388 Details : lov_update_create_set uses set->set_success as index for created
389 objects, so if some requests will be failed, they will have hole
390 at end of array and qos_shrink_lsm for allocate correct lsm.
392 Severity : enhancement
394 Description: Update OFED support to 1.4.1
399 Description: Reduce small size read RPC
400 Details : Set read-ahead limit for every file and only do read-ahead when
401 available read-ahead pages are bigger than 1M to avoid small size
404 Severity : enhancement
406 Description: move AT tunable parameters for more consistent usage
407 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
409 Severity : enhancement
411 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
412 Details : allow skip disconnected ost for send statfs request and hide error
418 Description: fix racy locking of mballoc block bitmaps causing BUG
419 Details : The locking of the mballoc buddy bitmap and the in-memory
420 block bitmap was using two different spin locks in some
421 cases. This made it possible to incorrectly access the
422 mballoc bitmap while another process was modifying it,
423 causing a sanity assertion to fail. While no on-disk corruption
424 was reported, there was some risk of this happening.
427 Frequency : rare, on llog test 6
429 Descriptoin: don't allow connect to already connected import
430 Details : allowing connect to already connected import is hide connecting problem.
435 Descriptoin: allow kill process which wait statahead result
436 Details : for some reasons 'ls' can stick in waiting result from statahead,
437 in this case need way for kill this process.
439 Severity : enhancement
441 Description: Add state history info file, enhance import info file
442 Details : Track import connection state changes in a new osc/mdc proc file;
443 add overview-type data to the osc/mdc import proc file.
445 Severity : enhancement
447 Description: MDS create should not wait for statfs RPC while holding DLM lock.
449 Severity : enhancement
451 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
456 Descriptoin: OST grant too much space to client even there are not enough space.
457 Details : Client will shrink its grant cache to OST if there are no write
458 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
459 this grant cache if there are already not enough avaible space
460 (left_space < total_clients * 32M).
463 Frequency : start MDS on uncleanly shutdowned MDS device
465 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
466 Details : stay in waiting mds<>ost recovery finished produce random bugs
467 due race between two ll_sync thread for one lov target. send
468 ACTIVATE event only if connect realy finished and import have
472 Frequency : rare, connect and disconnect target at same time
474 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
475 Details : don't call obd_disconnect under lov_lock. this long time
476 operation and can block ptlrpcd which answer to connect request.
481 Descriptoin: don't lose wakeup for imp_recovery_waitq
482 Details : recover_import_no_retry or invalidate_import and import_close can
483 both sleep on imp_recovery_waitq, but we was send only one wakeup
487 Frequency : always with long access acl
489 Descriptoin: mds can't pack reply with long acl.
490 Details : mds don't control size of acl but they limited by reint/getattr
493 Severity : enhancement
495 Description: Update to SLES10 kernel-2.6.16.60-0.33.
497 Severity : enhancement
499 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
502 Frequency : start MDS on uncleanly shutdowned MDS device
504 Descriptoin: aborting recovery hang on MDS
505 Details : don't throttle destroy RPCs for the MDT.
508 Frequency : on remount
510 Description: external journal device not working after the remount
511 Details : clear dev_rdonly flag for external journal devices in
517 Description: shutdown vs evict race
518 Details : client_disconnect_export vs connect request race.
519 if client will evicted at this time - we start invalidate
520 thread without referece to import and import can be freed
524 Frequency : rare, need acl's on inode.
526 Description: client can't handle ost additional correctly
527 Details : if ost was added after client connected to mds client can have
528 hit lnet_try_match_md ... to big messages to wide striped files.
529 in this case need teach client to handle config events about add
530 lov target and update client max ea size at that event.
532 Severity : enhancement
534 Description: Changelogs
535 Details : Changelogs are a lightweight mechanism to track filesystem
536 metadata and namespace changes. The changelog is recorded
537 permanently on the MDTs, and is periodically "consumed" / purged
538 when records are no longer needed.
540 Severity : enhancement
542 Description: compact fld format with extents
543 Details : Store range of seq rather than every seq in FLD. Seq
544 controller update FLD rather than clients. In Case of CMD, mdt0
545 has FLD, all other metadata server act as non persistent proxy
546 for FLD queries and cache fld entries in fld cache.
551 Description: don't skip ost target if they assigned to file
552 Details : Drop slow OSCs if we can, but not for requested start idx.
553 This means "if OSC is slow and it is not the requested
554 start OST, then it can be skipped, otherwise skip it only
555 if it is inactive/recovering/out-of-space.
559 Description: more cleanup in mds_lov
560 Details : not send LOV EA under replay, we can't know about they size at this
561 time. Don't allow client connect to mds before any ost connected,
562 for avoid problems with LOV EA size and returning EIO to client.
564 Severity : enhancement
566 Description: Interoperability at server side (Disk interoperability)
568 Severity : enhancement
570 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
572 Severity : enhancement
574 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
576 Severity : enhancement
578 Description: New client IO stack (CLIO).
580 Severity : enhancement
582 Description: Commit on sharing. Eliminate inter-client dependencies between
583 uncommitted transactions by doing transaction commits.
584 Thereby clients may recovery independently.
587 Frequency : Create a symlink file with a very long name
589 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
590 Details : If there is no extra space in the request for early cancels,
591 ldlm_req_handles_avail() returns 0 instead of a negative value.
593 Severity : enhancement
595 Description: Add /proc entry for import status
596 Details : The mdc, osc, and mgc import directories now have
597 an import directory that contains useful import data for debugging
600 Severity : enhancement
602 Description: Re-disable certain /proc logging
603 Details : Enable and disable client's offset_stats, extents_stats and
604 extents_stats_per_process stats logging on the fly.
607 Frequency : Only on FC kernels 2.6.22+
609 Description: oops in statahead
610 Details : Do not drop reference count for the dentry from VFS when lookup,
611 VFS will do that by itself.
613 Severity : enhancement
615 Description: Generic /proc file permissions
616 Details : Set /Proc file permissions in a more generic way to enable non-
617 root users operate on some /proc files.
621 Description: Hitting mdc_commit_close() ASSERTION
622 Details : Properly handle request reference release in
623 ll_release_openhandle().
627 Description: quota recovery deadlock during mds failover
628 Details : This patch includes att18982, att18236, att18237 in bz14840.
630 1. fix osts hang when mds does failover with quotaon
631 2. prevent watchdog storm when osts threads wait for the
636 Frequency : only patchless client
637 Description: add workaround for race between add/remove dentry from hash
639 Severity : enhancement
641 Description: Allow OST glimpses to return PW locks
645 Description: LBUG when llog conf file is full
646 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
650 Description: Prevent import from entering FULL state when server in recovery
654 Description: service mount cannot take device name with ":"
655 Details : Only when device name contains ":/" will mount treat it as
661 Description: replace ptlrpcd with the statahead thread to interpret the async
662 statahead RPC callback
666 Frequency : on recovery
667 Description: I/O failures after umount during fail back
668 Details : if client reconnected to restarted server we need join to recovery
669 instead of find server handler is changed and process self
670 eviction with cancel all locks.
672 Severity : enhancement
674 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
676 Severity : enhancement
678 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
680 Severity : enhancement
682 Description: Upcall on Lustre log has been dumped
683 Details : Allow for a user mode script to be called once a Lustre log has
684 been dumped. It passes the filename of the dumped log to the
685 script, the location of the script can be specified via
686 /proc/sys/lnet/debug_log_upcall.
691 Description: avoid idr_remove called for id which is not allocated.
692 Details : Move assigment s_dev for clustered nfs to end of initialization,
693 to avoid problem with error handling.
698 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
699 Details : When connection is reused this not moved from CONN_UNUSED_HASH
700 into CONN_USED_HASH and this prodice warning when put connection
701 again in unused hash.
704 Severity : enhancement
706 Description: File striping can now be set to use an arbitrary pool of OSTs.
708 Severity : enhancement
710 Description: Export bytes_read/bytes_write count on OSC/OST.
714 Description: Early reply size mismatch, MGC loses connection
715 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
716 the connect flags are properly negotiated.
721 Description: quotas are not honored with O_DIRECT
722 Details : all writes with the flag O_DIRECT will use grants which leads to
723 this problem. Now using OBD_BRW_SYNC to guard this.
727 Description: add quota statistics
728 Details : 1. sort out quota proc entries and proc code.
729 2. add quota statistics
731 Severity : enhancement
733 Description: enable quota support for HEAD.
737 Description: Properly propagate oinfo flags from lov to osc for statfs
738 Details : restore missing copy oi_flags to lov requests.
740 Severity : enhancement
742 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
743 : l_getgroups(8), lst(8), routerstat(8)
744 Details : included man pages for llobdstat(8), llstat(8),
745 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
747 Severity : enhancement
749 Description: configure's --enable-quota should check the
750 : kernel .config for CONFIG_QUOTA
751 Details : configure is terminated if --enable-quota is passed but
752 : no quota support is in kernel
756 Description: Remove portals compatibility
757 Details : Remove portals compatibility, not interoperable with releases
762 Description: Resolve device initialization race
763 Details : Prevent proc handler from accessing devices added to the
764 obd_devs array but yet be intialized.
766 Severity : enhancement
768 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
770 Severity : enhancement
772 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
777 Description: Using wrong pointer in osc_brw_prep_request
778 Details : Access to array[-1] can produce panic if kernel compiled with
779 CONFIG_PAGE_ALLOC enabled
783 Description: Client runs out of low memory
784 Details : Consider only lowmem when counting initial number of llap pages
788 Description: *optional* service tags registration
789 Details : if the "service tags" package is installed on a Lustre node
790 When the filesystem is mounted, a local-node service tag will
791 be created. See http://inventory.sun.com/ for more information
792 about the Service Tags asset management system.
796 Description: Kernel BUG tries to release flock
797 Details : Lustre does not destroy flock lock before last reference goes
798 away. So always drop flock locks when client is evicted and
799 perform unlock regardless of successfulness of speaking to MDS.
803 Description: add refcount protection for osc callbacks, avoid panic on shutdown
807 Description: sanity test 65a fails if stripecount of -1 is set
808 Details : handle -1 striping on filesystem in ll_dirstripe_verify
813 Description: ASSERTION(CheckWriteback(page,cmd)) failed
814 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
817 Severity : enhancement
819 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
823 Description: do not process already freed flock
824 Details : flock can possibly be freed by another thread before it reaches
825 to ldlm_flock_completion_ast.
829 Description: LBUG during stress test
830 Details : Need properly lock accesses the flock deadlock detection list.
834 Description: oops in page fault handler
835 Details : kernel page fault handler can return two special 'pages' in error
836 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
840 Description: timeout with invalidate import.
841 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
842 should be handled by ptlrpcd. This produce long age waiting and
843 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
845 Severity : enhancement
847 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
851 Description: enable MGS and MDT services start separately
852 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
853 a co-located MGS without starting the MGS, which is a complement
854 to 'nosvc' mount option.
858 Frequency : after recovery
859 Description: precreate to many object's after del orphan.
860 Details : del orphan st in oscc last_id == next_id and this triger growing
861 count of precreated objects. Set flag LOW to skip increase count
862 of precreated objects.
866 Frequency : rare, on clear nid stats
867 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
868 Details : when clean nid stats sometimes try destroy live entry,
869 and this produce panic in free.
873 Description: Stack overflow during MDS log replay
874 ease stack pressure by using a thread dealing llog_process.
878 Description: wait until IO finished before start new when do lock cancel.
879 Details : VM protocol want old IO finished before start new, in this case
880 need wait until PG_writeback is cleared until check dirty flag and
881 call writepages in lock cancel callback.
883 Severity : enhancement
885 Description: using special macro for print time and cleanup in includes.
889 Description: mds_mfd_close() ASSERTION(rc == 0)
890 Details : In mds_mfd_close(), we need protect inode's writecount change
891 within its orphan write semaphore to prevent possible races.
895 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
899 Frequency : rare, on shutdown ost
900 Description: don't hit live lock with umount ost.
901 Details : shrink_dcache_parent can be in long loop with destroy dentries,
902 use shrink_dcache_sb instead.
906 Description: don't panic with use echo client
907 Details : echo client pass NULL as client nid pointer and this produce null
912 Description: fix build on ppc32
913 Details : compile code with -m64 flag produce wrong object file for ppc32.
917 Description: add message levels for liblustreapi
921 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
922 Details : Move the 'good_osts' check before the 'total_bavail' check. This
923 will result in an -EAGAIN and in the exit call path we call
924 alloc_rr() which will with increasing aggressiveness attempt to
925 aquire precreated objects on the minimum number of required OSCs.
929 Description: Use old size assignment to avoid deadlock
930 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
931 the scheduling while holding a spinlock. We do not need locking
932 for size in ll_update_inode() because size is only updated from
933 the MDS for directories or files without objects, so there is no
934 other place to do the update, and concurrent access to such inodes
935 are protected by the inode lock.
939 Description: resolve "_IOWR redefined" build error on SLES10
943 Description: dump the memory debugging after all modules are unloaded to
944 suppress false negative in conf_sanity test 39
946 Severity : enhancement
948 Description: build kernel-ib packages for OFED 1.3 in our release cycle
953 Description: fix SLES kernel versioning
954 Details : the kernel version for our SLES 10 kernel did not include a "-"
955 before the "smp" at the end. while this was not a problem in
956 general, it did mean that software trying to use the kernel
957 version to try to detect a vendor specific kernel would fail.
958 this was most evident by the OFED build scripts.
962 Description: Don't update lov_desc members until making sure they are valid
963 Details : When updating lov_desc members via proc fs, need fix their
964 validities before doing the real update.
968 Description: don't put request into delay list while invalidate in flight.
969 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
970 import in flight. this produce timeout for invalidate and sometimes
971 can cause stale data.
975 Frequency : on ppc only
976 Description: not convert ost objects for directory because it's not exist.
977 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
981 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
983 Severity : enhancement
985 Description: Update to support 2.6.22.14 vanilla kernel.
989 Frequency : rare, on recovery
990 Description: read procfs can produce deadlock in some situation
991 Details : Holding lprocfs lock which send rpc can produce block for destroy
992 obd objects and this also block reconnect with -EALREADY. This
993 isn't fix all lprocfs bugs - but make it rare.
995 Severity : enhancement
997 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1000 Frequency : frequent on X2 node
1002 Description: mdc_set_open_replay_data LBUG
1003 Details : Set replay data for requests that are eligible for replay.
1007 Description: lustre_mgs: operation 101 on unconnected MGS
1008 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1009 MGC, and late on MGC cannot successfully connect to MGS and a lot
1010 of the error messages complaining that MGS is not connected.
1014 Frequency : on network error
1015 Description: panic with double free request if network error
1016 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1017 but it's true only for synchronus enqueue, for async enqueue
1018 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1021 Severity : enhancement
1023 Description: client-side metadata stat-ahead during readdir(directory readahead)
1024 Details : perform client-side metadata stat-ahead when the client detects
1025 readdir and sequential stat of dir entries therein
1028 Frequency : on start mds
1030 Description: Implement get_info(last_id) in obdfilter.
1033 Frequency : occasional
1035 Description: Correctly check stale fid, not start epoch if ost not support SOM
1036 Details : open with flag O_CREATE need set old fid in op_fid3 because
1037 op_fid2 was overwritten with new generated fid, but mds can answer
1038 with one of these two fids and both is not stale. Setattr
1039 incorrectly started an epoch and assume will be called
1040 done_writing, but without SOM done_writing ever being called.
1043 Frequency : rare, depends on device drivers and load
1045 Description: MDS or OSS nodes crash due to stack overflow
1046 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1047 In some cases, in conjunction with device drivers that use a lot
1048 of stack the MDS (or possibly OSS) service threads could overflow
1049 the stack. One change which was identified to consume additional
1050 stack has been reworked to avoid the extra stack usage.
1053 Frequency : occasional
1055 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1056 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1057 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1058 oscc_has_objects and see OSC has no objests and call
1059 oscc_internal_create to resend create request.
1061 Severity : enhancement
1063 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1065 Severity : enhancement
1067 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1070 Frequency : very rare
1072 Description: Fix replay if there is an un-replied request and open
1073 Details : In some cases, older replay request will revert the
1074 mcd->mcd_last_xid on MDS which is used to record the client's
1075 latest sent request.
1077 Severity : enhancement
1079 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1081 Severity : enhancement
1083 Description: Add rhel5 support to HEAD.
1085 Serverity : enhancement
1087 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1092 Description: Don't allow skipping OSTs if index has been specified.
1093 Details : Don't allow skipping OSTs if index has been specified, make
1094 locking in internal create lots better.
1098 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1099 Details : access to bitfield in structure is always rounded to long
1100 and this produce problem with not atomic change any bit.
1104 Description: Lustre make rpms failed.
1105 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1106 builds Lustre rpms from tarball.
1109 Frequency : rare on shutdown ost
1111 Description: If llog cancel was not send before clean_exports phase, this can
1112 produce deadlock in llog code.
1113 Details : If llog thread has last reference to obd and call class_import_put
1114 this produce deadlock because llog_cleanup_commit_master wait when
1115 last llog_commit_thread exited, but this never success because was
1116 called from llog_commit_thread.
1120 Description: allow userland application know is lost one of stripes.
1121 Details : fill lvb_blocks with error code on ost and return it to
1122 application if error flag found.
1126 Description: NULL lov_tgts causing MDS oops
1127 Details : more safe checks for NULL lov_tgts for avoid oops.
1129 Severity : enhancement
1131 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1135 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1136 Details : If an OST is down the MDS will hang indefinitely in
1137 obd_statfs_rqset() waiting for the statfs data. While for
1138 MDS QOS usage of statfs, it should not stuck in waiting.
1140 Severity : enhancement
1142 Description: remote_acl support
1143 Details : Support ACL-based permission check for remote user.
1144 Support setfacl/getfacl for remote user with the utils
1145 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1146 the system "{s,g}etfacl" utils.
1148 Severity : enhancement
1150 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1152 Severity : enhancement
1154 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1158 Description: make mgs_setparam() handle fsname containing dash
1159 Details : fsname containing a dash does not work with lctl conf_param
1161 Severity : enhancement
1163 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1165 Severity : enhancement
1167 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1171 Frequency : rare, at shutdown
1172 Description: access already free / zero obd_namespace.
1173 Details : if client_disconnect_export was called without force flag set,
1174 and exist connect request in flight, this can produce access to
1175 NULL pointer (or already free pointer) when connect_interpret
1176 store ocd flags in obd_namespace.
1180 Frequency : only at startup
1181 Description: not alloc memory with spinlock held.
1182 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1183 if any spinlock held.
1185 Severity : enhancement
1187 Description: make lustre randomly fail allocating memory
1188 Details : Make lustre randomly failed allocating memory for testing purpose.
1190 Severity : enhancement
1192 Description: lost problems with lov objid file
1193 Details : Fixes some scability and access to not inited memory problems
1194 in work with lov objdid file.
1199 Description: lfs find does not continue on file error
1200 Details : Continue other files processing when a file/dir is absent.
1204 Description: Inconsistent usage of lustre_pack_reply()
1205 Details : Standardize the usage of lustre_pack_reply() such that it
1206 always generate a CERROR on failure.
1209 Frequency : occasional
1211 Description: MDS hang or stay in waiting lock
1212 Details : If client receive lock with CBPENDING flag ldlm need send lock
1213 cancel as separate rpc, to avoid situation when cancel request
1214 can't processed due all i/o threads stay in wait lock.
1218 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1222 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1226 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1230 Description: Update to RHEL4 latest kernel.
1232 Severity : enhancement
1234 Description: Build SLES10 patchless client fails
1235 Details : The configure was broken by run ./configure with
1236 --with-linux-obj=.... argument for patchless client. When the
1237 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1238 can't search header adequately. Use absolute path such as
1239 -I($LINUX)/include instead.
1243 Description: interrupt oig_wait produce painc on resend.
1244 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1245 private set, and this produce situation when rq_ptlrpcd_data not
1246 copyed to new allocated request and triggered LBUG on assert
1247 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1248 ptlrpcd set if request is changed and can be safety changed to use
1253 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1254 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1255 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1256 these macros and replace them with their difinitions instead.
1260 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1261 Details : Update which_patch & target file for SLES10 latest kernel.
1264 Description: Remove client patches from SLES 10 kernel.
1265 Details : This causes SLES 10 clients to behave as patchless clients
1266 even on a Lustre-patched (server) kernel.
1268 Severity : enhancement
1270 Description: self-adjustable client's lru lists
1271 Details : use adaptive algorithm for managing client cached locks lru
1272 lists according to current server load, other client's work
1273 pattern, memory activities, etc. Both, server and client
1274 side namespaces provide number of proc tunables for controlling
1277 Severity : enhancement
1279 Description: light-weight GSS support
1280 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1281 and reduce performance overhead.
1283 Severity : enhancement
1285 Description: Linux keyring support
1286 Details : Support using service of Linux keyring for Lustre GSS internal
1287 context refresh/cache mechanism.
1291 Description: Fix errors in lfs documentation
1292 Details : Fixes man pages
1296 Description: don't use GFP_* in generic Lustre code.
1297 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1301 Description: obdclass is limited by single OBD_ALLOC(idarray)
1302 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1306 Description: warnings with build patchless client with vanila 2.6.19 and up
1307 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1308 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1312 Frequency : Only for SLES
1313 Description: sanity_quota fail test_1
1314 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1315 file, which makes TSTID[2] inunique.
