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
15 Severity : enhancement
17 Description: lfs quota output is not very convenient for awk/sed-parsing
18 Details : Some positions in lfs quota output table could be empty or
19 non-empty which made it hard to parse it with scripts, now
20 a dash is put instead of space where there is not supposed
25 Description: Update to OFED 1.5.1
26 Details : OFED 1.5.1 will be used in environments that don't use
31 Description: Remove lnet. prefix from lctl params display
32 Details : Also renamed lnet.memused to lnet_memused to
37 Description: Use conf_param -d to remove permanant settings.
41 Description: Add list_param -R
42 Details : lctl list_param now supports recursive mode.
45 Bugzilla : 21829, 22312
46 Description: Fix llobdstat.
47 Details : 1. Now updating the write_bytes counter
48 2. Port script fixes from 1.8.
52 Description: decrease the usage of memory on clients.
53 Details : 1. On clients, recycle dentries and inodes unused.
54 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
59 Description: llobdstat fix and enhancement
60 Details : add a counter to set a limit to how many samples will be returned
61 fix a wildcard in the path to limit to obdfilter stats only
65 Description: lfs check only allowed for root
66 Details : Code cleanups and added a regresion test of non-root use of
71 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
72 Details : Fix hr->hr_index overload.
76 Description: Fix FIEMAP functionality.
80 Description: FC11 patchless client support added
84 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
85 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
89 Description: New acc-sm script and various test-framework enhancments.
93 Description: Previously landed patch was reverted.
97 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
101 Description: Add ldiskfs maxdirsize mount option.
105 Description: Fix mkfs.lustre support for 16TB LUNs.
109 Description: Print mmp_check_interval
110 Details : print mmp_check_interval and make it possible to abort
111 mount operation in case it takes too long.
115 Description: Add cascading_rw to lustre mpi tests.
119 Description: Store SOM data in EA.
120 Details : Also several SOM related cleanups.
124 Description: Add refcount to md_open_data
128 Description: Use refcount on lov_request_set to prevent memory corruption.
132 Description: Statahead: use parent dir's i_mutex for synchronization.
136 Description: loadgen improvements.
140 Description: Enable peer health detection in ksocklnd and ko2iblnd.
144 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
148 Description: Fix race on obd_devcie:md_stats
152 Description: Fix SLES 11 client builds.
153 Details : Resolved uninitalized pointer gcc warnings.
157 Description: Allow non-root access for "lfs check".
158 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
162 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
164 Severity : enhancement
166 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
167 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
171 Description: Sanity test_27p fails.
172 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
173 Sanity tests 27 were cleaned up and improved.
177 Description: Unify req format on client/servers
178 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
179 buffers and swabbers approach..
184 Description: drop join file stuff.
186 Severity : enhancement
188 Description: Add 2.6.27 and SLES11 (patchless client) support
192 Description: Fix in ptlrpc_expire_one_request() to print the signed time
195 Severity : enhancement
197 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
200 Severity : enhancement
202 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
204 Severity : enhancement
206 Description: correctly handle big reply message.
207 Details : send LNet event if reply is bigger then buffer and adjust this buffer
212 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
214 Severity : enhancement
216 Description: Add support for OEL5.
220 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
224 Description: client could not reconnect to OST because of an active request.
225 Details : The client now retries to reconnect to the same server, if a
226 connect request failed with EBUSY or -EAGAIN. Test case 35b
227 conf-sanity.sh has been created.
231 Description: Don't allow make backward step on assignin osc next id.
232 Details : race between allocation next id and ll_sync thread can be cause
233 of set wrong osc next id and can be kill valid ost objects.
237 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
241 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
245 Description: Error handling in osc_statfs_interpret() has been improved.
246 Details : Check in osc_statfs_interpret() for EBADR.
250 Description: Do not send statfs() requests to OSTs disabled by administrator.
251 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
253 Severity : enhancement
255 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
257 Severity : enhancement
259 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
261 Severity : enhancement
263 Description: Enhancement for lfs(1) command to use numeric uid/gid.
267 Description: Improved cleanup procedure for conf-sanity test_46a.
271 Description: Do not start bulk transfer, if client tried to reconnect.
272 Details : Do not start bulk transfer, if client tried to reconnect,
273 but failed due to active rpcs. Check for exp_abort_active_req
274 has been implemented.