1319 Frequency : Only for SLES9
1320 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1321 Details : Update target/ChangeLog/which_patch .
1325 Frequency : Only for test_10 in sanity-quota.sh
1326 Description: a bug in quota test code
1327 Details : checking a test flag in a wrong place
1331 Frequency : Only for test_13 in sanity-quota.sh
1332 Description: a bug in quota test script
1333 Details : shouldn't check overfull allocated quota
1337 Description: jbd statistics
1338 Details : Port older jbd statistics patch for sles10
1340 Severity : enhancement
1342 Description: Add printing inode info into message about error in writepage.
1347 Description: reply_lock_interpret crash due to race with it and lock cancel.
1348 Details : Do not replay locks that are being cancelled. Do not reference
1349 locks by their address during replay, just by their handle.
1353 Frequency : When flocks are used.
1354 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1355 Details : Flock locks might destroy just granted lock if it could be merged
1356 with another existing flock, this is done in completion handler,
1357 so teach ldlm_cli_enquque_fini that this is a valid case for
1363 Description: Oops in read and write path when failing to allocate lock.
1364 Details : Check if lock allocation failed and return error back.
1368 Description: lstripe command fails for valid OST index
1369 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1370 instead of lov->desc.ld_active_tgt_count.
1374 Description: sanity.sh failed test 103
1375 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1380 Description: df doesn't work properly if diskfs blocksize != 4K
1381 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1385 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1386 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1388 Severity : enhancement
1390 Description: Adaptive timeouts
1391 Details : RPC timeouts adapt to changing server load and network
1392 conditions to reduce resend attempts and improve recovery time.
1396 Description: llapi_file_create() does not allow some changes
1397 Details : add llapi_file_open() that allows specifying the mode and
1398 open flags, and also returns an open file handle.
1402 Description: merge and cleanup kernel patches.
1403 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1407 Description: Add journal checksum support.(Kernel part)
1408 Details : The journal checksum feature adds two new flags i.e
1409 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1410 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1411 indicates that the commit block contains the checksum for
1412 the blocks described by the descriptor blocks. Now commit
1413 record can be sent to disk without waiting for descriptor
1414 blocks to be written to disk. This behavior is controlled
1415 using JBD2_FEATURE_ASYNC_COMMIT flag.
1419 Description: OSS needs mutliple precreate threads
1420 Details : Add ability to start more than one create thread per OSS.
1424 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1425 Details : execute lfs setstripe on client
1429 Description: mds_obd_create error creating tmp object
1430 Details : When the user sets quota on root, llog will be affected and can't
1431 create files and write files.
1434 Frequency : Always on ia64 patchless client, and possibly others.
1436 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1437 Details : This allows the patchless client to be loaded on architectures
1438 without this export.
1442 Description: RedHat Update kernel for RHEL5
1443 Details : Add the kernel config file for RHEL5.
1447 Description: RedHat Update kernel for RHEL5
1448 Details : Modify the kernel config file more closer RHEL5.
1452 Description: Build failure against Centos5 (RHEL5)
1453 Details : Define PAGE_SIZE when it isn't present.
1457 Description: test_7 on sanity-quota.sh can't be used on separate machines
1458 Details : add facet to handle it
1462 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1463 Details : with NFS, the anon dentry's parent was set to itself in
1464 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1465 corresponding dentry other than use rec->ur_name.
1467 Severity : enhancement
1469 Description: lfs setstripe enhancement
1470 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1474 Description: enable data checksumming by default
1475 Details : enable checksum by default, allow --disable-checksum
1476 configure option and "-o nochecksum" mount option. Checksums
1477 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1481 Description: lustre support for RHEL5
1482 Details : Add support for RHEL5.
1486 Description: Client eviction due to failover config
1487 Details : after a connection loss, the lustre client should attempt to
1488 reconnect to the last active server first before trying the
1489 other potential connections.
1493 Description: when mds and osts use different quota unit(32bit and 64bit),
1494 quota will be released repeatly.
1495 Details : void sending multiple quota reqs to mds, which will keep the
1496 status between the reqs.
1500 Description: osts not allocated evenly to files
1501 Details : change the condition to increase offset_idx
1504 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1506 Description: Data corruption for OSTs that are formatted larger than 2TB
1508 Details : When generating the bio request for lustre file writes the
1509 sector number would overflow a temporary variable before being
1510 used for the IO. The data reads correctly from Lustre (which
1511 will overflow in a similar manner) but other file data or
1512 filesystem metadata may be corrupted in some cases.
1516 Description: Tune the kernel for good SCSI performance.
1517 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1518 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1523 Description: rewrite ext2-derived code in obdclass/uuid.c
1524 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1525 from scratch preserving functionality.
1529 Description: "lfs find -obd UUID" prints directories
1530 Details : "lfs find -obd UUID" will return all directory names instead
1531 of just file names. It is incorrect because the directories
1532 do not reside on the OSTs.
1536 Description: use i_size_read and i_size_write in 2.6 port
1537 Details : replace inode->i_size access with i_size_read/write()
1540 Frequency : when removing large files
1542 Description: scheduling issue during removal of large Lustre files
1543 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1544 It causes scheduling issues when removing large files (17TB in the
1548 Frequency : only with liblustre clients on XT3
1550 Description: evictions taking too long
1551 Details : allow llrd to evict clients directly on OSTs
1554 Frequency : only on ppc
1556 Description: /proc/fs/lustre/devices broken on ppc
1557 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1558 We should make sure the type of 'index' is loff_t and then cast
1559 explicitly as needed below. Do not assign an explicitly cast
1563 Frequency : only for rhel5
1565 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1566 Details : Modify the target file & which_kernel.
1568 Severity : enhancement
1570 Description: omit set fsid for export NFS
1571 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1572 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1576 Frequency : Always, for kernels after 2.6.16
1577 Description: Fix warning idr_remove called for id=.. which is not allocated.
1578 Details : Last kernels save old s_dev before kill super and not allow
1579 to restore from callback - restore it before call kill_anon_super.
1583 Description: buffer overruns could theoretically occur
1584 Details : llapi_semantic_traverse() modifies the "path" argument by
1585 appending values to the end of the origin string, and a
1586 overrun may occur. Adding buffer overrun check in liblustreapi.
1590 Description: Fix error on 'ls .' at the top of the Lustre mount.
1591 Details : Don't revalidate dentry if it is a root dentry.
1595 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1596 Details : Modify vm-tunables-rhel4.patch.
1600 Description: Kernel config for 2.6.18-vanilla.
1601 Details : Modify targets/2.6-vanilla.target.in.
1602 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1603 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1604 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1605 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1609 Description: improve handling recoverable errors
1610 Details : if request processig with error which can be recoverable on server
1611 request should be resend, otherwise page released from cache and
1616 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1617 Details : Modify target file & which_patch.
1618 A flaw was found in the IA32 system call emulation provided
1619 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1620 value could be stored in the %RAX register, which could trigger an
1621 out-of-bounds system call table access. An untrusted local user
1622 could exploit this flaw to run code in the kernel
1623 (ie a root privilege escalation). (CVE-2007-4573).
1627 Description: O_DIRECT bypasses client statistics.
1628 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1629 read_ahead_stats were updated. copied the stats section from
1630 osc_send_oap_rpc() into async_internal().
1634 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1638 Description: change order of libsysio includes
1639 Details : '#include sysio.h' should always come before '#include xtio.h'
1643 Description: handle "serious error: objid * already exists" more gracefully
1644 Details : If LAST_ID value on disk is smaller than the objects existing in
1645 the O/0/d* directories, it indicates disk corruption and causes an
1646 LBUG(). If the object is 0-length, then we should use the existing
1647 object. This will help to avoid a full fsck in most cases.
1649 Severity : enhancement
1651 Description: adapt the lustre_config script to support the upgrade case
1652 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1653 server targets to 1.6 in parallel.
1657 Description: To avoid grant space > avaible space when the disk is almost
1658 full. Without this patch you might see the error "grant XXXX >
1659 available" or some LBUG about grant, when the disk is almost
1661 Details : In filter_check_grant, for non_grant cache write, we should
1662 check the left space by if (*left > ungranted + bytes), instead
1663 of (*left > ungranted), because only we are sure the left space
1664 is enough for another "bytes", then the ungrant space should be
1665 increase. In client, we should update cl_avail_grant only there
1666 is OBD_MD_FLGRANT in the reply.
1670 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1671 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1672 described in CVE-2007-4573. This problem could allow untrusted
1673 local users to gain root access.
1676 Frequency : when using O_DIRECT and quotas
1678 Description: Incorrect file ownership on O_DIRECT output files
1679 Details : block usage reported by 'lfs quota' does not take into account
1680 files that have been written with O_DIRECT.
1685 Description: touch file failed when fs is not full
1686 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1687 otherwise we can get ENOSP while fs is not full.
1691 Description: parallel lock callbacks
1692 Details : Instead of sending blocking and completion callbacks as separated
1693 requests, adding them to a set and sending in parallel.
1696 Frequency : only for Cray XT3
1697 Bugzilla : 12829/13455
1698 Description: Changing primary group doesn't change the group lustre assigns to
1700 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1701 overridden with the primary group provided by the group upcall,
1702 whereas the supplied fsgid can be trusted if it is in the list of
1703 supplementary groups returned by the group upcall.
1705 Severity : enhancement
1707 Description: Allow masking D_WARNING, D_ERROR messages from console
1708 Details : Console messages can now be disabled via lnet.printk.
1712 Description: User code with malformed file open parameter crashes client node
1713 Details : Before packing join_file req, all the related reference should be
1714 checked carefully in case some malformed flags cause fake
1715 join_file req on client.
1719 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1720 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1721 mode is equal to the granted mode. However, it should also check
1722 that LDLM_FL_CP_REQD is not set.
1726 Description: Heavy nfs access might result in deadlocks
1727 Details : After ELC code landed, it is now improper to enqueue any mds
1728 locks under och_sem, because enqueue might want to decide to
1729 cancel open locks for same inode we are holding och_sem for.
1733 Description: Client eviction while running blogbench
1734 Details : A lot of unlink operations with concurrent I/O can lead to a
1735 deadlock causing evictions. To address the problem, the number of
1736 oustanding OST_DESTROY requests is now throttled to
1737 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1738 callbacks are processed in priority.
1742 Description: enable ACLs on MDS by default
1743 Details : ACLs must be enabled on MDS by default.
1746 Frequency : PPC/PPC64 only
1748 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1749 Details : fix duplicated definitions between asm-ppc64/types.h and
1750 lustre_types.h on PPC.
1753 Frequency : PPC/PPC64 only
1755 Description: asm-ppc/segment.h does not exist
1756 Details : fix compile issue on PPC.
1760 Description: better handle error messages in extents code
1763 Frequency : RHEL4 only
1765 Description: mkfs is very slow on IA64/RHEL4
1766 Details : A performance regression has been discovered in the MPT Fusion
1767 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1768 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1769 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1771 Severity : enhancement
1773 Description: SNMP support enhancement
1774 Details : Adding total number of sampled request for an MDS node in snmp
1777 Severity : enhancement
1779 Description: Optimize ldlm waiting list processing for PR extent locks
1780 Details : When processing waiting list for read extent lock and meeting read
1781 lock that is same or wider to it that is not contended, skip
1782 processing rest of the list and immediatelly return current
1783 status of conflictness, since we are guaranteed there are no
1784 conflicting locks in the rest of the list.
1788 Description: Time out and refuse to reconnect
1789 Details : When the failover node is the primary node, it is possible
1790 to have two identical connections in imp_conn_list. We must
1791 compare not conn's pointers but NIDs, otherwise we can defeat
1792 connection throttling.
1796 Description: port llog fixes from b1_6 into HEAD
1797 Details : Port llog reference couting and some llog cleanups from b1_6
1798 (bug 10800) into HEAD, for protect from panic and access to already
1799 free llog structures.
1803 Description: Detect stride IO mode in read-ahead
1804 Details : When a client does stride read, read-ahead should detect that and
1805 read-ahead pages according to the detected stride pattern.
1809 Description: data checksumming impacts single node performance
1810 Details : add support for several checksum algorithm. Currently, only CRC32
1811 and Adler-32 are supported. The checksum type can be changed on
1812 the fly via /proc/fs/lustre/osc/*/checksum_type.
1816 Description: use adler32 for page checksums
1817 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1822 Description: build for x2 fails
1823 Details : fix compile issue on Cray systems.
1827 Description: Properly match for duplicate locks
1828 Details : Due to different lock order from skiplists code, we need to
1829 traverse entire list for now
1832 Frequency : only on PPC/SLES10
1834 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1835 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1836 be defined. Add a hack in mkfs_lustre.c to work around this compile
1841 Description: LASSERT on MDS when client holding flock lock dies
1842 Details : ldlm pool logic depends on number of granted locks equal to
1843 number of released locks which is not true for flock locks, so
1844 just exclude such locks from consideration.
1848 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1849 Details : Use fsfilt_sync() for both the whole filesystem sync and
1850 individual file sync to eliminate dangerous inode locking
1851 with I_LOCK that can lead to a deadlock.
1855 Description: performance in 1.6.3
1856 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1857 sure that 1MB requests are not fragmented by the block layer.
1859 Severity : enhancement
1861 Description: organize the server-side client stats on per-nid basis
1862 Details : Change the structure of stats under obdfilter and mds to
1870 The "uuid"s file would list the uuids of _active_ exports.
1871 And the clear entry is to clear all stats and stale nids.
1873 Severity : enhancement
1875 Description: eliminate client locks in face of contention
1876 Details : file contention detection and lockless i/o implementation
1877 for contended files.
1881 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1884 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1886 Description: MDS may lose file striping (and hence file data) in some cases
1887 Details : If there are additional extended attributes stored on the MDS,
1888 in particular ACLs, SELinux, or user attributes (if user_xattr
1889 is specified for the client mount options) then there is a risk
1890 of attribute loss. Additionally, the Lustre file striping
1891 needs to be larger than default (e.g. striped over all OSTs),
1892 and an additional attribute must be stored initially in the
1893 inode and then increase in size enough to be moved to the
1894 external attribute block (e.g. ACL growing in size) for file
1899 Description: skiplist implementation simplification
1900 Details : skiplists are used to group compatible locks on granted list
1901 that was implemented as tracking first and last lock of each
1902 lock group the patch changes that to using doubly linked lists
1906 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1907 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1908 because LASSERT on both the data supplied by a client, and the
1909 data on disk is dangerous and incorrect.
1911 Severity : enhancement
1913 Description: Slow truncate/writes to huge files at high offsets.
1914 Details : Directly associate cached pages to lock that protect those pages,
1915 this allows us to quickly find what pages to write and remove
1916 once lock callback is received.
1920 Description: more ldlm soft lockups
1921 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1922 starve other threads from the resource lock for a long time in
1923 case of long waiting queue, so change the debug level from
1924 D_OTHER to the less frequently used D_INFO.
1926 Severity : enhancement
1928 Description: add -gid, -group, -uid, -user options to lfs find
1932 Description: Hung threads in invalidate_inode_pages2_range
1933 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1934 once one is completed. As a result, some RPCs are stuck in the
1935 queue and are never sent.
1939 Description: filter threads hungs on waiting journal commit
1940 Details : Cleanup filter group llog code, then only filter group llog will
1941 be only created in the MDS/OST syncing process.
1945 Description: Procfs and llog threads access destoryed import sometimes.
1946 Details : Sync the import destoryed process with procfs and llog threads by
1947 the import refcount and semaphore.
1949 Severity : enhancement
1951 Description: openlock cache of b1_6 port to HEAD
1956 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1957 Details : If insertion of an extent fails, then discard the inode
1958 preallocation and free data blocks else it can lead to duplicate
1963 Description: don't always update ctime in ext3_xattr_set_handle()
1964 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1965 In some cases the ctime should not be updated, for example for
1966 2.0->1.8 compatibility it is necessary to delete an xattr and it
1967 should not update the ctime.
1971 Bugzilla : 15713/16362
1972 Description: Assertion in iopen_connect_dentry in 1.6.3
1973 Details : looking up an inode via iopen with the wrong generation number can
1974 populate the dcache with a disconneced dentry while the inode
1975 number is in the process of being reallocated. This causes an
1976 assertion failure in iopen since the inode's dentry list contains
1977 both a connected and disconnected dentry.
1981 Description: assertion failure in ldlm_handle2lock()
1982 Details : fix a race between class_handle_unhash() and class_handle2object()
1983 introduced in lustre 1.6.5 by bug 13622.
1988 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1989 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1994 Description: oops in sd_iostats_seq_show()
1995 Details : unloading/reloading the scsi low level driver triggers a kernel
1996 bug when trying to access the sd iostat file.
2001 Description: Kernel panics during QLogic driver reload
2002 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2003 patch, causing memory corruption.
2008 Description: journal_dev option does not work in b1_6
2009 Details : pass mount option during pre-mount.
2011 Severity : enhancement
2013 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2014 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2015 extent information of a file. It can be used to map logical blocks
2016 in a file to physical blocks in the block device.
2020 Description: LDLM soft lockups - improvement
2021 Details : It is be possible to send the lock handle along with each read
2022 or write request because the client is already doing a lock match
2023 itself so there isn't any reason the OST should have to re-do that
2027 Frequency : only on Cray X2
2029 Description: X2 build failures
2030 Details : fix build failures on Cray X2.
2034 Description: xid & resent requests
2035 Details : Initialize RPC XID from clock at startup (randomly if clock is
2038 Severity : enhancement
2040 Description: Add lustre_start utility to start or stop multiple Lustre servers
2045 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2046 Details : In case of memory pressure, list_del() can be called twice on
2047 req->rq_history_list, causing a kernel oops.
2051 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2052 Details : fix stack overflow in the distributed lock manager by defering
2053 export eviction after a failed AST to the elt thread instead of
2054 handling it in the dlm interpret routine.
2058 Description: Convert some comments to new format.
2059 Details : Update documenting comments to match doxygen conventions.
2063 Description: Grammar fixes.
2064 Details : A couple of trivial spelling fixes.
2068 Description: OSD_COUNTERS-mandatory
2069 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2070 to simplify the code.
2074 Description: simplify cmm_device freeing logic.
2075 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2079 Description: Add lockdep support to dt_object_operations locking interface.
2080 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2081 parameter indicating lock ordering. Update mdd code to use new
2086 Description: Introduce failloc constants for lockless IO tests.
2087 Details : Add two new failloc constants to test lockless IO. Only one of
2088 them in implemented---another is checked in yet to be landed
2093 Description: Add lockdep support for inode mutex.
2094 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2095 in the situations where Lustre has to lock more than one inode
2100 Description: Add optional invariants checking support.
2101 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2102 configure switch. This macro is to be used for consistency and
2103 sanity checks that are too expensive to be left in `production'
2108 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2109 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2114 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2115 Details : Use spin_lock_nested() in (the only) situation where more than
2116 one ldlm_lock is locked simultaneously. Also, fix possible
2117 dead-lock in ldlm_lock_change_resource() by enforcing particular
2122 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2123 Details : Instead of specifying each ldlm_lock call-back through separate
2124 parameter, wrap them into struct ldlm_callback_suite.
2128 Description: Kill join_lru obd method and its callers.
2129 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2130 in memory---a requirement implemented through ->o_join_lru() obd
2131 method in HEAD. Remove this method and its users.
2135 Description: Add asynchronous ldlm ENQUEUE completion handler.
2136 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2137 case that stock ldlm_completion_ast() cannot handle as it waits
2138 until lock is granted. Introduce new ldlm_completion_ast_async()
2139 for this. Also comment ldlm_completion_ast().
2143 Description: ldlm_error <-> errno conversion.
2144 Details : Add functions to map (rather arbitrary) between LDLM error codes
2145 and standard errno values. CLIO needs this to prevent LDLM specific
2146 constants from escaping ldlm and osc.
2150 Description: Kill unused ldlm_handle2lock_ns() function.
2151 Details : Kill unused ldlm_handle2lock_ns() function.
2155 Description: Add lu_ref support to ldlm_lock
2156 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2157 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2158 and ldlm_resource. LDLM interface has to be changed, because code that
2159 releases a reference on a lock, has to "know" what reference this is.
2160 In the most frequent case
2162 lock = ldlm_handle2lock(handle);
2164 LDLM_LOCK_PUT(lock);
2166 no changes are required. When any other reference (received _not_ from
2167 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2168 instead of LDLM_LOCK_PUT().
2170 Arguably, changes are pervasive, and interface requires some discipline
2171 for proper use. On the other hand, it was very instrumental in finding
2172 a few leaked lock references.
2176 Description: Add ldlm_lock_addref_try().
2177 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2178 attempts to addref a lock that might be being canceled
2183 Description: Add ldlm_weigh_callback().
2184 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2185 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2186 hard-coded `number of pages' logic.
2190 Description: Add lockdep annotations to llog code.
2191 Details : Use appropriately tagged _nested() locking calls in the places
2192 where llog takes more than one ->lgh_lock lock.
2196 Description: Add loi_kms_set().
2197 Details : Wrap kms updates into a helper function.
2201 Description: Constify instances of struct lsm_operations.
2202 Details : Constify instances of struct lsm_operations.
2206 Description: lu_conf support.
2207 Details : On a server, a file system object is uniquely identified
2208 by a fid, which is sufficient to locate and load all object
2209 state (inode). On a client, on the other hand, more data are
2210 necessary instantiate an object. Change lu_object_find() and
2211 friends to take additional `lu_conf' argument describing object.
2212 Typically this includes layout information.
2216 Description: lu_context fixes.
2217 Details : Introduce new lu_context functions that are needed on the client
2218 side, where some system threads (ptlrpcd) are shared by multiple
2219 modules, and so cannot be stopped during module shutdown.
2223 Description: Add start and stop methods to lu_device_type_operations.
2224 Details : Introduce two new methods in lu_device_type_operations, that are
2225 invoked when first instance of a given type is created and last one
2226 is destroyed respectively. This is need by CLIO.
2230 Description: Add lu_ref support to struct lu_device.
2231 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2232 track leaked references.
2236 Description: Introduce lu_kmem_descr.
2237 Details : lu_kmem_descr and its companion interface allow to create
2238 and destroy a number of kmem caches at once.
2242 Description: Fix lu_object finalization race.
2243 Details : Fix a race between lu_object_find() finding an object and its
2244 concurrent finalization. This race is (most likely) not possible
2245 on the server, but might happen on the client.
2249 Description: Introduce lu_ref interface.
2250 Details : lu_ref is a debugging module allowing to track references to
2251 a given object. It is quite cpu expensive, and has to be
2252 explicitly enabled with --enable-lu_ref. See usage description
2257 Description: Factor lu_site procfs stats into a separate function.
2258 Details : Separate lu_site stats printing code into a separate function
2259 that can be reused on a client.
2263 Description: Constify instances of struct {lu,dt,md}_device_operations.
2264 Details : Constify instances of struct {lu,dt,md}_device_operations.
2268 Description: Introduce struct md_site and move meta-data specific parts of
2269 struct lu_site here.
2270 Details : Move md-specific fields out of struct lu_site into special struct
2271 md_site, so that lu_site can be used on a client.
2275 Description: Kill mdd_lov_destroy().
2276 Details : Remove unused mdd code.
2280 Description: Add st_block checking to multistat.c.
2281 Details : Add st_block checking to multistat.c.
2285 Description: Add lu_ref support to struct obd_device.
2286 Details : Add lu_ref tracking to obd_device.
2290 Description: Kill obd_set_fail_loc().
2291 Details : Remove unused code.
2295 Description: Add special type for ptlrpc_request interpret functions.
2296 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2297 there. Actual usage will be in CLIO.
2301 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2302 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2303 doesn't work with lockdep.
2307 Description: Add rwv.c test program.
2308 Details : New testing program exercising readv(2) and writev(2) (Qian).
2312 Description: Add sendfile.c test program.
2313 Details : New testing program exercising sendfile(2) (Jay).
2317 Description: Ratelimit a message that can be very frequent.
2318 Details : Ratelimit a memory allocation failure message that can
2323 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2324 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2328 Description: Liblustre build fixes.
2329 Details : Liblustre build fixes.
2333 Description: libcfs: add cfs_{need,cond}_resched() interface.
2334 Details : libcfs: add cfs_{need,cond}_resched() definition and
2335 implementations for Linux, NT, and liblustre.
2337 Severity : enhancement
2339 Description: More exported tunables for mballoc
2340 Details : Add support for tunable preallocation window and new tunables for
2341 large/small requests
2345 Description: Detect corruption of block bitmap and checking for preallocations
2346 Details : Checks validity of on-disk block bitmap. Also it does better
2347 checking of number of applied preallocations. When corruption is
2348 found, it turns filesystem readonly to prevent further corruptions.
2352 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2353 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2354 window to 0. If it does hit cache miss, and read-pattern is still
2355 stride-io mode, does not reset the stride window, but also does not
2356 increase the stride window length in this case.
2360 Frequency : only for big-endian servers
2361 Description: Check if system is big-endian while mounting fs with extents feature
2362 Details : Mounting a filesystem with extents feature will fail on big-endian
2363 systems since ext3-based ldiskfs is not supported on big-endian
2364 systems. This can be over-riden with "bigendian_extents" mount option.
2366 Severity : enhancement
2368 Description: The root squash functionality
2369 Details : A security feature, which is to prevent users from being able
2370 to mount lustre on their desktop, run as root, and delete
2371 all of the files in the filesystem. The goal is accomplished by
2372 remapping user id (UID) and group id (GID) of the root user to
2373 a UID and GID specified by the system administartor via Lustre
2374 configuration management server (MGS). The functionality also
2375 allows to specify sets of clients for which the remapping does
2380 Description: Excessive recovery window
2381 Details : With AT enabled, the recovery window can be excessively long (6000+
2382 seconds). To address this problem, we no longer use
2383 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2384 timeout no longer depends on the service time, it is set to
2385 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2386 time via pb_service_time.
2390 Description: Watchdog triggered on MDS failover
2391 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2392 the OSTs know that the MDS "UUID" can be reused for the same export
2393 from a different NID, so we do not need to wait for the export to be
2397 Frequency : rare, only if using MMP with Linux RAID
2399 Description: MMP doesn't work with Linux RAID
2400 Details : While using HA for Lustre servers with Linux RAID, it is possible
2401 that MMP will not detect multiple mounts. To make this work we
2402 need to unplug the device queue in RAID when the MMP block is being
2403 written. Also while reading the MMP block, we should read it from
2404 disk and not the cached one.
2406 Severity : enhancement
2408 Description: open file using fid
2409 Details : A file can be opened using just its fid, like
2410 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2413 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2416 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2417 Details : add proper locking for creation of jbd_4k slab cache
2421 Description: MMP check in ext3_remount() fails without displaying any error
2422 Details : When multiple mount protection fails during remount, proper error
2425 Severity : enhancement
2427 Description: Allow stripe size to be up to 4G-64k
2428 Details : Fix math logic to allow large stripe sizes.
2432 Description: add check for >8TB ldiskfs filesystems
2433 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2434 Don't allow >8TB ldiskfs filesystems to be mounted without
2435 force_over_8tb mount option
2440 Descriptoin: OSS complains "too many blocks" and crashes
2441 Details : In filter_commitrw_write():
2442 1. lquota_chkquota();
2443 2. obdo_from_inode();
2444 3. lquota_pending_commit();
2445 "2." has changed uid and gid in this case, then caused
2446 lquota_chkquota() and lquota_pending_commit() to use
2447 different uid and gid. That is the root of the bug.
2451 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2452 Details : Detect and print out the 'stripe count = -1' case properly.
2456 Description: Out or order replies might be lost on replay
2457 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2458 request with tid smaller then one currently being added, add it
2459 to the start, not end of the list.
2461 --------------------------------------------------------------------------------
2463 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2465 * Support for kernels:
2466 2.6.5-7.283 (SLES 9),
2467 2.6.9-55.EL (RHEL 4),
2468 2.6.16.46-0.14 (SLES 10),
2469 2.6.18.8 vanilla (kernel.org)
2470 * Client support for unpatched kernels:
2471 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2472 2.6.16 - 2.6.22 vanilla (kernel.org)
2473 * Due to recently discovered recovery problems, we do not recommend
2474 using patchless RHEL 4 clients with this or any earlier release.
2475 * Recommended e2fsprogs version: 1.39.cfs8
2476 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2477 * Starting with this release, the ldiskfs backing filesystem required
2478 by Lustre is now in its own package, lustre-ldiskfs. This package
2479 should be installed. It is versioned separately from Lustre and
2480 may be released separately in future.
2484 Description: block reactivating mgc import until all deactivates complete
2485 Details : Fix race when failing back MDT/MGS to itself (testing)
2488 Frequency : at statup only
2490 Description: mds_lov_synchronize race leads to various problems
2491 Details : simultaneous MDT->OST connections at startup can cause the
2492 sync to abort, leaving the OSC in a bad state.
2494 Severity : enhancement
2496 Description: add optional extra BUILD_VERSION info
2497 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2498 to override the lustre version.
2501 Frequency : 2.6.18 servers only
2503 Description: ll_kern_mount() doesn't release the module reference
2504 Details : The ldiskfs module reference count never drops down to 0
2505 because ll_kern_mount() doesn't release the module reference.
2510 Description: server LBUG when using old ost_num_threads parameter
2511 Details : Accept the old ost_num_threads parameter but warn that it
2512 is deprecated, and fix an off-by-one error that caused an LBUG.
2517 Description: Transient SCSI error results in persistent IO issue
2518 Details : iobuf->dr_error is not reinitialized to 0 between two
2522 Frequency : sometimes when underlying device returns I/O errors
2524 Description: OSTs not going read-only during write failures
2525 Details : OSTs are not remounted read-only when the journal commit threads
2526 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2527 instead of the ext3 wrappers.
2531 Description: poor connect scaling with increasing client count
2532 Details : Don't run filter_grant_sanity_check for more than 100 exports
2533 to improve scaling for large numbers of clients.
2536 Frequency : SLES10 only
2538 Description: sanity-quota.sh quotacheck failed: rc = -22
2539 Details : Quotas cannot be enabled on SLES10.
2542 Frequency : liblustre clients only
2544 Description: getdirentries does not give error when run on compute nodes
2545 Details : getdirentries does not fail when the size specified as an argument
2546 is too small to contain at least one entry
2548 Severity : enhancement
2550 Description: Add LNET router traceability for debug purposes
2551 Details : If a checksum failure occurs with a router as part of the
2552 IO path, the NID of the last router that forwarded the bulk data
2553 is printed so it can be identified.
2558 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2559 Details : Due to a race condition, liblustre clients were occasionally
2560 evicted incorrectly.
2562 Severity : enhancement
2564 Description: lfs setstripe use optional parameters instead of postional
2567 Severity : enhancement
2569 Description: Nanosecond timestamp support for ldiskfs
2570 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2571 resolution timestamps. There is not yet support for this at
2572 the Lustre filesystem level.
2575 Frequency : during server recovery
2577 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2578 Details : request with rq_no_resend flag not awake l_wait_event if they get
2582 Frequency : nfs export on patchless client
2584 Description: connectathon hang when test nfs export over patchless client
2585 Details : Disconnected dentry cannot be found with lookup, so we do not need
2586 to unhash it or make it invalid
2589 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2590 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2591 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2596 Description: open req refcounting wrong on reconnect
2597 Details : If reconnect happened between getting open reply from server and
2598 call to mdc_set_replay_data in ll_file_open, we will schedule
2599 replay for unreferenced request that we are about to free.
2600 Subsequent close will crash in variety of ways.
2601 Check that request is still eligible for replay in
2602 mdc_set_replay_data().
2607 Description: disable writes to filesystem when reading health_check file
2608 Details : the default for reading the health_check proc file has changed
2609 to NOT do a journal transaction and write to disk, because this
2610 can cause reads of the /proc file to hang and block HA state
2611 checking on a healthy but otherwise heavily loaded system. It
2612 is possible to return to the previous behaviour during configure
2613 with --enable-health-write.
2615 Severity : enhancement
2617 Description: 64-bit inode version
2618 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2619 to the inode. This will be required for version-based recovery.
2624 Description: MDS fails to start if a duplicate client export is detected
2625 Details : in some rare cases it was possible for a client to connect to
2626 an MDS multiple times. Upon recovery the MDS would detect this
2627 and fail during startup. Handle this more gracefully.
2629 Severity : enhancement
2631 Description: Add -o localflock option to simulate old noflock
2633 Details : This will achieve local-only flock/fcntl locks
2639 Description: log_commit_thread vs filter_destroy race leads to crash
2640 Details : Take import reference before releasing llog record semaphore
2645 Description: Wrong request locking in request set processing
2646 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2647 imp_delayed_list, in this place should be used imp_lock.
2650 Frequency : when reconnection
2652 Description: Grant Leak when osc reconnect to OST
2653 Details : When osc reconnect ost, OST(filter) should check whether it
2654 should grant more space to client by comparing fed_grant and
2655 cl_avail_grant, and return the granted space to client instead
2656 of "new granted" space, because client will call osc_init_grant
2657 to update the client grant space info.
2660 Frequency : when client reconnect to OST
2662 Description: Grant Leak when osc do resend and replay bulk write
2663 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2664 bulk write request, because the grant info while be sync between
2665 OSC and OST when reconnect, and we should ignore the grant info
2666 these of resend/replay write req.
2671 Description: Grant space more than avaiable left space sometimes.
2672 Details : When then OST is about to be full, if two bulk writing from
2673 different clients came to OST. Accord the avaliable space of the
2674 OST, the first req should be permitted, and the second one
2675 should be denied by ENOSPC. But if the seconde arrived before
2676 the first one is commited. The OST might wrongly permit second
2677 writing, which will cause grant space > avaiable space.
2680 Frequency : when client is evicted
2682 Description: Grant might be wrongly erased when osc is evicted by OST
2683 Details : when the import is evicted by server, it will fork another
2684 thread ptlrpc_invalidate_import_thread to invalidate the
2685 import, where the grant will be set to 0. While the original
2686 thread will update the grant it got when connecting. So if
2687 the former happened latter, the grant will be wrongly errased
2688 because of this race.
2693 Description: Checking Stale with correct fid
2694 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2695 to check whether it is stale, because sometimes, we want the
2696 enqueue happened anyway, and op_data.fid2 will not be initialized.
2698 Severity : enhancement
2700 Description: update patchless client
2701 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2704 Frequency : only with 2.4 kernel
2706 Description: random memory corruption
2707 Details : size of struct ll_inode_info is to big for union inode.u and this
2708 can be cause of random memory corruption.
2713 Description: Memory leak in recovery
2714 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2715 It should also check obd_fail before fsfilt_start, otherwise if
2716 fsfilt_start return -EROFS,(failover mds during mds recovery).
2717 then the req will return with repmsg->transno = 0 and rc = EROFS.
2718 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2719 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2720 be freed no matter whether fsfilt_commit success or not.
2723 Frequency : only with huge count clients
2725 Description: Prevents from taking the superblock lock in llap_from_page for
2727 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2728 for llap_from_page call in ll_removepage prevents from taking the
2729 superblock lock for a soon died page.
2734 Description: Not check open intent error before release open handle
2735 Details : in some rare cases, the open intent error is not checked before
2736 release open handle, which may cause
2737 ASSERTION(open_req->rq_transno != 0), because it tries to release
2738 the failed open handle.
2743 Description: Set cat log bitmap only after create log success.
2744 Details : in some rare cases, the cat log bitmap is set too early. and it
2745 should be set only after create log success.
2749 Description: Accessing a block bevice can re-enable I/O when Lustre is
2750 tearing down a device.
2751 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2756 Description: service threads may hog cpus when there are a lot of requests
2757 Details : Insert cond_resched to give other threads a chance to use some CPU
2762 Description: the cat log was not initialized in recovery
2763 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2764 unlink log on mds will not be initialized until mds post
2765 recovery. And also in mds post recovery, the unlink log will
2766 initialization will be done asynchronausly, so there will be race
2767 between add unlink log and unlink log initialization.
2771 Description: brw_stats were being printed incorrectly
2772 Details : brw_stats were being printed as log2 but all of them were not
2773 recorded as log2. Also remove some code duplication arising from
2774 filter_tally_{read,write}.
2778 Frequency : rare, only in recovery.
2779 Description: ASSERTION(req->rq_type != LI_POISON) failed
2780 Details : imp_lock should be held while iterating over imp_sending_list for
2781 prevent destroy request after get timeout in ptlrpc_queue_wait.
2785 Description: replay-single.sh test 52 fails
2786 Details : A lock's skiplist need to be cleanup when it being unlinked
2787 from its resource list.
2791 Description: Short directio read returns full requested size rather than
2793 Details : Direct I/O operations should return actual amount of bytes
2794 transferred rather than requested size.
2796 Severity : enhancement
2798 Description: metadata RPC reduction (e.g. for rm performance)
2799 Details : decrease the amount of synchronous RPC between clients and servers
2800 by canceling conflicing lock before the operation on the client
2801 and packing thier handles into the main operation RPC to server.
2803 Severity : enhancement
2805 Description: Async OSC create to avoid the blocking unnecessarily.
2806 Details : If a OST has no remain object, system will block on the creating
2807 when need to create a new object on this OST. Now, ways use
2808 pre-created objects when available, instead of blocking on an
2809 empty osc while others are not empty. If we must block, we block
2810 for the shortest possible period of time.
2814 Description: improve handling recoverable errors
2815 Details : if request processig with error which can be recoverable on server
2816 request should be resend, otherwise page released from cache and
2819 Severity : enhancement
2821 Description: refine locking for avoid write wrong info into lov_objid file
2822 Details : fix possible races with add new target and write/update data in
2825 --------------------------------------------------------------------------------
2827 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2832 Frequency : on some architectures
2834 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2835 Details : Uninitialized flags sometimes cause configuration commands to
2839 Frequency : patchless clients only
2841 Description: missing __iget() symbol export
2842 Details : The __iget() symbol export is missing. To avoid the need for
2843 this on patchless clients the deathrow inode reaper is turned
2844 off, and we depend on the VM to clean up old inodes. This
2845 dependency was during via the fix for bug 12181.
2847 --------------------------------------------------------------------------------
2849 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2851 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2852 INTEROPERATE with older versions automatically. In many cases a
2853 special upgrade step is needed. Please read the
2854 user documentation before upgrading any part of a live system.
2855 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2856 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2857 * WARNING: Lustre configuration and startup changes are required with
2858 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2862 Severity : enhancement
2864 Description: MountConf
2865 Details : Lustre configuration is now managed via mkfs and mount
2866 commands instead of lmc and lconf. New obd types (MGS, MGC)
2867 are added for dynamic configuration management. See
2868 https://mail.clusterfs.com/wikis/lustre/MountConf for
2871 Severity : enhancement
2873 Description: dynamic OST addition
2874 Details : OSTs can now be added to a live filesystem
2876 Severity : enhancement
2878 Description: startup order invariance
2879 Details : MDTs and OSTs can be started in any order. Clients only
2880 require the MDT to complete startup.
2882 Severity : enhancement
2884 Description: parallel, asynchronous orphan cleanup
2885 Details : orphan cleanup is now performed in separate threads for each
2886 OST, allowing parallel non-blocking operation.
2888 Severity : enhancement
2890 Description: optimized stripe assignment
2891 Details : stripe assignments are now made based on ost space available,
2892 ost previous usage, and OSS previous usage, in order to try
2893 to optimize storage space and networking resources.
2895 Severity : enhancement
2897 Description: Permanently set tunables
2898 Details : All writable /proc/fs/lustre tunables can now be permanently
2899 set on a per-server basis, at mkfs time or on a live system.
2901 Severity : enhancement
2903 Description: Lustre message v2
2904 Details : Add lustre message format v2.
2906 Severity : enhancement
2908 Description: client OST exclusion list
2909 Details : Clients can be started with a list of OSTs that should be
2910 declared "inactive" for known non-responsive OSTs.
2914 Description: SPEC SFS validation failure on NFS v2 over lustre.
2915 Details : Changes the blocksize for regular files to be 2x RPC size,
2916 and not depend on stripe size.
2918 Severity : enhancement
2920 Description: Multiple MD RPCs in flight.
2921 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2922 To avoid overly-overloading MDS, introduce a limit on number of
2923 MDS RPCs in flight for a single client and add /proc controls
2924 to adjust this limit.
2926 Severity : enhancement
2928 Description: client read/write statistics
2929 Details : Add client read/write call usage stats for performance
2930 analysis of user processes.
2931 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2932 file access. extents_stats shows chunk size distribution.
2933 extents_stats_per_process show chunk size distribution per
2936 Severity : enhancement
2938 Description: mds statistics
2939 Details : Add detailed mds operations statistics in
2940 /proc/fs/lustre/mds/*/stats.
2944 Description: Failure of copying files with lustre special EAs.
2945 Details : Client side always return success for setxattr call for lustre
2946 special xattr (currently only "trusted.lov").
2951 Description: Refcount LNET uuids
2952 Details : The global LNET uuid list grew linearly with every startup;
2953 refcount repeated list entries instead of always adding to
2956 Severity : enhancement
2958 Description: Dynamic service threads
2959 Details : Within a small range, start extra service threads
2960 automatically when the request queue builds up.
2963 Frequency : mixed-endian client/server environments
2965 Description: mixed-endian crashes
2966 Details : The new msg_v2 system had some failures in mixed-endian
2969 Severity : enhancement
2971 Description: Easy OST removal
2972 Details : OSTs can be permanently deactivated with e.g. 'lctl
2973 conf_param lustre-OST0001.osc.active=0'
2975 Severity : enhancement
2977 Description: MGS proc entries
2978 Details : Added basic proc entries for the MGS showing what filesystems
2981 Severity : enhancement
2983 Description: provide MGS failover
2984 Details : Added config lock reacquisition after MGS server failover.
2986 Severity : enhancement
2988 Description: add Linux 2.4 support
2989 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2993 Description: a large application tries to do I/O to the same resource and dies
2994 in the middle of it.
2995 Details : Check the req->rq_arrival time after the call to
2996 ost_brw_lock_get(), but before we do anything about
2997 processing it & sending the BULK transfer request. This
2998 should help move old stale pending locks off the queue as
2999 quickly as obd_timeout.