278 Description: Prevent inconsistences between linux and lustre mount structures.
279 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
280 Make the sleep interruptible.
282 Severity : enhancement
284 Description: provide server to client comms path
285 Details : server to client communications path via new LDLM_SET_INFO rpc
289 Description: truncate starts GFP_FS allocation under transaction causing deadlock
290 Details : ldiskfs_truncate calls grab_cache_page which may start page
291 allocation under an open transaction. This may lead to
292 calling prune_icache with consequent lustre reentrance.
296 Description: send correctly lsm on open replay
297 Details : MDS is trust to LSM size on replay open, but client can set wrong size
302 Description: enable client interface failover
303 Details : When a child reconnects from another NID, properly update export
304 nid hash position and ldlm reverse import.
306 Severity : enhancement
308 Description: Communicate OST degraded/readonly state via statfs to MDS
309 Details : Flags in the statfs returned from OSTs indicate whether the
310 OST is in a degraded RAID state, or if the filesystem has
311 turned read-only after a filesystem error is detected.
316 Description: don't panic if EPROTO was hit when reading symlink
317 Details : correctly handling request reference in error cases.
319 Severity : enhancement
321 Description: Add LustreNetLink, a kernel-userspace communcation path.
323 Severity : enhancement
325 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
330 Description: access to llog context before init.
331 Details : move handling CATALOGS file at osc layer and forbid access to llog
335 Frequency : rare, only when client is writing to OST on local node
337 Description: Avoid deadlock for local client writes
338 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
339 memory freeing context. This allows OST threads to set the
340 PF_MEMALLOC flag on task structures in order to allocate memory
341 from reserved pools and complete IO.
342 Use GFP_HIGHUSER for OST allocations for non-local client writes,
343 so that the OST threads generate memory pressure and allow
344 inactive pages to be reclaimed.
346 Severity : enhancement
348 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
353 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
354 Details : move ldlm namespace creation in setup phase to avoid grab
355 _lprocfs_lock with cli_sem held.
359 Description: Temporarily disable grant shrink.
360 Details : Disable the feature for debugging.
363 Frequency : only when formatting test filesystems
365 Description: Unable to run several mkfs.lustre on loop devices at the same time
366 Details : mkfs.lustre returns error 256 on the concurrent loop devices
367 formatting. The solution is to proper handle the error.
369 Severity : enhancement
371 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
373 Severity : enhancement
375 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
380 Description: resolve race between obd_disconnect and class_disconnect_exports
381 Details : if obd_disconnect will be called to already disconnected export he
382 forget release one reference and osc module can't unloaded.
384 Severity : enhancement
386 Description: Allow tuning service thread via /proc
387 Details : For each service a new
388 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
389 created that can be used to set min/max thread counts, and get the
390 current number of running threads.
395 Descriptoin: don't return error if have partially created objects for file.
396 Details : lov_update_create_set uses set->set_success as index for created
397 objects, so if some requests will be failed, they will have hole
398 at end of array and qos_shrink_lsm for allocate correct lsm.
400 Severity : enhancement
402 Description: Update OFED support to 1.4.1
407 Description: Reduce small size read RPC
408 Details : Set read-ahead limit for every file and only do read-ahead when
409 available read-ahead pages are bigger than 1M to avoid small size
412 Severity : enhancement
414 Description: move AT tunable parameters for more consistent usage
415 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
417 Severity : enhancement
419 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
420 Details : allow skip disconnected ost for send statfs request and hide error
426 Description: fix racy locking of mballoc block bitmaps causing BUG
427 Details : The locking of the mballoc buddy bitmap and the in-memory
428 block bitmap was using two different spin locks in some
429 cases. This made it possible to incorrectly access the
430 mballoc bitmap while another process was modifying it,
431 causing a sanity assertion to fail. While no on-disk corruption
432 was reported, there was some risk of this happening.