3002 Frequency : when an incorrect nid is specified during startup
3004 Description: ptlrpc connect to non-existant node causes kernel crash
3005 Details : LNET can't be re-entered from an event callback, which
3006 happened when we expire a message after the export has been
3007 cleaned up. Instead, hand the zombie cleanup off to another
3010 Severity : enhancement
3012 Description: plain/inodebits lock performance improvement
3013 Details : Grouping plain/inodebits in granted list by their request modes
3014 and bits policy, thus improving the performance of search through
3018 Frequency : only if OST filesystem is corrupted
3020 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3021 Details : for a short time RPCs with bulk IO are in the replay list,
3022 but replay of bulk IOs is unimplemented. If the OST filesystem
3023 is corrupted due to disk cache incoherency and then replay is
3024 started it is possible to trip an assertion. Avoid putting
3025 committed RPCs into the replay list at all to avoid this issue.
3028 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3030 Description: System hang on startup
3031 Details : This bug allowed the liblustre (e.g. catamount) client to
3032 return to the app before handling all startup RPCs. This
3033 could leave the node unresponsive to lustre network traffic
3034 and manifested as a server ptllnd timeout.
3036 Severity : enhancement
3038 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3039 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3040 boolean to control whether to print peer debug info when a
3041 client's RPC times out.
3044 Frequency : only for kernels with patches from Lustre below 1.4.3
3046 Description: Remove old rdonly API
3047 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3050 Frequency : only for devices with external journals
3052 Description: Set external device read-only also
3053 Details : During a commanded failover stop, we set the disk device
3054 read-only while the server shuts down. We now also set any
3055 external journal device read-only at the same time.
3058 Frequency : when upgrading from 1.4 while trying to change parameters
3060 Description: The wrong (new) MDC name was used when setting parameters for
3061 upgraded MDT's. Also allows changing of OSC (and MDC)
3062 parameters if --writeconf is specified at tunefs upgrade time.
3065 Frequency : when setting specific ost indicies
3067 Description: QOS code breaks on skipped indicies
3068 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3069 created with --index need not be sequential.
3071 Severity : enhancement
3073 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3074 Details : The uninit_groups feature works in conjunction with the kernel
3075 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3076 up the pass1 processing of e2fsck. This is a read-only feature
3077 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3078 mount filesystems that have had this feature enabled.
3080 Severity : enhancement
3082 Description: Improve multi-block allocation algorithm to avoid fragmentation
3083 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3084 allocation locality and avoid filesystem fragmentation.
3086 ------------------------------------------------------------------------------
3088 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3090 * Support for kernels:
3091 2.4.21-47.0.1.EL (RHEL 3)
3092 2.6.5-7.283 (SLES 9)
3093 2.6.9-42.0.10.EL (RHEL 4)
3094 2.6.12.6 vanilla (kernel.org)
3095 2.6.16.27-0.9 (SLES 10)
3096 * Recommended e2fsprogs version: 1.39.cfs5
3098 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3102 Frequency : occasional, depends on client load and configuration
3103 Bugzilla : 12181, 12203
3104 Description: data loss for recently-modified files
3106 Details : In some cases it is possible that recently written or created
3107 files may not be written to disk in a timely manner (this should
3108 normally be within 30s unless client IO load is very high).
3109 The problem appears as zero-length files or files that are a
3110 multiple of 1MB in size after a client crash or client eviction
3111 that are missing data at the end of the file.
3113 This problem is more likely to be hit on clients where files are
3114 repeatedly created and unlinked in the same directory, clients
3115 have a large amount of RAM, have many CPUs, the filesystem has
3116 many OSTs, the clients are rebooted frequently, and/or the files
3117 are not accessed by other nodes after being written.
3119 The presence of the problem can be detected by looking at
3120 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3121 smaller than the second (nr_unused) then dirty files will not
3122 be flushed automatically to disk. "sync; sleep 10" should be
3123 run several times on the node before unmounting it to update
3124 Lustre (this is also safe to run on nodes without this problem).
3126 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3127 that can cause this same problem, so customers using that kernel
3128 also need to update the kernel in addition to Lustre. In order
3129 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3131 It is normal that files written just before a client crash (less
3132 than 30s) may not yet have been flushed to disk, even for local
3136 Frequency : frequent on thin XT3 nodes
3138 Description: UUID collision on thin XT3 Linux nodes
3139 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3140 randomly, since we relied on an insufficiently-seeded PRNG.
3145 Description: OSS hangs after "All ost request buffers busy"
3146 Details : A deadlock between quota and journal operations caused OSS
3147 hangs after printing "All ost request buffers busy."
3150 Frequency : always on liblustre builds
3152 Description: Cleanup compiler warnings on liblustre
3155 Frequency : always on liblustre builds on XT3
3157 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3161 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3162 > 32000 subdirectories
3163 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3164 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3165 subdirectory count crosses 32000. This will aid e2fsck to
3166 correctly handle more than 32000 subdirectories.
3169 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3171 Description: System hang on startup
3172 Details : This bug allowed the liblustre (e.g. catamount) client to
3173 return to the app before handling all startup RPCs. This
3174 could leave the node unresponsive to lustre network traffic
3175 and manifested as a server ptllnd timeout.
3177 Severity : enhancement
3179 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3180 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3181 boolean to control whether to print peer debug info when a
3182 client's RPC times out.
3187 Description: make O_SYNC working on 2.6 kernels
3188 Details : 2.6 kernels use different method for mark pages for write,
3189 so need add a code to lustre for O_SYNC work.
3194 Description: Failure to close file and release space on NFS
3195 Details : Put inode details into lock acquired in ll_intent_file_open.
3196 Use mdc_intent_lock in ll_intent_open to properly
3197 detect all kind of errors unhandled by mdc_enqueue.
3202 Description: proc file read during shutdown sometimes raced obd removal,
3204 Details : Add lock to prevent obd access after proc file removal.
3207 Frequency : Only for files larger than 4GB on 32-bit clients.
3209 Description: improperly doing page alignment of locks
3210 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3211 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3216 Description: RPCs being resent when they shouldn't be
3217 Details : Some RPCs that should not be resent are being resent. This
3218 can cause inconsistencies in the RPC state machine. Do not
3219 resend such requests.
3222 Frequency : rare, only with NFS export
3224 Description: Crash on NFS re-export node
3225 Details : under very unusual load conditions an assertion is hit in
3226 ll_intent_file_open()
3229 Frequency : only if OST filesystem is corrupted
3231 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3232 Details : for a short time RPCs with bulk IO are in the replay list,
3233 but replay of bulk IOs is unimplemented. If the OST filesystem
3234 is corrupted due to disk cache incoherency and then replay is
3235 started it is possible to trip an assertion. Avoid putting
3236 committed RPCs into the replay list at all to avoid this issue.
3241 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3242 Details : Large single O_DIRECT read and write calls can fail to allocate
3243 a sufficiently large buffer to process the request. In case of
3244 allocation failure the allocation is retried with a smaller
3245 buffer and broken into smaller requests.
3247 Severity : enhancement
3249 Description: Add -o localflock option to simulate old noflock behaviour.
3250 Details : This will achieve local-only flock/fcntl locks coherentness.
3255 Description: versioning check is incomplete
3256 Details : Checking the version difference of client vs. server, report
3257 error if the gap is too big.
3262 Description: add support PG_writeback bit
3263 Details : add support for PG_writeback bit for Lustre, for more carefull
3264 work with page cache in 2.6 kernel. This also fix some deadlocks
3265 and remove hack for work O_SYNC with 2.6 kernel.
3267 Severity : enhancement
3269 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3270 Details : The uninit_groups feature works in conjunction with the kernel
3271 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3272 up the pass1 processing of e2fsck. This is a read-only feature
3273 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3274 mount filesystems that have had this feature enabled.
3276 Severity : enhancement
3278 Description: Improve multi-block allocation algorithm to avoid fragmentation
3279 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3280 allocation locality and avoid filesystem fragmentation.
3282 ------------------------------------------------------------------------------
3284 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3286 * Support for kernels:
3287 2.6.9-42.0.3.EL (RHEL 4)
3288 2.6.5-7.276 (SLES 9)
3289 2.4.21-47.0.1.EL (RHEL 3)
3290 2.6.12.6 vanilla (kernel.org)
3291 2.6.16.21-0.8 (SLES10)
3292 * Recommended e2fsprogs version: 1.39.cfs2-0
3294 * The backwards-compatible /proc/sys/portals symlink has been removed
3295 in this release. Before upgrading, please ensure that you change
3296 any configuration scripts or /etc/sysctl.conf files that access
3297 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3298 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3299 in advance of the upgrade on any system running Lustre 1.4.6 or
3300 newer, since /proc/sys/lnet was added in that version.
3301 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3305 Frequency : only when quota is used
3307 Description: avoid scanning export list for quota master
3308 Details : Change the algorithms to avoid scanning export list in order
3309 to improve the efficiency.
3312 Frequency : MDS failover only, very rarely
3314 Description: "went back in time" messages on mds failover
3315 Details : The greatest transno may be lost when the current operation
3316 finishes with an error (transno==0) and the client's last_rcvd
3317 record is over-written. Save the greatest transno in the
3318 mds_last_transno for this case.
3321 Frequency : always for specific kernels and striping counts
3323 Description: client may get "Matching packet too big" without ACL support
3324 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3325 when trying to access files in certain configurations. The
3326 clients should in fact be denied when mounting because they do
3327 not understand ACLs.
3330 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3332 Description: many clients connecting with IO in progress causes connect timeouts
3333 Details : Avoid synchronous journal commits to avoid delays caused by many
3334 clients connecting/disconnecting when bulk IO is in progress.
3335 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3336 OST_IO_PORTAL to avoid delays behind potentially many pending
3340 Frequency : occasionally with multiple writers to a single file
3342 Description: shared writes to file may result in wrong size reported by stat()
3343 Details : Allow growing of kms when extent lock is cancelled
3346 Frequency : always with random mmap IO to multi-striped file
3348 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3349 Details : Only if the hole is at the end of OST object so that kms is too
3350 small. Fix is to increase kms accordingly in ll_nopage.
3353 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3355 Description: writes to a missing object would leak memory on the OST
3356 Details : If there is an inconsistency between the MDS and OST filesystems,
3357 such that the MDS references an object that doesn't exist, writes
3358 to that object will leak memory due to incorrect cleanup in the
3359 error handling path, eventually running out of memory on the OST.
3364 Description: Creating too long symlink causes lustre errors
3365 Details : Check symlink and name lengths before sending requests to MDS.
3368 Frequency : only if flock is enabled (not on by default)
3370 Description: posix locks not released on fd closure on 2.6.9+
3371 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3372 kernels, this caused such locks not to be released on fd close and
3373 then assertions on fs unmount about still used locks.
3376 Frequency : MDS failover only, very rarely
3378 Description: clients may get ASSERTION(granted_lock != NULL)
3379 Details : When request was taking a long time, and a client was resending
3380 a getattr by name lock request. The were multiple lock requests
3381 with the same client lock handle and
3382 mds_getattr_name->fixup_handle_for_resent_request found one of the
3383 lock handles but later failed with ASSERTION(granted_lock != NULL).
3388 Description: handle->h_buffer_credits > 0, assertion failure
3389 Details : h_buffer_credits is zero after truncate, causing assertion
3390 failure. This patch extends the transaction or creates a new
3394 Frequency : NFS re-export or patchless client
3395 Bugzilla : 11179, 10796
3396 Description: Crash on NFS re-export node (__d_move)
3397 Details : We do not want to hash the dentry if we don't have a lock.
3398 But if this dentry is later used in d_move, we'd hit uninitialised
3399 list head d_hash, so we just do this to init d_hash field but
3400 leave dentry unhashed.
3403 Frequency : NFS re-export or patchless client
3405 Description: NFS exports has problem with symbolic link
3406 Details : lustre client didn't properly install dentry when re-exported
3407 to NFS or running patchless client.
3410 Frequency : NFS re-export or patchless client
3412 Description: Various nfs/patchless fixes.
3413 Details : fixes reuse disconected alias for lookup process - this fixes
3414 warning "find_exported_dentry: npd != pd",
3415 fix permission error with open files at nfs.
3416 fix apply umask when do revalidate.
3419 Frequency : occasional
3421 Description: Crash on NFS re-export node
3422 Details : calling clear_page() on the wrong pointer triggered oops in
3423 generic_mapping_read().
3426 Frequency : rarely, using O_DIRECT IO
3428 Description: unaligned directio crashes client with LASSERT
3429 Details : check for unaligned buffers before trying any requests.
3432 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3434 Description: stale data returned from RAID cache
3435 Details : If only a small amount of IO is done to the RAID device before
3436 reading it again it is possible to get stale data from the RAID
3437 cache instead of reading it from disk.
3440 Frequency : always for sles10 kernel
3442 Description: sles10 support
3443 Details : ll_follow_link: compile fixes and using of nd_set_link
3444 under newer kernels.
3447 Frequency : depends on arch, kernel and compiler version, always on sles10
3450 Description: recursive or deep enough symlinks cause stack overflow
3451 Details : getting rid of large stack-allocated variable in
3455 Frequency : depends on hardware
3457 Description: lustre write performance loss in the SLES10 kernel
3458 Details : the performance loss is caused by using of write barriers in the
3459 ext3 code. The SLES10 kernel turns barrier support on by
3460 default. The fix is to undo that change for ldiskfs.
3462 ------------------------------------------------------------------------------
3464 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3466 * Support for kernels:
3467 2.6.9-42.0.3EL (RHEL 4)
3468 2.6.5-7.276 (SLES 9)
3469 2.4.21-47.0.1.EL (RHEL 3)
3470 2.6.12.6 vanilla (kernel.org)
3474 Frequency : quota enabled and large files being deleted
3476 Description: releasing more than 4GB of quota at once hangs OST
3477 Details : If a user deletes more than 4GB of files on a single OST it
3478 will cause the OST to spin in an infinite loop. Release
3479 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3484 Description: statfs data retrieved from /proc may be stale or zero
3485 Details : When reading per-device statfs data from /proc, in the
3486 {kbytes,files}_{total,free,avail} files, it may appear
3487 as zero or be out of date.
3490 Frequency : systems with MD RAID1 external journal devices
3492 Description: lconf's call to blkid is confused by RAID1 journal devices
3493 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3494 of returning all block devices that match the journal UUID.
3497 Frequency : always, for aggregate stripe size over 4GB
3499 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3500 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3501 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3504 Frequency : always if "lfs find" used on a local file/directory
3506 Description: "lfs find" segfaults if used on a local file/directory
3507 Details : The case where a directory component was not specified wasn't
3508 handled correctly. Handle this properly.
3511 Frequency : always on ppc64
3513 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3514 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3515 perform endianness conversion.
3518 Frequency : rarely (truncate to non-zero file size after write under load)
3519 Bugzilla : 10730, 10687
3520 Description: Files padded with zeros to next 4K multiple
3521 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3522 it is possible that files that are truncated to a non-zero size
3523 immediately after being written are filled with zero bytes beyond
3524 the truncated size. No file data is lost.
3526 Severity : enhancement
3528 Description: Allow recovery/failover for liblustre clients.
3529 Details : liblustre clients were unaware of failover configurations until
3532 Severity : enhancement
3534 Description: user file locks should fail when not mounting with flock option
3535 Details : Set up an error-returning stub in ll_file_operations.lock field
3536 to prevent incorrect behaviour when client is mounted without
3537 flock option. Also, set up properly f_op->flock field for
3541 Frequency : always on ia64
3543 Description: "lfs df" loops on printing out MDS statfs information
3544 Details : The obd_ioctl_data was not initialized and in some systems
3545 this caused a failure during the ioctl that did not return
3546 an error. Initialize the struct and return an error on failure.
3549 Frequency : SLES 9 only
3551 Description: Error of copying files with lustre special EAs as root
3552 Details : Client side always return success for setxattr call for lustre
3553 special xattr (currently only "trusted.lov").
3556 Frequency : rarely on clusters with both ia64+i386 clients
3558 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3559 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3560 clients are doing concurrent writes to the same file it is
3561 possible that the ia64 clients may LASSERT because the OST
3562 extent locks are not PAGE_SIZE aligned. Ensure that grown
3563 locks are always aligned on the request boundary.
3566 Frequency : specific use, occasional
3568 Description: Overwriting in use executable truncates on-disk binary image
3569 Details : If one node attempts to overwrite an executable in use by
3570 another node, we now correctly return ETXTBSY instead of
3571 truncating the file.
3573 Severity : enhancement
3575 Description: Async OSC create to avoid the blocking unnecessarily.
3576 Details : If a OST has no remain object, system will block on the creating
3577 when need to create a new object on this OST. Now, ways use
3578 pre-created objects when available, instead of blocking on an
3579 empty osc while others are not empty. If we must block, we block
3580 for the shortest possible period of time.
3585 Description: chmod on Lustre root is propagated to other clients
3586 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3587 invalid by the follow_mount time.
3590 Frequency : liblustre clients only
3592 Description: Race in 'instant cancel' lock handling could lead to such locks
3593 never to be granted in case of SMP MDS
3594 Details : Do not destroy not yet granted but cbpending locks in
3598 Frequency : replay/resend of open
3600 Description: non null lock assetion failure in mds_intent_policy
3601 Details : Trying to replay/resend lockless open requests resulted in
3602 mds_open() returning 0 with no lock. Now it sets a flag if
3603 a lock is going to be returned.
3605 Severity : enhancement
3607 Description: Checksum enhancements
3608 Details : New checksum enhancements allow for resending RPCs that failed
3611 Severity : enhancement
3613 Description: Tunables on number of dirty pages in cacche
3614 Details : Allow to set limit on number of dirty pages cached.
3619 Description: client crash on unmount - lock still has references
3620 Details : In some error handling cases it was possible to leak a lock
3621 reference on a client while accessing a file. This was not
3622 harmful to the client during operation, but would cause the
3623 client to crash when the filesystem is unmounted.
3626 Frequency : specific case, rare
3628 Description: ETXTBSY on mds though file not in use by client
3629 Details : ETXTBSY is no longer incorrectly returned when attempting to
3630 chmod or chown a directory that the user previously tried to
3631 execute or a currently-executing binary.
3634 Frequency : extremely rare except on liblustre-based clients
3636 Description: Lustre space not freed when files are deleted
3637 Details : Clean up open-unlinked files after client eviction. Previously
3638 the unlink was skipped and the files remained as orphans.
3643 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3644 Details : In some cases it was possible to send a blocking callback to
3645 a client doing a glimpse, even though that client didn't get
3646 a lock granted. When the glimpse lock is cancelled on the OST
3647 the freed lock is left on the waiting list and corrupted the list.
3650 Frequency : all core dumps
3652 Description: Broke core dumps to lustre
3653 Details : Negative dentry may be unhashed if parent does not have UPDATE
3654 lock, but some callers, e.g. do_coredump, expect dentry to be
3655 hashed after successful create, hash it in ll_create_it.
3657 ------------------------------------------------------------------------------
3659 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3661 * Support for kernels:
3662 2.6.9-42.0.2.EL (RHEL 4)
3663 2.6.5-7.276 (SLES 9)
3664 2.4.21-40.EL (RHEL 3)
3665 2.6.12.6 vanilla (kernel.org)
3669 Frequency : always on RHEL 3
3671 Description: Number of open files grows over time
3672 Details : The number of open files grows over time, whether or not
3673 Lustre is started. This was due to a filp leak introduced
3674 by one of our kernel patches.
3676 ------------------------------------------------------------------------------
3678 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3680 * Support for kernels:
3681 2.6.9-42.EL (RHEL 4)
3682 2.6.5-7.276 (SLES 9)
3683 2.4.21-40.EL (RHEL 3)
3684 2.6.12.6 vanilla (kernel.org)
3689 Bugzilla : 5719, 9635, 9792, 9684
3690 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3691 Details : If a server is under heavy load and cannot reply to new
3692 connection requests before the client resends the (re)connect,
3693 the connection handling code can behave badly if two service
3694 threads are concurrently handing separate (re)connections from
3695 the same client. Add better locking to the connection handling
3696 code, and ensure that only a single connection will be processed
3697 for a given client UUID, even if the lock is dropped.
3699 Severity : enhancement
3701 Description: add TCP zero-copy support to kernel
3702 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3703 sends if the hardware supports scatter-gather and checksumming.
3704 This allows socklnd to do client-write and server-read more
3705 efficiently and reduce CPU utilization from skbuf copying.
3708 Frequency : only if NFS exporting from client
3710 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3711 Details : If setting attributes on a file created under NFS that had
3712 never been opened it would be possible to oops the client
3713 if the file had no objects.
3716 Frequency : always for liblustre
3718 Description: liblustre client does MDS+OSTs setattr RPC for each write
3719 Details : When doing a write from a liblustre client, the client
3720 incorrectly issued an RPC to the MDS and each OST the file was
3721 striped over in order to update the timestamps. When writing
3722 with small chunks and many clients this could overwhelm the MDS
3723 with RPCs. In all cases it would slow down the write because
3724 these RPCs are unnecessary.
3726 Severity : enhancement
3728 Description: allow number of MDS service threads to be changed at module load
3729 Details : It is now possible to change the number of MDS service threads
3730 running. Adding "options mds mds_num_threads={N}" to the MDS's
3731 /etc/modprobe.conf will set the number of threads for the next
3732 time Lustre is restarted (assuming the "mds" module is also
3733 reloaded at that time). The default number of threads will
3734 stay the same, 32 for most systems.
3739 Description: OST crash if filesystem is unformatted or corrupt
3740 Details : If an OST is started on a device that has never been formatted
3741 or if the filesystem is corrupt and cannot even mount then the
3742 error handling cleanup routines would dereference a NULL pointer.
3747 Description: NULL pointer deref in llap_from_page.
3748 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3749 incorrect) mapping if the page was truncated/reclaimed while it was
3750 searched for. Check for this condition and skip such pages when
3751 doing readahead. Introduce extra check to llap_from_page() to
3752 verify page->mapping->host is non-NULL (so page is not anonymous).
3755 Frequency : Sometimes when using sys_sendfile
3757 Description: "page not covered by a lock" warnings from ll_readpage
3758 Details : sendfile called ll_readpage without right page locks present.
3759 Now we introduced ll_file_sendfile that does necessary locking
3760 around call to generic_file_sendfile() much like we do in
3764 Frequency : with certain MDS communication failures at client mount time
3766 Description: NULL pointer deref after failed client mount
3767 Details : a client connection request may delayed by the network layer
3768 and not be sent until after the PTLRPC layer has timed out the
3769 request. If the client fails the mount immediately it will try
3770 to clean up before the network times out the request. Add a
3771 reference from the request import to the obd device and delay
3772 the cleanup until the network drops the request.
3775 Frequency : occasionally during client (re)connect
3777 Description: assertion failure during client (re)connect
3778 Details : processing a client connection request may be delayed by the
3779 client or server longer than the client connect timeout. This
3780 causes the client to resend the connection request. If the
3781 original connection request is replied in this interval, the
3782 client may trip an assertion failure in ptlrpc_connect_interpret()
3783 which thought it would be the only running connect process.
3786 Frequency : only with obd_echo servers and clients that are rebooted
3788 Description: kernel BUG accessing uninitialized data structure
3789 Details : When running an obd_echo server it did not start the ping_evictor
3790 thread, and when a client was evicted an uninitialized data
3791 structure was accessed. Start the ping_evictor in the RPC
3792 service startup instead of the OBD startup.
3794 Severity : enhancement
3795 Bugzilla : 10193 (patchless)
3796 Description: Remove dependency on various unexported kernel interfaces.
3797 Details : No longer need reparent_to_init, exit_mm, exit_files,
3798 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3801 Frequency : rare (only users of deprecated and unsupported LDAP config)
3803 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3804 Details : LDAP apparently contains 'lustreName' attributes instead of
3805 'name'. A simple remapping of the name is sufficient.
3808 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3810 Description: waiting_locks_callback trips kernel BUG if client is evicted
3811 Details : Running with the dump_on_timeout debug flag turned on makes
3812 it possible that the waiting_locks_callback() can try to dump
3813 the Lustre kernel debug logs from an interrupt handler. Defer
3814 this log dumping to the expired_lock_main() thread.
3816 Severity : enhancement
3818 Description: Support NFS exporting on 2.6 kernels.
3819 Details : Implement non-rawops metadata methods for NFS server to use without
3820 changing NFS server code.
3823 Frequency : very rare (synthetic metadata workload only)
3825 Description: two racing renames might cause an MDS thread to deadlock
3826 Details : Running the "racer" program may cause one MDS thread to rename
3827 a file from being the source of a rename to being the target of
3828 a rename at exactly the same time that another thread is doing
3829 so, and the second thread has already enqueued these locks after
3830 doing a lookup of the target and is trying to relock them in
3831 order. Ensure that we don't try to re-lock the same resource.
3834 Frequency : only very large systems with liblustre clients
3836 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3837 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3838 all OSTs in parallel. This allows the eviction of stale liblustre
3839 clients to proceed much faster than if they were done in series,
3840 and also offers similar improvements for other set_info RPCs.
3845 Description: excessive CPU usage during initial read phase on client
3846 Details : During the initial read phase on a client, it would agressively
3847 retry readahead on the file, consuming too much CPU and impacting
3848 performance (since 1.4.5.8). Improve the readahead algorithm
3849 to avoid this, and also improve some other common cases (read
3850 of small files in particular, where "small" is files smaller than
3851 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3856 Description: MDS crash when receiving packet with unknown intent.
3857 Details : Do not LBUG in unknown intent case, just return -EFAULT
3859 Severity : enhancement
3860 Bugzilla : 9293, 9385
3861 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3862 Details : Do not serialize getattr (non-intent version) and statfs.
3865 Frequency : occasional, when OST network is overloaded/intermittent
3867 Description: client evicted by OST after bulk IO timeout
3868 Details : If a client sends a bulk IO request (read or write) the OST
3869 may evict the client if it is unresposive to its data GET/PUT
3870 request. This is incorrect if the network is overloaded (takes
3871 too long to transfer the RPC data) or dropped the OST GET/PUT
3872 request. There is no need to evict the client at all, since
3873 the pinger and/or lock callbacks will handle this, and the
3874 client can restart the bulk request.
3877 Frequency : Always when mmapping file with no objects
3879 Description: client crashes when mmapping file with no objects
3880 Details : Check that we actually have objects in a file before doing any
3881 operations on objects in ll_vm_open, ll_vm_close and
3887 Description: Request leak when working with deleted CWD
3888 Details : Introduce advanced request refcount tracking for requests
3889 referenced from lustre intent.
3891 Severity : Enhancement
3893 Description: Cache open file handles on client.
3894 Details : MDS now will return special lock along with openhandle, if
3895 requested and client is allowed to hold openhandle, even if unused,
3896 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3897 closing files for every read/write openration.
3899 Severity : Enhancement
3901 Description: Cache open negative dentries on client when possible.
3902 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3903 negative dentries on lock revocation.
3908 Description: Remounting a client read-only wasn't possible with a zconf mount
3909 Details : It wasn't possible to remount a client read-only with llmount.
3911 Severity : enhancement
3912 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3913 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3914 MPICH is included to add Lustre-specific ADIO interfaces.
3915 This is based closely on the UFS ADIO layer and only differs
3916 in file creation, in order to allow the OST striping to be set.
3917 This is user-contributed code and not supported by CFS.
3922 Description: extended inode attributes (immutable, append-only) work improperly
3923 when 2.4 and 2.6 kernels are used on client/server or vice versa
3924 Details : Introduce kernel-independent values for these flags.
3926 Severity : enhancement
3929 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3930 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3931 granularity. Now, Fractional MB granularity is be supported,
3932 this is very useful for low memory system.
3934 Severity : enhancement
3936 Description: Getattr by fid
3937 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3938 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3939 have locks and avoids getattr rpc requests for it on every stat.
3942 Frequency : Always, for filesystems larger than 2TB
3944 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3945 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3946 filesystems larger than 2TB to be mounted with mballoc enabled.
3949 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3951 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3952 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3953 filesystems larger than 2TB is mounted then the kernel will
3954 silently corrupt the start of the filesystem. CONFIG_LBD is
3955 enabled for all CFS-supported kernels, but the possibility of
3956 this happening with a modified kernel config exists.
3958 Severity : enhancement
3960 Description: add client O_DIRECT support for 2.6 kernels
3961 Details : It is now possible to do O_DIRECT reads and writes to files
3962 in the Lustre client mountpoint on 2.6 kernel clients.
3964 Severity : enhancement
3966 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3967 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3968 parallel, not waiting for response from every OST before sending
3969 a rpc to the next OST.
3974 Description: setattr vs write race when updating file timestamps
3975 Details : Client processes that update a file timestamp into the past
3976 right after writing to the file (e.g. tar) it is possible that
3977 the updated file modification time can be reset to the current
3978 time due to a race between processing the setattr and write RPC.
3980 Severity : enhancement
3982 Description: Bring 'lfs find' closer in line with regular Linux find.
3983 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3984 -print0 options and obtains all the needed info through the lustre
3987 Severity : enhancement
3989 Description: support up to 1024 configured devices on one node
3990 Details : change obd_dev array from statically allocated to dynamically
3991 allocated structs as they are first used to reduce memory usage
3996 Description: Flush dirty partially truncated pages during truncate
3997 Details : Immediatelly flush partially truncated pages in filter_setattr,
3998 this way we completely avoid having any pages in page cache on OST
3999 and can retire ugly workarounds during writes to flush such pages.
4004 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4005 Details : protect inode from truncation within vfs_unlink() context
4006 just take a reference before calling vfs_unlink() and release it
4007 when parent's i_sem is free.
4012 Description: last_id value checked outside lock on OST caused LASSERT failure
4013 Details : If there were multiple MDS->OST object precreate requests in
4014 flight, it was possible that the OST's last object id was checked
4015 outside a lock and incorrectly tripped an assertion. Move checks
4016 inside locks, and discard old precreate requests.
4019 Frequency : always, if extents are used on OSTs
4021 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4022 Details : OSTs using the extents format would not zero the high 16 bits of
4023 the index physical block number. This is not a problem for any
4024 OST filesystems smaller than 16TB, and no kernels support ext3
4025 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4026 new/modified files) and can be fixed for existing filesystems
4027 with e2fsprogs-1.39-cfs1.
4032 Description: import connection selection may be incorrect if timer wraps
4033 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4034 import connections to be ignored if the 32-bit jiffies counter
4035 wraps. Use a 64-bit jiffies counter.
4038 Frequency : very large clusters immediately after boot
4040 Description: LNET request buffers exhausted under heavy short-term load
4041 Details : If a large number of client requests are generated on a service
4042 that has previously never seen so many requests it is possible
4043 that the request buffer growth cannot keep up with the spike in
4044 demand. Instead of dropping incoming requests, they are held in
4045 the LND until the RPC service can accept more requests.
4048 Frequency : Sometimes during replay
4050 Description: Assertion failure in ll_local_open after replay.
4051 Details : If replay happened on an open request reply before we were able
4052 to set replay handler, reply will become not swabbed tripping the
4053 assertion in ll_local_open. Now we set the handler right after
4054 recognising of open request
4057 Frequency : very rare
4059 Description: kernel reports "badness in vsnprintf"
4060 Details : Reading from the "recovery_status" /proc file in small chunks
4061 may cause a negative length in lprocfs_obd_rd_recovery_status()
4062 call to vsnprintf() (which is otherwise harmless). Exit early
4063 if there is no more space in the output buffer.
4065 Severity : enhancement
4067 Description: clear OBD RPC statistics by writing to them
4068 Details : It is now possible to clear the OBD RPC statistics by writing
4069 to the "stats" file.
4074 Description: Client mtime is not the same on different clients after utimes
4075 Details : In some cases, the client was using the utimes() syscall on
4076 a file cached on another node. The clients now validate the
4077 ctime from the MDS + OSTs to determine which one is right.
4082 Description: Inability to activate failout mode
4083 Details : lconf script incorrectly assumed that in pythong string's numeric
4084 value is used in comparisons.
4087 Frequency : always with multiple stripes per file
4089 Description: Inefficient object allocation for mutli-stripe files
4090 Details : When selecting which OSTs to stripe files over, for files with
4091 a stripe count that divides evenly into the number of OSTs,
4092 the MDS is always picking the same starting OST for each file.
4093 Return the OST selection heuristic to the original design.
4098 Description: mount failures may take full timeout to return an error
4099 Details : Under some heavy load conditions it is possible that a
4100 failed mount can wait for the full obd_timeout interval,
4101 possibly several minutes, before reporting an error.
4102 Instead return an error as soon as the status is known.
4104 Frequency : quota enabled and large files being deleted
4106 Description: releasing more than 4GB of quota at once hangs OST
4107 Details : If a user deletes more than 4GB of files on a single OST it
4108 will cause the OST to spin in an infinite loop. Release
4109 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4114 Description: statfs data retrieved from /proc may be stale or zero
4115 Details : When reading per-device statfs data from /proc, in the
4116 {kbytes,files}_{total,free,avail} files, it may appear
4117 as zero or be out of date.
4120 Frequency : systems with MD RAID1 external journal devices
4122 Description: lconf's call to blkid is confused by RAID1 journal devices
4123 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4124 of returning all block devices that match the journal UUID.
4127 Frequency : always, for aggregate stripe size over 4GB
4129 Description: assertion fails when trying to use 4GB stripe size
4130 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4131 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4134 Frequency : always on ppc64
4136 Description: the first write on an ext3 filesystem with mballoc got stuck
4137 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4138 perform endianness conversion.
4140 ------------------------------------------------------------------------------
4142 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4144 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4145 INTEROPERATE with older versions automatically. Please read the
4146 user documentation before upgrading any part of a live system.
4147 * WARNING: Lustre networking configuration changes are required with
4148 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4151 * Support for kernels:
4152 2.6.9-22.0.2.EL (RHEL 4)
4153 2.6.5-7.244 (SLES 9)
4154 2.6.12.6 vanilla (kernel.org)
4157 Severity : enhancement
4158 Bugzilla : 7981/8208
4159 Description: Introduced Lustre Networking (LNET)
4160 Details : LNET is new networking infrastructure for Lustre, it includes
4161 a reorganized network configuration mode (see the user
4162 documentation for full details) as well as support for routing
4163 between different network fabrics. Lustre Networking Devices
4164 (LNDs) for the supported network fabrics have also been
4165 created for this new infrastructure.
4167 Severity : enhancement
4168 Description: Introduced Access control lists
4169 Details : clients can set ACLs on files and directories in order to have
4170 more fine-grained permissions than the standard Unix UGO+RWX.
4171 The MDS must be started with the "-o acl" mount option.
4173 Severity : enhancement
4174 Description: Introduced filesystem quotas
4175 Details : Administrators may now establish per-user quotas on the
4178 Severity : enhancement
4180 Description: Configuration change for the XT3
4181 The PTLLND is now used to run Lustre over Portals on the XT3
4182 The configure option(s) --with-cray-portals are no longer used.
4183 Rather --with-portals=<path-to-portals-includes> is used to
4184 enable building on the XT3. In addition to enable XT3 specific
4185 features the option --enable-cray-xt3 must be used.
4190 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4191 Details : A race between allocating a new llcd and re-getting the llcd_lock
4192 allowed another thread to grab newly-allocated llcd.
4194 Severity : enhancement
4196 Description: 2.6 OST async journal commit and locking fix to improve performance
4197 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4198 2.6 kernels are now async as they already were in 2.4 kernels so
4199 that they can commit concurrently with the network bulk transfer.
4200 For block-allocated files the filter allocation semaphore is held
4201 to avoid filesystem fragmentation during allocation. BKL lock
4202 removed for 2.6 xattr operations where it is no longer needed.
4207 Description: lconf incorrectly determined whether two IP networks could talk
4208 Details : In some more complicated routing and multiple-network
4209 configurations, lconf will avoid trying to make a network
4210 connection to a disjoint part of the IP space. It was doing the
4211 math incorrectly for one set of cases.
4216 Description: Fix for potential infinite loop processing records in an llog.
4217 Details : If an llog record is corrupted/zeroed, it is possible to loop
4218 forever in llog_process(). Validate the llog record length
4219 and skip the remainder of the block on error.
4222 Frequency : occasional (liblustre only)
4224 Description: liblustre could not open files whose last component is a symlink
4225 Details : sysio_path_walk() would incorrectly pass the open intent to
4226 intermediate path components.
4229 Frequency : rare (liblustre only with non-standard tuning)
4230 Bugzilla : 7201 (7350)
4231 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4232 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4233 after it has been released, only do this for referenced locks
4235 Severity : enhancement
4237 Description: specifying an (invalid) directory default stripe_size of -1
4238 would reset the directory default striping
4239 Details : stripe_size -1 was used internally to signal directory stripe
4240 removal, now use "all default" to signal dir stripe removal
4241 as a directory striping of "all default" is not useful
4244 Frequency : common for large clusters running liblustre clients
4246 Description: doing an ls when liblustre clients are running is slow
4247 Details : sending a glimpse AST to a liblustre client waits for every AST
4248 to time out, as liblustre clients will not respond. Since they
4249 cannot cache data we refresh the OST lock LVB from disk instead.
4251 Severity : enhancement
4253 Description: doing an ls at the same time as file IO can be slow
4254 Details : enqueue and other "small" requests can be blocked behind many
4255 large IO requests. Create a new OST IO portal for non-IO
4256 requests so they can be processed faster.
4259 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4261 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4262 created file with mode 000
4263 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4264 allow an ftruncate() as a non-root user to a file with mode 000.
4265 Lustre now respects this flag to disable mode checks when
4266 truncating a file owned by the user
4269 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4272 Description: Revoking locks from clients that went dead or catatonic might take
4274 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4275 cancellation of such locks instant on servers without waiting for
4276 any reply from clients. Clients drops these locks when cancel
4277 notification from server is received without replying.
4280 Frequency : liblustre-only, when liblustre client dies or becomes busy
4282 Description: Doing ls on Linux clients can take a long time with active
4284 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4285 granting such locks to it in the first place if possible. Locks
4286 are taken by proxy on the OST during the read or write and
4287 dropped immediately afterward. Add connect flags handling, do
4288 not grant locks to liblustre clients for glimpse ASTs.
4290 Severity : enhancement
4292 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4293 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4294 sequential reads and behaves badly with applications doing random
4295 reads. Improve it by reading ahead at least read region, and
4296 avoiding excessive large RPC for small reads.
4298 Severity : enhancement
4300 Description: Creating more than 1000 files for a single job may cause a load
4301 imbalance on the OSTs if there are also a large number of OSTs.
4302 Details : qos_prep_create() uses an OST index reseed value that is an
4303 even multiple of the number of available OSTs so that if the
4304 reseed happens in the middle of the object allocation it will
4305 still utilize the OSTs as uniformly as possible.
4310 Description: OST or MDS may oops in ping_evictor_main()
4311 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4312 but doesn't restart at beginning of obd_exports_timed list
4315 Severity : enhancement
4317 Description: improve by-nid export eviction on the MDS and OST
4318 Details : allow multiple exports with the same NID to be evicted at one
4319 time without re-searching the exports list.
4322 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4324 Description: MDS may oops in groups_free()
4325 Details : in rare race conditions a newly allocated group_info struct is
4326 freed again, and this can be NULL. The 2.4 compatibility code
4327 for groups_free() checked for a NULL pointer, but 2.6 did not.
4330 Frequency : common for liblustre clients doing little filesystem IO
4331 Bugzilla : 9352, 7313
4332 Description: server may evict liblustre clients accessing contended locks
4333 Details : if a client is granted a lock or receives a completion AST
4334 with a blocking AST already set it would not reply to the AST
4335 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4338 Frequency : lfs setstripe, only systems with more than 160 OSTs
4340 Description: unable to set striping with a starting offset beyond OST 160
4341 Details : llapi_create_file() incorrectly limited the starting stripe
4342 index to the maximum single-file stripe count.
4345 Frequency : LDAP users only
4347 Description: lconf did not handle in-kernel recovery with LDAP properly
4348 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4350 Severity : enhancement
4352 Description: bind OST threads to NUMA nodes to improve performance
4353 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4354 node and do their allocations there to localize memory access
4356 Severity : enhancement
4358 Description: llmount can determine client NID directly from Myrinet (GM)
4359 Details : the client NID code from gmnalnid was moved directly into
4360 llmount, removing the need to use this or specifying the
4361 client NID explicitly when mounting GM clients with zeroconf
4364 Frequency : if client is started with down MDS
4366 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4367 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4368 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4369 ptlrpc_import_delayed_req() because we didn't check if the
4370 request was interrupted, and we also didn't break out of the
4371 event loop if there was no timeout
4376 Description: data loss during non-page-aligned writes to a single file from
4377 both multiple nodes and multiple threads on one node at same time
4378 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4379 inconsistency led to false short-reads, that were cached and later
4380 used by ->prepare_write() to fill in partially written page,
4381 leading to data loss.
4384 Frequency : always, if lconf --abort_recovery used
4386 Description: lconf --abort_recovery fails with 'Operation not supported'
4387 Details : lconf was attempting to abort recovery on the MDT device and not
4390 Severity : enhancement
4392 Description: remove cleanup logs
4393 Details : replace lconf-generated cleanup logs with lustre internal
4394 cleanup routines. Eliminates the need for client-cleanup and
4397 Severity : enhancement
4399 Description: add support for EAs (user and system) on lustre filesystems
4400 Details : it is now possible to store extended attributes in the Lustre
4401 client filesystem, and with the user_xattr mount option it
4402 is possible to allow users to store EAs on their files also
4404 Severity : enhancement
4406 Description: Add possibility (config option) to show minimal available OST free
4408 Details : When compiled with --enable-mindf configure option, statfs(2)
4409 (and so, df) will return least minimal free space available from
4410 all OSTs as amount of free space on FS, instead of summary of
4411 free spaces of all OSTs.
4413 Severity : enhancement
4415 Description: do not expand extent locks acquired on OST-side
4416 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4417 by server: they are not cached anyway.
4420 Frequency : when mmap is used/binaries executed from Lustre
4422 Description: Unmmap pages before throwing them away from read cache.
4423 Details : llap_shrink cache now attempts to unmap pages before discarding
4424 them (if unmapping failed - do not discard). SLES9 kernel has
4425 extra checks that trigger if this unmapping is not done first.
4430 Description: lconf didn't resolve symlinks before checking to see whether a
4431 given mountpoint was already in use
4434 Frequency : when migrating failover services
4435 Bugzilla : 6395, 9514
4436 Description: When migrating a subset of services from a node (e.g. failback
4437 from a failover service node) the remaining services would
4438 time out and evict clients.
4439 Details : lconf --force (implied by --failover) sets the global obd_timeout
4440 to 5 seconds in order to quickly disconnect, but this caused
4441 other RPCs to time out too quickly. Do not change the global
4442 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4444 Severity : enhancement
4445 Frequency : if MDS is started with down OST
4446 Bugzilla : 9439,5706
4447 Description: Allow startup/shutdown of an MDS without depending on the
4448 availability of the OSTs.