435 Frequency : rare, on llog test 6
437 Descriptoin: don't allow connect to already connected import
438 Details : allowing connect to already connected import is hide connecting problem.
443 Descriptoin: allow kill process which wait statahead result
444 Details : for some reasons 'ls' can stick in waiting result from statahead,
445 in this case need way for kill this process.
447 Severity : enhancement
449 Description: Add state history info file, enhance import info file
450 Details : Track import connection state changes in a new osc/mdc proc file;
451 add overview-type data to the osc/mdc import proc file.
453 Severity : enhancement
455 Description: MDS create should not wait for statfs RPC while holding DLM lock.
457 Severity : enhancement
459 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
464 Descriptoin: OST grant too much space to client even there are not enough space.
465 Details : Client will shrink its grant cache to OST if there are no write
466 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
467 this grant cache if there are already not enough avaible space
468 (left_space < total_clients * 32M).
471 Frequency : start MDS on uncleanly shutdowned MDS device
473 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
474 Details : stay in waiting mds<>ost recovery finished produce random bugs
475 due race between two ll_sync thread for one lov target. send
476 ACTIVATE event only if connect realy finished and import have
480 Frequency : rare, connect and disconnect target at same time
482 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
483 Details : don't call obd_disconnect under lov_lock. this long time
484 operation and can block ptlrpcd which answer to connect request.
489 Descriptoin: don't lose wakeup for imp_recovery_waitq
490 Details : recover_import_no_retry or invalidate_import and import_close can
491 both sleep on imp_recovery_waitq, but we was send only one wakeup
495 Frequency : always with long access acl
497 Descriptoin: mds can't pack reply with long acl.
498 Details : mds don't control size of acl but they limited by reint/getattr
501 Severity : enhancement
503 Description: Update to SLES10 kernel-2.6.16.60-0.33.
505 Severity : enhancement
507 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
510 Frequency : start MDS on uncleanly shutdowned MDS device
512 Descriptoin: aborting recovery hang on MDS
513 Details : don't throttle destroy RPCs for the MDT.
516 Frequency : on remount
518 Description: external journal device not working after the remount
519 Details : clear dev_rdonly flag for external journal devices in
525 Description: shutdown vs evict race
526 Details : client_disconnect_export vs connect request race.
527 if client will evicted at this time - we start invalidate
528 thread without referece to import and import can be freed
532 Frequency : rare, need acl's on inode.
534 Description: client can't handle ost additional correctly
535 Details : if ost was added after client connected to mds client can have
536 hit lnet_try_match_md ... to big messages to wide striped files.
537 in this case need teach client to handle config events about add
538 lov target and update client max ea size at that event.
540 Severity : enhancement
542 Description: Changelogs
543 Details : Changelogs are a lightweight mechanism to track filesystem
544 metadata and namespace changes. The changelog is recorded
545 permanently on the MDTs, and is periodically "consumed" / purged
546 when records are no longer needed.
548 Severity : enhancement
550 Description: compact fld format with extents
551 Details : Store range of seq rather than every seq in FLD. Seq
552 controller update FLD rather than clients. In Case of CMD, mdt0
553 has FLD, all other metadata server act as non persistent proxy
554 for FLD queries and cache fld entries in fld cache.
559 Description: don't skip ost target if they assigned to file
560 Details : Drop slow OSCs if we can, but not for requested start idx.
561 This means "if OSC is slow and it is not the requested
562 start OST, then it can be skipped, otherwise skip it only
563 if it is inactive/recovering/out-of-space.
567 Description: more cleanup in mds_lov
568 Details : not send LOV EA under replay, we can't know about they size at this
569 time. Don't allow client connect to mds before any ost connected,
570 for avoid problems with LOV EA size and returning EIO to client.