4449 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4450 Add appropriate locking and lov-osc refcounts for safe
4451 cleaning. Add osc abort_inflight calls in case the
4452 synchronize never started.
4455 Frequency : occasional (Cray XT3 only)
4457 Description: root not authorized to access files in CRAY_PORTALS environment
4458 Details : The client process capabilities were not honoured on the MDS in
4459 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4460 been accessed by an authorized user then root was able to access
4461 the file on the local client also. The root user capabilities
4462 are now allowed on the MDS, as this environment has secure UID.
4465 Frequency : occasional
4467 Description: ldiskfs "too long searching" message happens too often
4468 Details : A debugging message (otherwise harmless) prints too often on
4469 the OST console. This has been reduced to only happen when
4470 there are fragmentation problems on the filesystem.
4475 Description: Division by zero in statfs when all OSCs are inactive
4476 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4477 lov_statfs divides by value returned by lov_get_stripecnt().
4481 Bugzilla : 9489, 3273
4482 Description: First write from each client to each OST was only 4kB in size,
4483 to initialize client writeback cache, which caused sub-optimal
4484 RPCs and poor layout on disk for the first writen file.
4485 Details : Clients now request an initial cache grant at (re)connect time
4486 and so that they can start streaming writes to the cache right
4487 away and always do full-sized RPCs if there is enough data.
4488 If the OST is rebooted the client also re-establishes its grant
4489 so that client cached writes will be honoured under the grant.
4494 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4495 Details : Now I/O RPCs go to different portal number and (presumably) fast
4496 lock requests (and glimses) and other RPCs get their own service
4497 threads pool that should be able to service those RPCs
4500 Severity : enhancement
4502 Description: Ability to exchange lustre version between client and servers and
4503 issue warnings at client side if client is too old. Also for
4504 liblustre clients there is ability to refuse connection of too old
4506 Details : New 'version' field is added to connect data structure that is
4507 filled with version info. That info is later checked by server and
4511 Frequency : rare, liblustre only.
4512 Bugzilla : 9296, 9581
4513 Description: Two simultaneous writes from liblustre at offset within same page
4514 might proceed at the same time overwriting eachother with stale
4516 Details : I/O lock withing llu_file_prwv was released too early, before data
4517 actually was hitting the wire. Extended lock-holding time until
4518 server acknowledges receiving data.
4521 Frequency : extremely rare. Never observed in practice.
4523 Description: avoid generating lustre_handle cookie of 0.
4524 Details : class_handle_hash() generates handle cookies by incrementing
4525 global counter, and can hit 0 occasionaly (this is unlikely, but
4526 not impossible, because initial value of cookie counter is
4527 selected randonly). Value of 0 is used as a sentinel meaning
4528 "unassigned handle" --- avoid it. Also coalesce two critical
4529 sections in this function into one.
4531 Severity : enhancement
4533 Description: allow liblustre clients to delegate truncate locking to OST
4534 Details : To avoid overhead of locking, liblustre client instructs OST to
4535 take extent lock in ost_punch() on client's behalf. New connection
4536 flag is added to handle backward compatibility.
4538 Severity : enhancement
4539 Bugzilla : 4928, 7341, 9758
4540 Description: allow number of OST service threads to be specified
4541 Details : a module parameter allows the number of OST service threads
4542 to be specified via "options ost ost_num_threads={N}" in the
4543 OSS's /etc/modules.conf or /etc/modprobe.conf.
4547 Bugzilla : 6146, 9635, 9895
4548 Description: servers crash with bad pointer in target_handle_connect()
4549 Details : In rare cases when a client is reconnecting it was possible that
4550 the connection request was the last reference for that export.
4551 We would temporarily drop the export reference and get a new
4552 one, but this may have been the last reference and the export
4553 was just destroyed. Get new reference before dropping old one.
4555 Severity : enhancement
4556 Frequency : if client is started with failover MDS
4558 Description: Allow multiple MDS hostnames in the mount command
4559 Details : Try to read the configuration from all specified MDS
4560 hostnames during a client mount in case the "primary"
4563 Severity : enhancement
4565 Description: Stop sending data to evicted clients as soon as possible.
4566 Details : Check if the client we are about to send or are sending data to
4567 was evicted already. (Check is done every second of waiting,
4568 for which l_wait_event interface was extended to allow checking
4569 of exit condition at specified intervals).
4572 Frequency : rare, normally only when NFS exporting is done from client
4574 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4576 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4579 Frequency : always, if srand() is called before liblustre initialization
4581 Description: Liblustre uses system PRNG disturbing its usage by user application
4582 Details : Introduce internal to lustre fast and high-quality PRNG for
4583 lustre usage and make liblustre and some other places in generic
4584 lustre code to use it.
4586 Severity : enhancement
4587 Bugzilla : 9477, 9557, 9870
4588 Description: Verify that the MDS configuration logs are updated when xml is
4589 Details : Check if the .xml configuration logs are newer than the config
4590 logs stored on the MDS and report an error if this is the case.
4591 Request --write-conf, or allow starting with --old_conf.
4593 Severity : enhancement
4595 Description: Handle symlinks in the path when checking if Lustre is mounted.
4596 Details : Resolve intermediate symlinks when checking if a client has
4597 mounted a filesystem to avoid duplicate client mounts.
4602 Description: lconf can hit an error exception but still return success.
4603 Details : The lconf command catches the Command error exception at the top
4604 level script context and will exit with the associated exit
4605 status, but doesn't ensure that this exit status is non-zero.
4610 Description: failure of ptlrpc thread startup can cause oops
4611 Details : Starting a ptlrpc service thread can fail if there are a large
4612 number of threads or the server memory is very fragmented.
4613 Handle this without oopsing.
4616 Frequency : always, only if liblustre and non-default acceptor port was used
4618 Description: liblustre cannot connect to servers with non-default acceptor port
4619 Details : tcpnal_set_default_params() was not called and was therefore
4620 ignoring the environment varaible TCPNAL_PORT, as well as other
4621 TCPNAL_ environment variables
4626 Description: two objects could be created on the same OST for a single file
4627 Details : If an OST is down, in some cases it was possible to create two
4628 objects on a single OST for a single file. No problems other
4629 than potential performance impact and spurious error messages.
4633 Bugzilla : 5681, 9562
4634 Description: Client may oops in ll_unhash_aliases
4635 Details : Client dcache may become inconsistent in race condition.
4636 In some cases "getcwd" can fail if the current directory is
4642 Description: Inode refcounting problems in NFS export code
4643 Details : link_raw functions used to call d_instantiate without obtaining
4644 extra inode reference first.
4648 Bugzilla : 9942, 9903
4649 Description: Referencing freed requests leading to crash, memleaks with NFS.
4650 Details : We used to require that call to ll_revalidate_it was always
4651 followed by ll_lookup_it. Also with revalidate_special() it is
4652 possible to call ll_revalidate_it() twice for the same dentry
4653 even if first occurence returned success. This fix changes semantic
4654 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4655 reference on a request referred from the intent.
4656 ll_intent_release() then releases such a request.
4659 Frequency : rare, normally benchmark loads only
4661 Description: unlinked inodes were kept in memory on the client
4662 Details : If a client is repeatedly creating and unlinking files it
4663 can accumulate a lot of stale inodes in the inode slab cache.
4664 If there is no other client load running this can cause the
4665 client node to run out of memory. Instead flush old inodes
4666 from client cache that have the same inode number as a new inode.
4669 Frequency : SLES9 2.6.5 kernel and long filenames only
4670 Bugzilla : 9969, 10379
4671 Description: utime reports stale NFS file handle
4672 Details : SLES9 uses out-of-dentry names in some cases, which confused
4673 the lustre dentry revalidation. Change it to always use the
4677 Frequency : rare, unless heavy write-truncate concurrency is continuous
4678 Bugzilla : 4180, 6984, 7171, 9963, 9331
4679 Description: OST becomes very slow and/or deadlocked during object unlink
4680 Details : filter_destroy() was holding onto the parent directory lock
4681 while truncating+unlinking objects. For very large objects this
4682 may block other threads for a long time and slow overall OST
4683 responsiveness. It may also be possible to get a lock ordering
4684 deadlock in this case, or run out of journal credits because of
4685 the combined truncate+unlink. Solution is to do object truncate
4686 first in one transaction without parent lock, and then do the
4687 final unlink in a new transaction with the parent lock. This
4688 reduces the lock hold time dramatically.
4691 Frequency : rare, 2.4 kernels only
4693 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4694 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4695 before unmounting their filesystem in order to avoid deadlock.
4696 The kernel_locked() function in 2.4 kernels only checks whether
4697 the kernel lock is held, not whether it is this process that is
4698 holding it as 2.6 kernels do.
4703 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4704 Details : The client ptlrpc code may be trying to reconnect to a down
4705 server before a previous connection attempt has timed out.
4706 Increase the reconnect interval to be longer than the connection
4707 timeout interval to avoid sending duplicate connections to
4711 Frequency : echo_client brw_test command
4713 Description: fix echo_client to work with OST preallocated code
4714 Details : OST preallocation code (5137) didn't take echo_client IO path
4715 into account: echo_client calls filter methods outside of any
4716 OST thread and, hence, there is no per-thread preallocated
4717 pages and buffers to use. Solution: hijack pga pages for IO. As
4718 a byproduct, this avoids unnecessary data copying.
4722 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4723 Description: Client can oops in mdc_commit_close() after open replay
4724 Details : It was possible for the MDS to return an open request with no
4725 transaction number in mds_finish_transno() if the client was
4726 evicted, but without actually returning an error. Clients
4727 would later try to replay that open and may trip an assertion
4728 Simplify the client close codepath, and always return an error
4729 from the MDS in case the open is not successful.
4732 Frequency : rare, 2.6 OSTs only
4734 Description: OST may deadlock under high load on fragmented files
4735 Details : If there was a heavy load and highly-fragmented OST filesystems
4736 it was possible to have all the OST threads deadlock waiting on
4737 allocation of biovecs, because the biovecs were not released
4738 until the entire RPC IO was completed. Instead, release biovecs
4739 as soon as they are complete to ensure forward IO progress.
4741 Severity : enhancement
4743 Description: Support for specifying external journal device at mount
4744 Details : If an OST or MDS device is formatted with an external journal
4745 device, this device major/minor is stored in the ext3 superblock
4746 and may not be valid for failover. Allow detecting and
4747 specifying the external journal at mount time.
4752 Description: Mounting an MDS with pending unlinked files may cause oops
4753 Details : target_finish_recovery() calls mds_postrecov() which returned
4754 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4755 considers this an error and immediately begins cleaning up the
4756 lov, just after starting the mds_lov process
4758 Severity : enhancement
4760 Description: Implement 'lfs df' to report actual free space on per-OST basis
4761 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4762 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4763 usage of objects; '-h' to report in human readable format.
4765 ------------------------------------------------------------------------------
4767 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4774 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4775 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4776 at startup during orphan unlinking. Instead, skip buddy update
4777 before setup, as it will be regenerated after recovery is complete.
4782 Description: If an OST is inactive, its locks might reference stale inodes.
4783 Details : lov_change_cbdata() must iterate over all namespaces, even if
4784 they are inactive to clear inode references from the lock.
4786 Severity : enhancement
4787 Frequency : occasional, if non-standard max_dirty_mb used
4789 Description: Client will block write RPCs if not enough grant
4790 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4791 then the client will block writes while waiting for another RPC
4792 to complete instead of consuming its dirty limit. With change
4793 we get improved performance when max_dirty_mb is small.
4795 Severity : enhancement
4796 Bugzilla : 3389, 6253
4797 Description: Add support for supplementary groups on the MDS.
4798 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4799 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4800 lookups for user supplementary groups into a cache.
4804 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4805 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4806 for liblustre O_EXCL is enforced by MDS.
4810 Description: readdir never returns NULL in liblustre.
4811 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4812 dentry in struct dirent.
4816 Frequency : liblustre only, depends on application IO pattern
4817 Description: liblustre clients evicted if not contacting servers
4818 Details : Don't put liblustre clients into the ping_evictor list, so
4819 they will not be evicted by the pinger ever.
4821 Severity : enhancement
4823 Description: Add ability to evict clients by NID from MDS.
4824 Details : By echoing "nid:$NID" string into
4825 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4826 $NID would be instantly evicted from this MDS and from all active
4827 OSTs connected to it.
4831 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4832 Details : lookup_it_finish() used to query file size from OSTs that was not
4837 Description: service threads change working directory to that of init
4838 Details : Starting lustre service threads may pin the working directory
4839 of the parent thread, making that filesystem busy. Threads
4840 now change to the working directory of init to avoid this.
4844 Frequency : during shutdown only
4845 Description: shutdown with a failed MDS or OST can cause unmount to hang
4846 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4851 Frequency : 2.6 only
4852 Description: chmod/chown may include an extra supplementary group
4853 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4854 supplementary group and if none is specified this is used.
4857 Bugzilla : 5479 (6816)
4859 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4860 Details : If lookup is in progress on a file that is unlinked we might try
4861 to revalidate the inode and fail in revalidate after lookup is
4862 complete and ll_file_open() enqueues the open again but
4863 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4866 Frequency : always, if lconf --abort_recovery used
4868 Description: lconf --abort_recovery fails with 'Operation not supported'
4869 Details : lconf was attempting to abort recovery on the MDT device and not
4872 ------------------------------------------------------------------------------
4874 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4879 Frequency : rare (only unsupported configurations with a node running as an
4881 Bugzilla : 6514, 5137
4882 Description: Mounting a Lustre file system on a node running as an OST could
4884 Details : OSTs now preallocates memory needed to write out data at
4885 startup, instead of when needed, to avoid having to
4886 allocate memory in possibly low memory situations.
4887 Specifically, if the file system is mounted on on OST,
4888 memory pressure could force it to try to write out data,
4889 which it needed to allocate memory to do. Due to the low
4890 memory, it would be unable to do so and the node would
4891 become unresponsive.
4893 Severity : enhancement
4895 Description: Addition of lconf --service command line option
4896 Details : lconf now accepts a '--service <arg>' option, which is
4897 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4899 Severity : enhancement
4901 Description: Failover mode is now the default for OSTs.
4902 Details : By default, OSTs will now run in failover mode. To return to
4903 the old behaviour, add '--failout' to the lmc line for OSTs.
4905 Severity : enhancement
4907 Description: Health checks are now provided for MDS and OSTs
4908 Details : Additional detailed health check information on MSD and OSTs
4909 is now provided through the procfs health_check value.
4912 Frequency : occasional, depends on IO load
4914 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4915 after numerous create/delete cycles
4916 Details : The ext3 inode allocation policy would not allocate new inodes
4917 very well on the OSTs because there are no new directories
4918 being created. Instead we look for groups with free space if
4919 the parent directories are nearly full.
4924 Description: Network or server problems during mount may cause partially
4925 mounted clients instead of returning an error.
4926 Details : The config llog parsing code may overwrite the error return
4927 code during mount error handling, returning success instead
4933 Description: MDS can fail to allocate large reply buffers
4934 Details : After long uptimes the MDS can fail to allocate large reply
4935 buffers (e.g. zconf client mount config records) due to memory
4936 fragmentation or consumption by the buffer cache. Preallocate
4937 some large reply buffers so that these replies can be sent even
4938 under memory pressure.
4942 Frequency : rare (liblustre)
4943 Description: fsx running with liblustre complained that using truncate() to
4944 extend the file doesn't work. This patch corrects that issue.
4945 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4946 ATTR_SIZE and lsm use in llu_setattr_raw.
4950 Frequency : rare, only 2.6 kernels
4951 Description: Unusual file access patterns on the MDS may result in inode
4952 data being lost in very rare circumstances.
4953 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4954 code in the 2.6 kernel caused the inode and/or EA data not to
4955 be read from disk, causing single-file corruption.
4959 Frequency : rare, only 2.6 filesystems using extents
4960 Description: Heavy concurrent write and delete load may cause data corruption.
4961 Details : It was possible under high-load situations to have an extent
4962 metadata block in the block device cache from a just-unlinked
4963 file overwrite a newly-allocated data block. We now unmap any
4964 metadata buffers that alias just-allocated data blocks.
4968 Frequency : filesystems with default stripe_count larger than 77
4969 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4970 Details : lconf specifies an inode size of 4096 bytes when the default
4971 stripe_count is larger than 77. This conflicts with the default
4972 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4973 this case to avoid pinning too much memory for large EAs.
4975 ------------------------------------------------------------------------------
4977 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4982 Frequency : rare (extremely heavy IO load with hundreds of clients)
4984 Description: Client is evicted, gets IO error writing to file
4985 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4986 caused clients to be evicted for AST timeouts. The fixes in
4987 bug 5192 mean we no longer need to have such short AST timeouts
4988 so ldlm_timeout has been increased.
4991 Frequency : occasional during --force or --failover shutdown under load
4992 Bugzilla : 5949, 4834
4993 Description: Server oops/LBUG if stopped with --force or --failover under load
4994 Details : a collection of import/export refcount and cleanup ordering
4995 issues fixed for safer force cleanup
4998 Frequency : only filesystems larger than 120 OSTs
4999 Bugzilla : 5990, 6223
5000 Description: lfs getstripe would oops on a very large filesystem
5001 Details : lov_getconfig used kfree on vmalloc'd memory
5004 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5005 Bugzilla : 6242, 6243
5006 Description: reading from files that had been truncated to a non-zero size
5007 but never opened returned no data
5008 Details : ll_file_read() reads zeros from no-object files to EOF
5013 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5014 Details : There's little harm in aborting MDS/OSS recovery and letting it
5015 try again, so I removed the LASSERT and return an error instead.
5017 Severity : enhancement
5019 Description: New debugging infrastructure for tracking down data corruption
5020 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5021 (b) cover more of the client-side code path, and (c) try to narrow
5022 down where problems occurred
5026 Bugzilla : 3819, 4364, 4397, 6313
5027 Description: Racing close and eviction MDS could cause assertion in mds_close
5028 Details : It was possible to get multiple mfd references during close and
5029 client eviction, leading to one thread referencing a freed mfd.
5031 Severity: : enhancement
5032 Bugzilla : 3262, 6359
5033 Description: Attempts to reconnect to servers are now more aggressive.
5034 Details : This builds on the enhanced upcall-less recovery that was added
5035 in 1.4.2. When trying to reconnect to servers, clients will
5036 now try each server in the failover group every 10 seconds. By
5037 default, clients would previously try one server every 25 seconds.
5042 Description: After recovery, certain operations trigger a failed
5043 assertion on a client.
5044 Details : Failing over an mds, using lconf -d --failover, while a
5045 client was doing a readdir() call would cause the client to
5046 LBUG after recovery completed and the readdir() was resent.
5048 Severity : enhancement
5050 Description: Default groups are now added by lconf
5051 Details : You can now run lconf --group <servicename> without having to
5052 manually add groups with lmc.
5055 Frequency : occasional
5057 Description: Nodes with an elan id of 0 trigger a failed assertion
5060 Frequency : always when accessing e.g. tty/console device nodes
5062 Description: tty and some other devices nodes cannot be used on lustre
5063 Details : file's private_data field is used by device data and lustre
5064 values in there got lost. New field was added to struct file to
5065 store fs-specific private data.
5068 Frequency : when exporting Lustre via NFS
5070 Description: NFSD failed occasionally when looking up a path component
5071 Details : NFSD is looking up ".." which was broken in ext3 directories
5072 that had grown large enough to become hashed.
5075 Frequency : Clusters with multiple interfaces not on the same subnet
5077 Description: Nodes will repeatedly try to reconnect to an interface which it
5078 cannot reach and report an error to the log.
5079 Details : Extra peer list entries will be created by lconf with some peers
5080 unreachable. lconf now validates the peer before adding it.
5083 Frequency : Only if a default stripe is set on the filesystem root.
5085 Description: Setting a default stripe on the filesystem root prevented the
5086 filesystem from being remounted.
5087 Details : The client was sending extra request flags in the root getattr
5088 request and did not allocate a reply buffer for the dir EA.
5091 Frequency : occasional, higher if lots of files are accessed by one client
5092 Bugzilla : 6159, 6097
5093 Description: Client trips assertion regarding lsm mismatch/magic
5094 Details : While revalidating inodes the VFS looks up inodes with ifind()
5095 and in rare cases can find an inode that is being freed.
5096 The ll_test_inode() code will free the lsm during ifind()
5097 when it finds an existing inode and then the VFS later attaches
5098 this free lsm to a new inode.
5102 Bugzilla : 6422, 7030
5103 Description: MDS deadlock between mkdir and client eviction
5104 Details : Creating a new file via mkdir or mknod (starting a transaction
5105 and getting the ns lock) can deadlock with client eviction
5106 (gets ns lock and trying to finish a synchronous transaction).
5109 Frequency : occasional
5110 Description: While starting a server, the fsfilt_ext3 module could not be
5112 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5113 kernels. Previously, lconf would still use the ext3 name
5114 when trying to load modules. Now, it will correctly use
5115 ext3 on 2.4 and ldiskfs on 2.6.
5117 Severity : enhancement
5118 Description: The default stripe count has been changed to 1
5119 Details : The interpretation of the default stripe count (0, to lfs
5120 or lmc) has been changed to mean striping across a single
5121 OST, rather than all available. For general usage we have
5122 found a stripe count of 1 or 2 works best.
5124 Severity : enhancement
5125 Description: Add support for compiling against Cray portals.
5126 Details : Conditional compiling for some areas that are different
5130 Frequency : occasional
5131 Bugzilla : 6409, 6834
5132 Description: Creating files with an explicit stripe count may lead to
5133 a failed assertion on the MDS
5134 Details : If some OSTs are full or unavailable, creating files may
5135 trigger a failed assertion on the MDS. Now, Lustre will
5136 try to use other servers or return an error to the
5140 Frequency : occasional
5142 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5143 caused lock timeout during readahead (since 1.4.2).
5144 Details : Processes doing readahead might match a lock that hasn't been
5145 granted yet if there are overlapping and conflicting lock
5146 requests. The readahead process waits on ungranted lock
5147 (original lock is CBPENDING), while OST waits for that process
5148 to cancel CBPENDING read lock and eventually evicts client.
5150 Severity : enhancement
5152 Description: Initial enabling of flock support for clients
5153 Details : Implements fcntl advisory locking and file status functions.
5154 This feature is provided as an optional mount flag (default
5155 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5156 locking are implemented yet, and those that are are not guaranteed
5157 to be completely correct in production environments.
5158 mount -t lustre -o [flock|noflock] ...
5161 Frequency : occasional
5163 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5164 assertion in the ext3 JBD (journaling) layer.
5165 Details : The b_committed_data struct is protected by the big kernel lock
5166 in 2.4 kernels, serializing journal_commit_transaction() and
5167 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5168 access to this struct. In 2.6 kernels there is finer grained
5169 locking to improve SMP performance of the JBD layer.
5173 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5178 Description: merge and cleanup kernel patches.
5181 -----------------------------------------------------------------------------
5183 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5185 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5186 versions of Lustre. Please update all servers and clients to
5187 version 1.4.2 or later at the same time. You must also run
5188 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5189 to update the configuration logs.
5191 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5192 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5193 - swab llog records of type '0' so we get proper header size/idx (5861)
5194 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5195 - fix rename of one directory over another leaking an inode (5953)
5196 - avoid SetPageDirty on 2.6 (5981)
5197 - don't re-add just-being-destroyed locks to the waiting list (5653)
5198 - when creating new directories, inherit the parent's custom
5199 striping settings if present parent (3048)
5200 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5201 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5202 - don't hold client locks on temporary worklist from l_lru (5666)
5203 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5204 - automatically evict dead clients (5921)
5205 - Update file size properly in create+truncate+fstat case (6196)
5206 - Do not unhash mountpoint dentries, do not allow removal of
5208 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5209 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5210 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5211 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5212 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5213 - Correctly handle reads of files with no objects (6243)
5214 - lctl recover will also mark a device active if deactivate used (5933)
5216 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5217 - allow --write-conf on an MDS with different nettype than client (5619)
5218 - don't write config llogs to MDS for mounts not from that MDS (5617)
5219 - lconf should create multiple TCP connections from a client (5201)
5220 - init scripts are now turned off by default; run chkconfig --on
5221 lustre and chkconfig --on lustrefs to use them
5222 - upcalls are no longer needed for clients to recover to failover
5224 - add --abort-recovery option to lconf to abort recovery on device
5226 - add support for an arbitrary number of OSTs (3026)
5227 - Quota support protocol changes.
5228 - forward compatibility changes to wire structs (6007)
5229 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5230 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5231 - improved "lustre status" script
5232 - initialize blocksize for non-regular files (6062)
5233 - added --disable-server and --disable-client configure options (5782)
5234 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5235 - Vanilla 2.4.29 support
5236 - increase maximum number of obd devices to 520 (6242)
5237 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5238 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5239 SLES 9 2.6.5 kernels
5240 - limit stripes per file to 160 (the maximum EA size) (6093)
5242 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5245 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5246 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5247 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5248 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5249 - small ext3 extents cleanups and fixes (5733)
5250 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5251 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5252 - avoid needless client->OST connect, fix handle mismatch (5317)
5253 - fix DLM error path that led to out-of-sync client, long delays (5779)
5254 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5255 - fix several locking issues related to i_size (5492,5624,5654,5672)
5256 - don't move pending lock onto export if it is already evicted (5683)
5257 - fix kernel oops when creating .foo in unlinked directory (5548)
5258 - fix deadlock in obdfilter statistics vs. object create (5811)
5259 - use time_{before,after} to avoid timer jiffies wrap (5882)
5260 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5261 - Do not leak request if server was not able to process it (5154)
5262 - If mds_open unable to find parent dir, make that negative lookup(5154)
5263 - don't create new directories with extent-mapping (5909, 5936)
5265 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5266 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5267 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5268 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5269 - support for RHEL4 kernel on i686 (5773)
5270 - provide error messages when incompatible logs are encountered (5898)
5272 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5273 * version 1.4.0.10 (1.4.1 release candidate 1)
5275 - don't keep a lock reference when lock is not granted (4238)
5276 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5277 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5278 - drop import inflight refcount on signal_completed_replay error (5255)
5279 - unlock page after async write error during send (3677)
5280 - handle missing objects in filter_preprw_read properly (5265)
5281 - no transno return for symlink open, don't save no-trasno open (3440)
5282 - don't try to complete elan receive that already failed (4012)
5283 - free RPC server reply state on error (5406)
5284 - clean up thread from ptlrpc_start_thread() on error (5160)
5285 - readahead could read extra page into cache that wasn't ejected (5388)
5286 - prevent races in class_attach/setup/cleanup/detach (5260)
5287 - don't dereference de->d_inode after l_dput of de (5458)
5288 - use "int" for stripe value returned from lock_to_stripe (5544)
5289 - mballoc allocation and error-checking fixes in 2.6 (5504)
5290 - block device patches to fix I/O request sizes in 2.6 (5482)
5291 - look up hostnames for IB nals (5602)
5292 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5293 - don't start multiple acceptors for the same port (5277)
5294 - fix incorrect LASSERT in mds_getattr_name (5635)
5295 - export a proc file for general "ping" checking (5628)
5296 - fix "lfs check" to not block when the MDS is down (5628)
5298 - service request history (4965)
5299 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5300 - create an "evict_client" /proc entry on OSTs, like the MDS has
5301 - fix mount usage message, return errors per mount(8) (5168)
5302 - change grep [] to grep "[]" in tests so they work in more UMLs
5303 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5304 - remove ancient LOV_MAGIC_V0 EA support (5047)
5305 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5306 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5307 - allow readahead allocations to fail when low on memory (5383)
5308 - mmap locking landed again, after considerable improvement (2828)
5309 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5311 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5314 - send OST transaction number in read/write reply to free req (4966)
5315 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5316 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5317 - the watchdog thread now runs as interruptible (5246)
5318 - flock/lockf fixes (but it's still disabled, pending 5135)
5319 - don't use EXT3 constants in llite code (5094)
5320 - memory shortage at startup could cause assertion (5176)
5322 - reorganization of lov code
5323 - single portals codebase
5325 - add extents/mballoc support (5025)
5326 - direct I/O reads in the obdfilter (4048)
5327 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5329 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5332 - send OST transaction number in read/write reply to free req (4966)
5333 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5334 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5335 - the watchdog thread now runs as interruptible (5246)
5336 - handle missing objects in filter_preprw_read properly (5265)
5337 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5338 - drop import inflight refcount on signal_completed_replay error (5255)
5339 - unlock page after async write error during send (3677)
5340 - return original error code on reconstructed replies (3761)
5341 - no transno return for symlink open, don't save no-trasno open (3440)
5343 - add pid to ldlm debugging output (4922)
5344 - bump the watchdog timeouts -- we can't handle 30sec yet
5345 - extra debugging for orphan dentry/inode bug (5259)
5347 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5350 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5351 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5352 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5353 - fix NULL dereference in /proc/sys/portals/routes (4827)
5354 - allow failed mdc_close() operations to be interrupted (4561)
5355 - stop precreate on OST before MDS would time out on it (4778)
5356 - don't send partial-page writes before EOF from client (4410)
5357 - discard client grant for sub-page writes on large-page clients (4520)
5358 - don't free dentries not owned by NFS code, check generation (4806)
5359 - fix lsm leak if mds_create_objects() fails (4801)
5360 - limit debug_daemon file size, always print CERROR messages (4789)
5361 - use transno after validating reply (3892)
5362 - process timed out requests if import state changes (3754)
5363 - update mtime on OST during writes, return in glimpse (4829)
5364 - add mkfsoptions to LDAP (4679)
5365 - use ->max_readahead method instead of zapping global ra (5039)
5366 - don't interrupt __l_wait_event() during strace
5368 - add software watchdogs to catch hung threads quickly (4941)
5369 - make lustrefs init script start after nfs is mounted
5370 - fix CWARN/ERROR duplication (4930)
5371 - return async write errors to application if possible (2248)
5372 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5373 - print NAL number in %x format (4645)
5374 - update barely-supported suse-2.4.21-171 series (4842)
5375 - support for sles 9 %post scripts
5376 - support for building 2.6 kernel-source packages
5377 - support for sles km_* packages
5379 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5382 - ignore -ENOENT errors in osc_destroy (3639)
5383 - notify osc create thread that OSC is being cleaned up (4600)
5384 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5385 - reconstruct ost_handle() like mds_handle() (4657)
5386 - create a new thread to do import eviction to avoid deadlock (3969)
5387 - let lconf resolve symlinked-to devices (4629)
5388 - don't unlink "objects" from directory with default EA (4554)
5389 - hold socknal file ref over connect in case target is down (4394)
5390 - allow more than 32000 subdirectories in a single directory (3244)
5391 - fix blocks count for O_DIRECT writes (3751)
5392 - OST returns ENOSPC from object create when no space left (4539)
5393 - don't send truncate RPC if file size isn't changing (4410)
5394 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5395 - bind to privileged port in socknal and tcpnal (3689)
5397 - rate limit CERROR/CWARN console message to avoid overload (4519)
5398 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5399 - basic mmap support (3918)
5400 - kernel patch series update from b1_4 (4711)
5402 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5405 - avoid crash during MDS cleanup with OST shut down (2775)
5406 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5407 - don't use bad inodes on the MDS (3744)
5408 - dynamic object preallocation to improve recovery speed (4236)
5409 - don't hold spinlock over lock dumping or change debug flags (4401)
5410 - don't zero obd_dev when it is force cleaned (3651)
5411 - print grants to console if they go negative (4431)
5412 - "lctl deactivate" will stop automatic recovery attempts (3406)
5413 - look for existing locks in ldlm_handle_enqueue() (3764)
5414 - don't resolve lock handle twice in recovery avoiding race (4401)
5415 - revalidate should check working dir is a directory (4134)
5417 - don't always mark "slow" obdfilter messages as errors (4418)
5419 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5422 - don't close LustreDB during write_conf until it is done (3860)
5423 - fix typo in lconf for_each_profile (3821)
5424 - allow dumping logs from multiple threads at one time (3820)
5425 - don't allow multiple threads in OSC recovery (3812)
5426 - fix debug_size parameters (3864)
5427 - fix mds_postrecov to initialize import for llog ctxt (3121)
5428 - replace config semaphore with spinlock (3306)
5429 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5430 - don't allow enqueue to complete on a destroyed export (3822)
5431 - down write_lock before checking llog header bitmap (3825)
5432 - recover from lock replay timeout (3764)
5433 - up llog sem before sending rpc (3652)
5434 - reduce ns lock hold times when setting kms (3267)
5435 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5436 - fix NULL deref and obd_dev leak on setup error (3312)
5437 - replace some LBUG about llog ops with error handling (3841)
5438 - don't match INVALID dentries from d_lookup and spin (3784)
5439 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5440 - fix invalid assertion in ptlrpc_set_wait (3880)
5442 - add libwrap support for the TCP acceptor (3996)
5443 - add /proc/sys/portals/routes for non-root route listing (3994)
5444 - allow setting MDS UUID in .xml (2580)
5445 - print the stack of a process that LBUGs (4228)
5447 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5450 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5451 - make sure to unset replay flag from failed open requests (3440)
5452 - if default stripe count is 0, use OST count for inode size (3636)
5453 - update parent mtime/ctime on client for create/unlink (2611)
5454 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5455 - free recovery state on server during a forced cleanup (3571)
5456 - unregister_reply for resent reqs (3063)
5457 - loop back devices mounting and status check on 2.6 (3563)
5458 - fix resource-creation race that can provoke i_size == 0 (3513)
5459 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5460 - more debugging for page-accounting assertion (3746)
5461 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5462 - avoid deadlock after precreation failure (3758)
5463 - fix race and lock order deadlock in orphan handling (3450, 3750)
5464 - add validity checks when grabbing inodes from l_ast_data (3599)
5466 - add /proc/.../recovery_status to obdfilter (3428)
5467 - lightweight CDEBUG infrastructure, debug daemon (3668)
5468 - change default OSC RPC parameters to be better on small clusters
5469 - turn off OST read cache for files smaller than 32MB
5470 - install man pages and include them in rpms (3100)
5471 - add new init script for (un)mounting lustre filesystems (2593)
5472 - run chkconfig in %post for init scripts (3701)
5473 - drop scimac NAL (unmaintained)
5475 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5478 - clean kiobufs before and after use (3485)
5479 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5480 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5481 - change init script to fail more clearly if not run as root (1528)
5482 - allow clients to reconnect during replay (1742)
5483 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5484 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5485 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5486 - fix cond_resched (3554)
5487 - only evict unfinished clients after recovery (3515)
5488 - allow bulk resend, prevent data loss (3570)
5489 - dynamic ptlrpc request buffer allocation (2102)
5490 - don't allow unlinking open directory if it isn't empty (2904)
5491 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5492 - remove extraneous obd dereference causing LASSERT failure (3334)
5493 - don't use get_cycles() when creating temp. files on the mds (3156)
5494 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5495 - handle EEXIST for set-stripe, set proper directory name (3336)
5497 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5498 - fix ksocknal_fmb_callback() error messages (2918)
5500 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5503 - don't copy lvb into (possibly NULL) reply on error (2983)
5504 - don't deref dentry after dput, don't free lvb on error (2922)
5505 - use the kms to determine writeback rpc length (2947)
5506 - increment oti_logcookies when osc is inactive (2948)
5507 - update client's i_blocks count via lvb messages (2543)
5508 - handle intent open/close of special files properly (1557)
5509 - mount MDS with errors=remount-ro, like obdfilter (2009)
5510 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5511 - don't use cancelling-locks' kms values (2947)
5512 - use highest lock extent for kms, not last one (2925)
5513 - don't dereference ERR_PTR() dentry in error handling path (3107)
5514 - fix thread race in portals_debug_dumplog() (3122)
5515 - create lprocfs device entries at setup instead of at attach (1519)
5516 - common AST error handler, don't evict client on completion race (3145)
5517 - zero nameidata in detach_mnt in 2.6 (3118)
5518 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5519 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5520 - initialize RPC timeout timer earlier for 2.6 (3219)
5521 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5522 - print nal/nid for unknown nid (3258)
5523 - additional checks for oscc recovery before doing precreate (3284)
5524 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5525 - don't crash in mdc_close for bad permissions on open (3285)
5526 - zero i_rdev for non-device files (3147)
5527 - clear page->private before handing to FS, better assertion (3119)
5528 - tune the read pipeline (3236)
5529 - fix incorrect decref of invalidated dentry (2350)
5530 - provide read-ahead stats and refine rpc in flight stats (3328)
5531 - don't hold journal transaction open across create RPC (3313)
5532 - update atime on MDS at close time (3265)
5533 - close LDAP connection when recovering to avoid server load (3315)
5534 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5535 - don't leak open file on MDS after open resend (3325)
5536 - serialize filter_precreate and filter_destroy_precreated (3329)
5537 - loop device shouldn't call sync_dev() for nul device (3092)
5538 - clear page cache after eviction (2766)
5539 - resynchronize MDS->OST in background (2824)
5540 - refuse to mount the same filesystem twice on same mountpoint (3394)
5541 - allow llmount to create routes for mounting behind routers (3320)
5542 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5543 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5544 - fix rare problem with rename on htree directories (3417)
5546 - allow default OST striping configuration per directory (1414)
5547 - fix compilation for qswnal for 2.6 kernels (3125)
5548 - increase maximum number of MDS request buffers for large systems
5549 - change liblustreapi to be useful for external progs like lfsck (3098)
5550 - increase local configuration timeout for slow disks (3353)
5551 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5553 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5556 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5557 - don't overwrite extent policy data in reply if lock was blocked (2901)
5558 - drop filter export grants atomically with removal from device (2663)
5559 - del obd_self_export from work_list in class_disconnect_exports (2908)
5560 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5561 - swab reply message in mdc_close, other PPC fixes (2464)
5562 - fix destroying of named logs (2325)
5563 - overwrite old logs when running lconf --write_conf (2264)
5564 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5565 - fix race in target_handle_connect (2898)
5566 - mds_reint_create() should take same inode create lock (2926)
5567 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5568 - don't close files for self_export to avoid uninitialized obd (2936)
5569 - allow MDS with the same name as client node (2939)
5570 - hold dentry reference for closed log files for unlink (2325)
5571 - reserve space for all logs during transactions (2059)
5572 - don't evict page beyond end of stripe extent (2925)
5573 - don't oops on a deleted current working directory (2399)
5574 - handle hard links to targets without a parent properly (2517)
5575 - don't dereference NULL lock when racing during eviction (2867)
5576 - don't grow lock extents when lots of conflicting locks (2919)
5578 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5581 - account for cache space usage on clients to avoid data loss (974)
5582 - lfsck support in lustre kernel code (2349)
5583 - reduce journal credits needed for BRW writes (2370)
5584 - orphan handling to avoid losing space on client/server crashes
5585 - ptlrpcd can be blocked, stopping ALL progress (2477)
5586 - use lock value blocks to assist in proper KMS, faster stat (1021)
5587 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5588 - recovery for initial connections (2355)
5589 - fixes for mds_cleanup_orphans (1934)
5590 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5591 - block all file creations until orphan recovery completes (1901)
5592 - client remove rq_connection from request struct (2423)
5593 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5594 - recovery timer race (2670)
5595 - mdc_close recovey bug (2532)
5596 - ptlrpc cleanup bug (2710)
5597 - mds timeout on local locks (2588)
5598 - namespace lock held during RPCs (2431)
5599 - handle interrupted sync write properly (2503)
5600 - don't try to handle a message that hasn't been replied to (2699)
5601 - client assert failure during cleanup after abort recovery (2701)
5602 - leak mdc device after failed mount (2712)
5603 - ptlrpc_check_set allows timedout requests to complete (2714)
5604 - wait for inflight reqs when ptlrpcd finishes (2710)
5605 - make sure unregistered services are removed from the srv_list
5606 - reset bulk XID's when resending them (caught by 1138 test)
5607 - unregister_bulk after timeout
5608 - fix lconf error (2694)
5609 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5610 - readahead locks pages, leaves pending causing memory pressure (2673)
5611 - increase OST request buffers to 4096 on large machines (2729)
5612 - fix up permission of existing directories in simple_mkdir (2661)
5613 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5614 - don't assign transno to errored transactions (2742)
5615 - don't delete objects on OST if given a bogus objid from MDS (2751)
5616 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5617 - if rq_no_resend, then timeout request after recovery (2432)
5618 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5619 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5620 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5621 - MDS recovery shouldn't delete objects that it has given out (2730)
5622 - if enqueue arrives after completion, don't clobber LVB (2819)
5623 - don't unlock pages twice when trigger_group_io returns error (2814)
5624 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5625 - don't write pages to disk if there was an error (1450)
5626 - don't ping imports that have recovery disabled (2676)
5627 - take buffered bytes into account when balancing socknal conn (2817)
5628 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5629 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5630 - remove little-endian swabbing of llog records (1987)
5631 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5632 - retry reposting request buffers if they fail (1191)
5633 - grow extent at grant time to avoid granting a revoked lock (2809)
5634 - lock revoke doesn't evict page if covered by a second lock (2765)
5635 - disable VM readahead to avoid reading outside lock extents (2805)
5637 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5638 - updated kernel patches for hp-2.4.20 kernel (2681)
5640 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5643 - fix truncated write corruption (2366)
5644 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5646 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5647 - ensure len(uuid) < 37 in lmc (1171)
5648 - fix ia64 OOPS in llog_test (2255)
5649 - zero end of page at obdfilter for partial page writes (2648)
5650 - don't leave stale dentries around after renames (bug 2428)
5651 - fix timeouts when evicting a client with a single lock held (2642)