572 Severity : enhancement
574 Description: Interoperability at server side (Disk interoperability)
576 Severity : enhancement
578 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
580 Severity : enhancement
582 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
584 Severity : enhancement
586 Description: New client IO stack (CLIO).
588 Severity : enhancement
590 Description: Commit on sharing. Eliminate inter-client dependencies between
591 uncommitted transactions by doing transaction commits.
592 Thereby clients may recovery independently.
595 Frequency : Create a symlink file with a very long name
597 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
598 Details : If there is no extra space in the request for early cancels,
599 ldlm_req_handles_avail() returns 0 instead of a negative value.
601 Severity : enhancement
603 Description: Add /proc entry for import status
604 Details : The mdc, osc, and mgc import directories now have
605 an import directory that contains useful import data for debugging
608 Severity : enhancement
610 Description: Re-disable certain /proc logging
611 Details : Enable and disable client's offset_stats, extents_stats and
612 extents_stats_per_process stats logging on the fly.
615 Frequency : Only on FC kernels 2.6.22+
617 Description: oops in statahead
618 Details : Do not drop reference count for the dentry from VFS when lookup,
619 VFS will do that by itself.
621 Severity : enhancement
623 Description: Generic /proc file permissions
624 Details : Set /Proc file permissions in a more generic way to enable non-
625 root users operate on some /proc files.
629 Description: Hitting mdc_commit_close() ASSERTION
630 Details : Properly handle request reference release in
631 ll_release_openhandle().
635 Description: quota recovery deadlock during mds failover
636 Details : This patch includes att18982, att18236, att18237 in bz14840.
638 1. fix osts hang when mds does failover with quotaon
639 2. prevent watchdog storm when osts threads wait for the
644 Frequency : only patchless client
645 Description: add workaround for race between add/remove dentry from hash
647 Severity : enhancement
649 Description: Allow OST glimpses to return PW locks
653 Description: LBUG when llog conf file is full
654 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
658 Description: Prevent import from entering FULL state when server in recovery
662 Description: service mount cannot take device name with ":"
663 Details : Only when device name contains ":/" will mount treat it as
669 Description: replace ptlrpcd with the statahead thread to interpret the async
670 statahead RPC callback
674 Frequency : on recovery
675 Description: I/O failures after umount during fail back
676 Details : if client reconnected to restarted server we need join to recovery
677 instead of find server handler is changed and process self
678 eviction with cancel all locks.
680 Severity : enhancement
682 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
684 Severity : enhancement
686 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
688 Severity : enhancement
690 Description: Upcall on Lustre log has been dumped
691 Details : Allow for a user mode script to be called once a Lustre log has
692 been dumped. It passes the filename of the dumped log to the
693 script, the location of the script can be specified via
694 /proc/sys/lnet/debug_log_upcall.
699 Description: avoid idr_remove called for id which is not allocated.
700 Details : Move assigment s_dev for clustered nfs to end of initialization,
701 to avoid problem with error handling.
706 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
707 Details : When connection is reused this not moved from CONN_UNUSED_HASH
708 into CONN_USED_HASH and this prodice warning when put connection
709 again in unused hash.
712 Severity : enhancement
714 Description: File striping can now be set to use an arbitrary pool of OSTs.
716 Severity : enhancement
718 Description: Export bytes_read/bytes_write count on OSC/OST.
722 Description: Early reply size mismatch, MGC loses connection
723 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
724 the connect flags are properly negotiated.
729 Description: quotas are not honored with O_DIRECT
730 Details : all writes with the flag O_DIRECT will use grants which leads to
731 this problem. Now using OBD_BRW_SYNC to guard this.
735 Description: add quota statistics
736 Details : 1. sort out quota proc entries and proc code.
737 2. add quota statistics
739 Severity : enhancement
741 Description: enable quota support for HEAD.
745 Description: Properly propagate oinfo flags from lov to osc for statfs
746 Details : restore missing copy oi_flags to lov requests.
748 Severity : enhancement
750 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
751 : l_getgroups(8), lst(8), routerstat(8)
752 Details : included man pages for llobdstat(8), llstat(8),
753 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
755 Severity : enhancement
757 Description: configure's --enable-quota should check the
758 : kernel .config for CONFIG_QUOTA
759 Details : configure is terminated if --enable-quota is passed but
760 : no quota support is in kernel
764 Description: Remove portals compatibility
765 Details : Remove portals compatibility, not interoperable with releases
770 Description: Resolve device initialization race
771 Details : Prevent proc handler from accessing devices added to the
772 obd_devs array but yet be intialized.
774 Severity : enhancement
776 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
778 Severity : enhancement
780 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
785 Description: Using wrong pointer in osc_brw_prep_request
786 Details : Access to array[-1] can produce panic if kernel compiled with
787 CONFIG_PAGE_ALLOC enabled
791 Description: Client runs out of low memory
792 Details : Consider only lowmem when counting initial number of llap pages
796 Description: *optional* service tags registration
797 Details : if the "service tags" package is installed on a Lustre node
798 When the filesystem is mounted, a local-node service tag will
799 be created. See http://inventory.sun.com/ for more information
800 about the Service Tags asset management system.
804 Description: Kernel BUG tries to release flock
805 Details : Lustre does not destroy flock lock before last reference goes
806 away. So always drop flock locks when client is evicted and
807 perform unlock regardless of successfulness of speaking to MDS.
811 Description: add refcount protection for osc callbacks, avoid panic on shutdown
815 Description: sanity test 65a fails if stripecount of -1 is set
816 Details : handle -1 striping on filesystem in ll_dirstripe_verify
821 Description: ASSERTION(CheckWriteback(page,cmd)) failed
822 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
825 Severity : enhancement
827 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
831 Description: do not process already freed flock
832 Details : flock can possibly be freed by another thread before it reaches
833 to ldlm_flock_completion_ast.
837 Description: LBUG during stress test
838 Details : Need properly lock accesses the flock deadlock detection list.
842 Description: oops in page fault handler
843 Details : kernel page fault handler can return two special 'pages' in error
844 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
848 Description: timeout with invalidate import.
849 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
850 should be handled by ptlrpcd. This produce long age waiting and
851 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
853 Severity : enhancement
855 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
859 Description: enable MGS and MDT services start separately
860 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
861 a co-located MGS without starting the MGS, which is a complement
862 to 'nosvc' mount option.
866 Frequency : after recovery
867 Description: precreate to many object's after del orphan.
868 Details : del orphan st in oscc last_id == next_id and this triger growing
869 count of precreated objects. Set flag LOW to skip increase count
870 of precreated objects.
874 Frequency : rare, on clear nid stats
875 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
876 Details : when clean nid stats sometimes try destroy live entry,
877 and this produce panic in free.
881 Description: Stack overflow during MDS log replay
882 ease stack pressure by using a thread dealing llog_process.
886 Description: wait until IO finished before start new when do lock cancel.
887 Details : VM protocol want old IO finished before start new, in this case
888 need wait until PG_writeback is cleared until check dirty flag and
889 call writepages in lock cancel callback.
891 Severity : enhancement
893 Description: using special macro for print time and cleanup in includes.
897 Description: mds_mfd_close() ASSERTION(rc == 0)
898 Details : In mds_mfd_close(), we need protect inode's writecount change
899 within its orphan write semaphore to prevent possible races.
903 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
907 Frequency : rare, on shutdown ost
908 Description: don't hit live lock with umount ost.
909 Details : shrink_dcache_parent can be in long loop with destroy dentries,
910 use shrink_dcache_sb instead.
914 Description: don't panic with use echo client
915 Details : echo client pass NULL as client nid pointer and this produce null
920 Description: fix build on ppc32
921 Details : compile code with -m64 flag produce wrong object file for ppc32.
925 Description: add message levels for liblustreapi
929 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
930 Details : Move the 'good_osts' check before the 'total_bavail' check. This
931 will result in an -EAGAIN and in the exit call path we call
932 alloc_rr() which will with increasing aggressiveness attempt to
933 aquire precreated objects on the minimum number of required OSCs.
937 Description: Use old size assignment to avoid deadlock
938 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
939 the scheduling while holding a spinlock. We do not need locking
940 for size in ll_update_inode() because size is only updated from
941 the MDS for directories or files without objects, so there is no
942 other place to do the update, and concurrent access to such inodes
943 are protected by the inode lock.
947 Description: resolve "_IOWR redefined" build error on SLES10
951 Description: dump the memory debugging after all modules are unloaded to
952 suppress false negative in conf_sanity test 39
954 Severity : enhancement
956 Description: build kernel-ib packages for OFED 1.3 in our release cycle
961 Description: fix SLES kernel versioning
962 Details : the kernel version for our SLES 10 kernel did not include a "-"
963 before the "smp" at the end. while this was not a problem in
964 general, it did mean that software trying to use the kernel
965 version to try to detect a vendor specific kernel would fail.
966 this was most evident by the OFED build scripts.
970 Description: Don't update lov_desc members until making sure they are valid
971 Details : When updating lov_desc members via proc fs, need fix their
972 validities before doing the real update.
976 Description: don't put request into delay list while invalidate in flight.
977 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
978 import in flight. this produce timeout for invalidate and sometimes
979 can cause stale data.
983 Frequency : on ppc only
984 Description: not convert ost objects for directory because it's not exist.
985 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
989 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
991 Severity : enhancement
993 Description: Update to support 2.6.22.14 vanilla kernel.
997 Frequency : rare, on recovery
998 Description: read procfs can produce deadlock in some situation
999 Details : Holding lprocfs lock which send rpc can produce block for destroy
1000 obd objects and this also block reconnect with -EALREADY. This
1001 isn't fix all lprocfs bugs - but make it rare.
1003 Severity : enhancement
1005 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1008 Frequency : frequent on X2 node
1010 Description: mdc_set_open_replay_data LBUG
1011 Details : Set replay data for requests that are eligible for replay.
1015 Description: lustre_mgs: operation 101 on unconnected MGS
1016 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1017 MGC, and late on MGC cannot successfully connect to MGS and a lot
1018 of the error messages complaining that MGS is not connected.
1022 Frequency : on network error
1023 Description: panic with double free request if network error
1024 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1025 but it's true only for synchronus enqueue, for async enqueue
1026 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1029 Severity : enhancement
1031 Description: client-side metadata stat-ahead during readdir(directory readahead)
1032 Details : perform client-side metadata stat-ahead when the client detects
1033 readdir and sequential stat of dir entries therein
1036 Frequency : on start mds
1038 Description: Implement get_info(last_id) in obdfilter.
1041 Frequency : occasional
1043 Description: Correctly check stale fid, not start epoch if ost not support SOM
1044 Details : open with flag O_CREATE need set old fid in op_fid3 because
1045 op_fid2 was overwritten with new generated fid, but mds can answer
1046 with one of these two fids and both is not stale. Setattr
1047 incorrectly started an epoch and assume will be called
1048 done_writing, but without SOM done_writing ever being called.
1051 Frequency : rare, depends on device drivers and load
1053 Description: MDS or OSS nodes crash due to stack overflow
1054 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1055 In some cases, in conjunction with device drivers that use a lot
1056 of stack the MDS (or possibly OSS) service threads could overflow
1057 the stack. One change which was identified to consume additional
1058 stack has been reworked to avoid the extra stack usage.
1061 Frequency : occasional
1063 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1064 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1065 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1066 oscc_has_objects and see OSC has no objests and call
1067 oscc_internal_create to resend create request.
1069 Severity : enhancement
1071 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1073 Severity : enhancement
1075 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1078 Frequency : very rare
1080 Description: Fix replay if there is an un-replied request and open
1081 Details : In some cases, older replay request will revert the
1082 mcd->mcd_last_xid on MDS which is used to record the client's
1083 latest sent request.
1085 Severity : enhancement
1087 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1089 Severity : enhancement
1091 Description: Add rhel5 support to HEAD.
1093 Serverity : enhancement
1095 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1100 Description: Don't allow skipping OSTs if index has been specified.
1101 Details : Don't allow skipping OSTs if index has been specified, make
1102 locking in internal create lots better.
1106 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1107 Details : access to bitfield in structure is always rounded to long
1108 and this produce problem with not atomic change any bit.
1112 Description: Lustre make rpms failed.
1113 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1114 builds Lustre rpms from tarball.
1117 Frequency : rare on shutdown ost
1119 Description: If llog cancel was not send before clean_exports phase, this can
1120 produce deadlock in llog code.
1121 Details : If llog thread has last reference to obd and call class_import_put
1122 this produce deadlock because llog_cleanup_commit_master wait when
1123 last llog_commit_thread exited, but this never success because was
1124 called from llog_commit_thread.
1128 Description: allow userland application know is lost one of stripes.
1129 Details : fill lvb_blocks with error code on ost and return it to
1130 application if error flag found.
1134 Description: NULL lov_tgts causing MDS oops
1135 Details : more safe checks for NULL lov_tgts for avoid oops.
1137 Severity : enhancement
1139 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1143 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1144 Details : If an OST is down the MDS will hang indefinitely in
1145 obd_statfs_rqset() waiting for the statfs data. While for
1146 MDS QOS usage of statfs, it should not stuck in waiting.
1148 Severity : enhancement
1150 Description: remote_acl support
1151 Details : Support ACL-based permission check for remote user.
1152 Support setfacl/getfacl for remote user with the utils
1153 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1154 the system "{s,g}etfacl" utils.
1156 Severity : enhancement
1158 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1160 Severity : enhancement
1162 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1166 Description: make mgs_setparam() handle fsname containing dash
1167 Details : fsname containing a dash does not work with lctl conf_param
1169 Severity : enhancement
1171 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1173 Severity : enhancement
1175 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1179 Frequency : rare, at shutdown
1180 Description: access already free / zero obd_namespace.
1181 Details : if client_disconnect_export was called without force flag set,
1182 and exist connect request in flight, this can produce access to
1183 NULL pointer (or already free pointer) when connect_interpret
1184 store ocd flags in obd_namespace.
1188 Frequency : only at startup
1189 Description: not alloc memory with spinlock held.
1190 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1191 if any spinlock held.
1193 Severity : enhancement
1195 Description: make lustre randomly fail allocating memory
1196 Details : Make lustre randomly failed allocating memory for testing purpose.
1198 Severity : enhancement
1200 Description: lost problems with lov objid file
1201 Details : Fixes some scability and access to not inited memory problems
1202 in work with lov objdid file.
1207 Description: lfs find does not continue on file error
1208 Details : Continue other files processing when a file/dir is absent.
1212 Description: Inconsistent usage of lustre_pack_reply()
1213 Details : Standardize the usage of lustre_pack_reply() such that it
1214 always generate a CERROR on failure.
1217 Frequency : occasional
1219 Description: MDS hang or stay in waiting lock
1220 Details : If client receive lock with CBPENDING flag ldlm need send lock
1221 cancel as separate rpc, to avoid situation when cancel request
1222 can't processed due all i/o threads stay in wait lock.
1226 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1230 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1234 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1238 Description: Update to RHEL4 latest kernel.
1240 Severity : enhancement
1242 Description: Build SLES10 patchless client fails
1243 Details : The configure was broken by run ./configure with
1244 --with-linux-obj=.... argument for patchless client. When the
1245 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1246 can't search header adequately. Use absolute path such as
1247 -I($LINUX)/include instead.
1251 Description: interrupt oig_wait produce painc on resend.
1252 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1253 private set, and this produce situation when rq_ptlrpcd_data not
1254 copyed to new allocated request and triggered LBUG on assert
1255 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1256 ptlrpcd set if request is changed and can be safety changed to use
1261 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1262 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1263 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1264 these macros and replace them with their difinitions instead.
1268 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1269 Details : Update which_patch & target file for SLES10 latest kernel.
1272 Description: Remove client patches from SLES 10 kernel.
1273 Details : This causes SLES 10 clients to behave as patchless clients
1274 even on a Lustre-patched (server) kernel.
1276 Severity : enhancement
1278 Description: self-adjustable client's lru lists
1279 Details : use adaptive algorithm for managing client cached locks lru
1280 lists according to current server load, other client's work
1281 pattern, memory activities, etc. Both, server and client
1282 side namespaces provide number of proc tunables for controlling
1285 Severity : enhancement
1287 Description: light-weight GSS support
1288 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1289 and reduce performance overhead.
1291 Severity : enhancement
1293 Description: Linux keyring support
1294 Details : Support using service of Linux keyring for Lustre GSS internal
1295 context refresh/cache mechanism.
1299 Description: Fix errors in lfs documentation
1300 Details : Fixes man pages
1304 Description: don't use GFP_* in generic Lustre code.
1305 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1309 Description: obdclass is limited by single OBD_ALLOC(idarray)
1310 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1314 Description: warnings with build patchless client with vanila 2.6.19 and up
1315 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1316 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1320 Frequency : Only for SLES
1321 Description: sanity_quota fail test_1
1322 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1323 file, which makes TSTID[2] inunique.
1327 Frequency : Only for SLES9
1328 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1329 Details : Update target/ChangeLog/which_patch .
1333 Frequency : Only for test_10 in sanity-quota.sh
1334 Description: a bug in quota test code
1335 Details : checking a test flag in a wrong place
1339 Frequency : Only for test_13 in sanity-quota.sh
1340 Description: a bug in quota test script
1341 Details : shouldn't check overfull allocated quota
1345 Description: jbd statistics
1346 Details : Port older jbd statistics patch for sles10
1348 Severity : enhancement
1350 Description: Add printing inode info into message about error in writepage.
1355 Description: reply_lock_interpret crash due to race with it and lock cancel.
1356 Details : Do not replay locks that are being cancelled. Do not reference
1357 locks by their address during replay, just by their handle.
1361 Frequency : When flocks are used.
1362 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1363 Details : Flock locks might destroy just granted lock if it could be merged
1364 with another existing flock, this is done in completion handler,
1365 so teach ldlm_cli_enquque_fini that this is a valid case for
1371 Description: Oops in read and write path when failing to allocate lock.
1372 Details : Check if lock allocation failed and return error back.
1376 Description: lstripe command fails for valid OST index
1377 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1378 instead of lov->desc.ld_active_tgt_count.
1382 Description: sanity.sh failed test 103
1383 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1388 Description: df doesn't work properly if diskfs blocksize != 4K
1389 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1393 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1394 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1396 Severity : enhancement
1398 Description: Adaptive timeouts
1399 Details : RPC timeouts adapt to changing server load and network
1400 conditions to reduce resend attempts and improve recovery time.
1404 Description: llapi_file_create() does not allow some changes
1405 Details : add llapi_file_open() that allows specifying the mode and
1406 open flags, and also returns an open file handle.
1410 Description: merge and cleanup kernel patches.
1411 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1415 Description: Add journal checksum support.(Kernel part)
1416 Details : The journal checksum feature adds two new flags i.e
1417 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1418 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1419 indicates that the commit block contains the checksum for
1420 the blocks described by the descriptor blocks. Now commit
1421 record can be sent to disk without waiting for descriptor
1422 blocks to be written to disk. This behavior is controlled
1423 using JBD2_FEATURE_ASYNC_COMMIT flag.
1427 Description: OSS needs mutliple precreate threads
1428 Details : Add ability to start more than one create thread per OSS.
1432 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1433 Details : execute lfs setstripe on client
1437 Description: mds_obd_create error creating tmp object
1438 Details : When the user sets quota on root, llog will be affected and can't
1439