5652 - set deadline for the initial HELLO message to drain (2634)
5653 - print out dotted-quad IP addresses in the socknal (2302)
5655 - additional debugging for MDS client eviction problem (2443)
5656 - fix mkfsoptions support for osts (2603, 2604)
5658 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5661 - add series for the vanilla 2.6.0 kernel
5662 - add series for the vanilla 2.4.24 kernel
5663 - add series for a cray x86/64 UL kernel drop
5664 - fix xattr patches for the vanilla 2.4.19 series
5666 - generate true UUIDs in lmc (1171)
5667 - have portals stack dumping break in UML (2466)
5668 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5669 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5670 - wake callback waiting threads on client eviction (2460)
5671 - Add --ptldebug and --subsystem to lmc (1719)
5672 - update assertion to allow safe interrupt allocation
5673 - set rq_no_resend for cancel requests (2432)
5674 - recalculate ptlrpcd timeout after resend (2494)
5675 - call vfs_rmdir when removing pending directories (2368)
5676 - fix renaming a file to itself (2429)
5677 - lmc creates a default one-stripe lov (2454)
5678 - expand procfs space to handle large clusters (2326)
5679 - increase UML stack to avoid overflow
5680 - update lconf's list of debug and subsystem masks
5681 - fix lfs find --obd (2510)
5682 - /proc tunable for disabling filter read caching (2591)
5683 - stop rpm packages from altering slapd.conf (2301)
5684 - disable nagle in the socknal under 0conf (2578)
5685 - choose mds inode size based on stripe count (2572)
5686 - fix kernel-source rpm problems (2516)
5688 - add --disable-doc to avoid pdf generation (2421)
5689 - update documentation, tests, type-os, comments
5690 - avoid format warnings on ia64
5691 - remove the TOE NAL
5692 - tiny code cleanups by removing unused fields
5694 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5697 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5698 - remove the most blatant lies from BUILDING file (2371)
5699 - change default debug level to reasonable production setting
5700 - reduce client side cache size to reduce cache flush time
5701 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5702 - make TCP zerocopy and pinger support enabled by default (2476)
5703 - sync writes completed after process exits caused crashes (2319)
5704 - maintain correct mount count on the MDS (2356)
5705 - backout 1557, because 2316 wasn't really fixed
5706 - better file I/O statistics gathering in /proc
5707 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5708 - another kernel patch to fix zero-copy TCP function export
5709 - don't take duplicate lock when processing re-sent getattr (2420)
5710 - lctl uses obd_self_export instead of creating new conn (2353)
5711 - MDS/OST recovery case which requires object creation asserted (2425)
5712 - move lfs from /usr/sbin to /usr/bin in packages
5713 - fix race between mds_client_add and mds_client_free (2417)
5714 - use kmalloc instead of slabs in portals (2430)
5715 - don't create duplicate records when a failover MDS is present (2442)
5716 - remove unnecessary mount age check (2332)
5717 - don't remove directory inodes from locks prematurely (2451)
5718 - don't break if MDS service name is the same as hostname (2103)
5719 - fix races in client write RPC generation when cache full (2482)
5721 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5724 - remove now-unused request->rq_obd (278)
5725 - if an allocation fails, print out how much memory we've used (1933)
5726 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5727 - add the "configurable stack size" patch to most series files (1256)
5728 - ability to write large log records, for 100+ OST configs (2306)
5729 - fix NULL deref when filter_prep fails (2314)
5730 - fix operator precedence error in filter_sync
5731 - dynamic allocation of socknal TX descriptors (2315)
5732 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5733 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5734 - ensure that CWARN messages in Portals always get to the syslog
5735 - __init/__exit are not for prototype decls (ldlm_init/exit)
5736 - x86-64 compile warning fixes
5737 - fix gateway LMC keyword conflict (2318)
5738 - fix MDS lock inversions in getattr/reint paths (1844)
5739 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5740 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5741 - DLM race condition prevented some lock evictions (2328)
5742 - ENOMEM detection and retry on socknal sends (2230)
5743 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5744 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5745 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5746 - better error messages when a client is rejected during recovery (1505)
5747 - avoid cancelling locks which were never granted, after failure (2330)
5748 - fix i_sem/journal inversion in mds_client_add (2333)
5749 - fix truncate/getattr lock cycle deadlock (2334)
5750 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5751 - fix two rare exit paths which could leak an l_lock() ref (2321)
5752 - fencepost error in MDS/OST orphan recovery (2226)
5753 - make log record alignment 8 bytes (1988)
5754 - lstripe now fails when requested offset > ost_count (2237)
5755 - ensure that all kernel series have a complete list.h (1607)
5756 - fix crashes in special-file operations (2316)
5757 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5759 - allow configurable automake binary, for testing new versions
5760 - small update to the lfs documentation
5762 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5764 * fix negative export reference count in fsfilt_sync (2312)
5766 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5767 * release candidate 0.9.1
5769 - orphans are moved into the PENDING directory for possible recovery
5770 - replayed opens now open by fid for orphan/rename safety (1042)
5771 - last close of an orphan inode generates a transno (683)
5772 - chdir() and mount() now pin the directory entry (1020)
5773 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5774 - discard very old requests without processing them (1502)
5775 - remove some common, well-understood CERRORs (1505)
5776 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5777 - clear "grant" flags in OST replies until OST grant code lands (1644)
5778 - fix read performance by not clobbering i_blksize on client (1598)
5779 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5780 - make LRU size a /proc tunable, clears locks when reduced (707)
5781 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5782 - server locks take references on exports now (1558)
5783 - build fixes for 2.4.20-rh trees (1663)
5784 - return an error from lov_create if all OSCs are inactive (1751)
5785 - fix import levels when a reconnect happens without a timeout (1597)
5786 - exit early from mds_open if we get a lookup error (1749)
5787 - partial page read at EOF wouldn't wait for disk before sending (1642)
5788 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5789 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5790 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5791 - write/truncate lock inversion (1639)
5792 - Don't auto-load obdclass, portals modules during cleanup (1495)
5793 - fix timestamps from jumping to "now" (1763)
5794 - extra journal assertions (1648)
5795 - add an extra multiunlink test (1771)
5796 - fix read_record/write_record API (1776)
5797 - fix leak of offset_extent, possible incorrect i_size later (1772)
5798 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5799 - Debugging for the kqswnal_get_idle_tx problems (1820)
5800 - Allow recovery to be attempted multiple times (1536)
5801 - Write out MDS last_rcvd file after it is first created (1600)
5802 - Fix tx_descriptor leak in failed transmit situations (1827)
5803 - ext3 journaling fixes for assertion failure after IO error (1871)
5804 - class_export_put() on freed export after completion AST error (1896)
5805 - Fix revalidate looping in VFS (1322)
5806 - Don't access a freed export during MDS_REINT timeout (1521)
5807 - Add open-unlink recovery support on the MDS (1673,1764)
5808 - Return an error if no MDS data was read from last_rcvd (1946)
5809 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5810 - Don't setup a disk device that doesn't match exported UUID (317)
5811 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5812 - avoid committing NULL handle in force close
5813 - local.sh is now a one-stripe LOV configuration
5814 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5815 - fix ext3 htree duplicate directory entry corruption (1516)
5816 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5817 - update to new LOV EA format (2097)
5818 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5820 2003-06-15 Phil Schwan <phil@clusterfs.com>
5823 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5824 - per-import/export recovery handling (958, 931, 959)
5825 - multiple last-rcvd slots, for serving multiple FSes (949)
5826 - connections are again shared between multiple imp/exports (963, 964)
5827 - "umount -f" would hang if any requests needed to be sent (393, 978)
5828 - avoid pinning large req buffer by copying for queued messages (989)
5829 - add "uuid" to "lctl device" command to help upcalls (991)
5830 - "open" RPCs with transnos would confuse recovery counters (1037)
5831 - do proper endian conversion of all wire messages (288, 340, 891)
5832 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5833 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5834 - fix ptlrpc_connection leak in target_handle_connect (1174)
5835 - fix import refcounting bug in OST and MDS cleanup (1134)
5836 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5837 - fix very unlikely obd_types race condition (501)
5838 - remove osc_open hack for echo_client (1187)
5839 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5840 - a failure in read_inode2 leads to deadlock (1139)
5841 - cancel ack-locks as soon as transaction is committed (1072)
5842 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5843 - make sure to commitrw after any preprw to avoid deadlock (1162)
5844 - failing to execute a file in a lustre FS would lock inode (1203)
5845 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5846 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5847 - obd_uuid2tgt would walk off the end of the list (1255)
5848 - on IA64 the osc would give portals incorrect bulk size (1258)
5849 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5850 - fix lock inversion caused by new llite matching code (1282)
5851 - limit the number of dirty pages on a client to 10MB (1286)
5852 - timed out locks were not being corrected cancelled (1289)
5853 - fix O_DIRECT above 4GB on IA-32 (1292)
5854 * major user-visible changes
5855 - fail out/fail over policy now controlled by the upcall (993)
5857 - add OBD_PING to check server availability and failure (954)
5858 - lustre messages are now sent in sending host order (288, 340, 891)
5859 - add eadatalen to MDS getattr reply (340)
5860 - OST read replies may contain second buffer, with per-page status (593)
5862 2003-03-11 Phil Schwan <phil@clusterfs.com>
5865 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5866 - failed open()s could cause deadlock; fixed (867, 869)
5867 - stop cancelling OST locks when files are closed (481)
5868 - overlapping XID spaces caused network corruption (851, 853)
5869 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5870 - setattr_raw added, to do single-RPC, server-side setattrs
5871 - lmc/lconf syntax change for OST UUIDs
5872 - fix crashy race condition between ptlrpc_free_req and osc_close
5873 - don't use request in mdc_enqueue if we hit a timeout (889)
5874 - don't set the inode i_size for regular files from the MDS (896)
5875 - handle out of order completion AST (842)
5876 - don't LBUG if a lock request times out after receiving AST (913)
5877 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5878 - if a bad lock AST arrives, send an error instead of dropping entirely
5879 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5880 - fix leak in bulk IO when only partially completed (899, 900, 926)
5881 - fix O_DIRECT for ia64 (55)
5882 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5883 - C-z after timeout could hang a process forever; fixed (977)
5885 - client-side I/O cache (678, 924, 929, 941, 970)
5887 - READPAGE and SETATTRs which don't take server-side locks get
5890 2003-02-11 Phil Schwan <phil@clusterfs.com>
5893 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5894 - Fix fsfilt_extN_readpage() to read a full page of directory
5895 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5896 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5897 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5898 - Most intent operations were removed, in favour of a new RPC mode
5899 that does a single RPC to the server and bypasses most of the VFS
5900 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5901 - Aggressively cancel local locks on DLM servers
5902 - mds_reint_unlink sends EA to the client if it's the last nlink.
5903 client uses that EA to unlink OST objects.
5904 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5905 - recursive symlinks were fixed (439)
5906 - fixed NULL deref in DEBUG_REQ
5907 - filter_update_lastobjid no longer calls sync, which annoyed extN
5908 - fixed multi-client small-writes to a single file problem (445)
5909 - fixed mtime updates during file writes (607)
5910 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5911 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5912 - fixed Linux OST size reporting problem (444, 656)
5913 - OST now updates object mtime with writes or setattr (607, 619)
5914 - client verifies file size before zeroing page past EOF (445)
5915 - OST now writes last allocated objid to disk with allocation (108)
5916 - LOV on echo now works (409)
5918 - mds_reint_unlink sends a new buffer, with the EA included. this
5919 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5920 set if a regular file was being unlinked, and it was the last link
5921 - use PtlGet from the target for bulk writes (315)
5922 - OST now updates object mtime with writes or setattr (607, 619)
5923 - LDLM now has a grant-time callback to revalidate locked items, if
5925 - Many MDS operations were reorganized to combat race conditions
5927 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5928 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5930 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5933 - Fully reactivate OST imports after reconnection (512, others)
5934 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5935 - More graceful error handling for truncating on dead OST (515)
5936 - Don't error out unless we're actually accessing dead stripes (474)
5937 - Fix garbage sizes when stripes are missing (410)
5938 - LRU counters were broken, causing constant lock purge (433, 432)
5939 - garbage on read from stripes with failed OSTs (441)
5940 - mark OSCs as active before reconnecting during recovery (438)
5941 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5942 - lfind did not preserve OST order in output (443)
5943 - symlinks cause hung clients, incorrect data (439)
5944 - stop dereferencing request after dropping refcount (457)
5945 - don't LASSERT(spin_is_locked) on non-SMP (455)
5946 - fixes for many rename() bugs
5947 - fstat didn't correctly synchronize attributes (399)
5948 - server must handle lock cancellation during blocking AST prep (487)
5949 - bulk descriptors were free()d too soon (511)
5950 - fix paths in lconf, which would load incorrect modules (451, 507)
5951 - fix confusing lconf 'host not found' error message (386)
5952 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5953 - fix race condition in mdc_blocking_ast() for inode access (526)
5954 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5955 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5956 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5957 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5958 - fix lstripe in several ways (kernel side) (527)
5959 - fix request leak in ldlm_cli_enqueue (262)
5960 - incorrect OSC was marked inactive after OST failure
5961 - call mds_fs_cleanup before unmounting filesystem (524)
5962 - fix races between taking ns_lock and ldlm_lock_change_resource
5963 - fix races updating LOV export open file list
5964 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5965 - fix recovery NULL deref in ldlm_cli_cancel_unused
5966 - fix some DLM races by using new hash table for lock handles (419)
5967 - permit the client to specify desired inodes, at replay
5968 - duplicate requests when we queue them for replay reintegration
5969 - fix last_rcvd offset calculation
5970 - sync after each recovered transaction, so we always make progress
5971 - never, not always, ERESTART requests without transnos
5972 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5974 - skip replay if the MDS says that the client is already connected
5975 - don't check for a recovery-enabled export to match lctl's UUID
5976 - don't INC_USE_COUNT for phantom exports
5977 - don't crash when cleaning up phantom exports (567)
5978 - don't double-finish or set replay data for errored mdc_open requests
5979 - abort requests when they time out, so we don't get old replies
5980 - send/receive replies for AST messages again
5981 - if the client says that it doesn't have the lock, cancel it on the
5983 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5984 mark it as destroyed, it will all work out when decref is called
5985 - fix module use counts (22, 581)
5987 - ASTs now expect a reply (server cancels lock on error reply)
5989 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5992 - fix many simultaneous client startup (392)
5993 - fix dentry->d_it clobbering
5994 - credentials weren't being shipped for readdir/getattr operations
5995 - remove invalid assertions triggered during some concurrent MD
5997 - proper Lustre versions added (336, 389)
5998 - fix memory leak for create error case (398)
5999 - fix LOV locking bug that would get cli/srv out of sync
6000 - fix echo client over LOV (409)
6001 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6002 - fix double-O_EXCL intent crash (424)
6003 - avoid sending multiple lock CANCELs (352)
6005 - MDS can do multi-client recovery (modulo bugs in new code)
6007 - many updates, edits, cleanups
6009 2002-11-18 Phil Schwan <phil@clusterfs.com>
6012 - fix null d_it dereference (346)
6013 - fix full OST/dbench hang (333)
6014 - fix permission problem with file removal (286)
6015 - fix removal of OSCs from LOV when they fail
6016 - fix NULL deref during bulk timeout (214)
6017 - fix problems related to multiple filesystems on one MDS (241)
6018 - fixed serious subtle metadata locking bugs
6019 - free locks on clients when inodes are removed due to memory
6021 - fix inode pointer in lock data (285)
6022 - partial support for multiple MDS on a single host (241)
6023 - data locks weren't cancelled at clear_inode time (290, 311)
6024 - intent locks could lead to unbounded lock growth (205)
6025 - added a maximum lock count, an LRU list, and a flusher
6026 - fix multiple rename (365)
6027 - properly abstracted the echo client
6028 - OSC locked 1 byte too many; fixed
6029 - rewrote brw callback code:
6030 - fixed recovery bugs related to LOVs (306)
6031 - fixed too-many-pages-in-one-write crash (191)
6032 - fixed (again) crash in sync_io_timeout (214)
6033 - probably fixed callback-related race (385)
6035 - Add capability to MDS protocol
6036 - LDLM cancellations and callbacks on different portals
6038 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6041 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6042 - defer open object destruction to close time (601981, 138)
6043 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6044 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6045 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6046 - don't call obd_setattr in ll_file_release for destroyed objects
6047 * protocol change to lustre_msg: move |version| and add |flags|
6048 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6049 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6050 * added replay of create, unlink, link and rename operations during
6051 MDS failover; recovery should be much more robust now
6052 * remove failed OSCs from LOVs (only lov_create uses this so far)
6053 * the lustre-HOWTO was brought (more) up to date (582544)
6055 2002-10-23 Phil Schwan <phil@clusterfs.com>
6058 - in-use dentries weren't being reused properly (617851)
6059 - prevent multiple LDLM setup (599178)
6060 - fix LOV size calculations for truncate (617853)
6061 - fix client handling of MDS intent errors (POSIX)
6062 - fix permission bug in lovstripe.c test (624321)
6063 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6064 - truncate past end of file could corrupt data
6065 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6066 - a race in recovery could return ETIMEDOUT to apps (623947)
6067 - building outside the source directory was fixed
6068 * the lustre-HOWTO was brought (more) up to date (582544)
6069 * major progress was made on recovery functionality
6071 2002-10-10 Phil Schwan <phil@clusterfs.com>
6074 - recovery deadlock fix
6075 - rm -rf causes LBUG fix (617817)
6076 - file open by multiple tasks fix (618962)
6077 - directory permissions bugs (602707 and 620007)
6078 - journal_stop fixed with locking (611313)
6079 - O_APPEND failures resolved (618273, perhaps 614459)
6080 - lconf PATH fix (619770)
6081 - IA64 build fix (621450)
6082 - RPC buffer sizes scale with amount of memory
6084 2002-10-01 Phil Schwan <phil@clusterfs.com>
6087 - locks would be cancelled without throwing away data pages,
6088 resulting in inconsistent data (605627)
6089 - inode attributes were not always being refreshed (605627, 612449)
6090 - lconf now continues to cleanup after lctl reports an error
6091 - MDS now enforces user permissions (602707)
6092 - lprocfs cleanup fixed, but not yet enabled (614157)
6093 - fixed infinite server hang, should a client not respond to an AST
6094 - avoid going into recovery if user calls readlink() with a buffer
6095 that's too small (613941)
6096 - AST RPCs no longer require replies (614867) -- this may be changed
6097 - don't crash server if client sends an IOV that's too big (611336)
6098 - fixed lock conversion deadlock (611892)
6099 - fixed the following of symlinks (614622)
6100 * recovery: the server can remove locks from a client that dies, other
6101 clients can make progress
6102 * more extN patch fixes
6103 * compile-time configurable ptlrpc buffer allocations
6105 - collaborative read cache document
6106 - Lustre Lite Performance CDR document-in-progress
6108 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6111 - fix typo in patch-2.4.18
6113 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6116 - clear ptlrpc request each time in handle_incoming_request()
6117 - unlink of files now destroys the object on the OST
6119 2002-09-19 Peter Braam <braam@clusterfs.com>
6121 * add hard link support
6122 * change obdfile creation method
6123 * kernel patch changed
6125 2002-09-19 Peter Braam <braam@clusterfs.com>
6128 - stack overflow bug in extN fixed
6130 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6132 * documentation updates
6133 - add man pages for config tools
6134 - update tests/README to describe testing with new config tools
6135 - finish metadata API descriptions
6136 * bug fixes and cleanups
6137 - statfs workaround for 16TB limit
6138 - LOV stripe allocation improved, can stripe on subset of OSTs
6139 - LOV file size/IO offset was wrong for files > 4GB in size
6140 - object EA data was being dropped, caused files to be unreadable
6141 - memory overflow with non-LOV OST caused memory corruption
6142 - fixed regression tests to work with new config tools, obdfilter
6143 - fixed bug when directory size became larger than 1 block
6144 - fixed bug (for single client case) when PWD was deleted
6145 - invalidate local directory pages when doing intent-based ops
6146 - avoid LDLM oops when lock callback contained bad data
6148 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6150 * documentation updates
6151 * bug fixes and cleanups
6152 - configuration tools
6155 - 64-bit compile warnings
6156 - 64-bit internal statfs data
6158 * test_brw on persistent OST devices
6160 * lprocfs (disabled)
6162 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6164 * documentation updates
6165 * bug fixes and cleanups
6166 * configuration tools
6168 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6171 * many small fixes to 0_5_4
6172 * io/network handling
6173 * thinkos in MDS operations
6175 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6178 * crucial basic fixes to 0.5.3
6179 * IOR, Iozone work over Elan
6182 2002-08-07 Phil Schwan <phil@clusterfs.com>
6183 * version 0_5_3, our first alpha
6184 * we use the new Portals iovs
6185 * documentation updates
6186 * bug fixes and cleanups
6187 * small changes in the DLM wire protocol
6189 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6190 * version 0_5_1 with some initial stability,
6191 * locking on MD and file I/O.
6192 * documentation updates
6193 * several bug fixes since 0.5.0
6194 * small changes in wire protocol
6196 2002-07-18 Phil Schwan <phil@clusterfs.com>
6198 * delivered as Lustre Light Alpha
6199 * fixed a crash after handling invalid MDS requests
6200 * fixed directory pages for architectures with non-4k pages sizes
6202 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6203 * release version v0_4_4
6204 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6206 * Fixes a number of interruption problems with OST operations.
6207 * Update documentation for portals header changes
6208 * Move all wire protocol structs/defines to lustre_idl.h
6209 * Fixes symlink length bug.
6210 * Add tcpdump to repository.
6212 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6213 * release version v0_4_3
6214 * Fixes statfs for inodes on extN.
6215 * Fixes bug in runtests which would delete /etc/hosts.
6216 * Use 64-bit object IDs wherever possible (not into VFS though)
6217 Remove ost_get_info, which is unused by lustre, and out of date.
6219 2002-07-03 Peter Braam <braam@clusterfs.com>
6220 * release version v0_4_2 Fixes a lookup error (type not passed)
6221 * move forward to head of Portals
6222 * move forward to latest Lustre kernel
6224 2002-06-25 Peter Braam <braam@clusterfs.com>
6225 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob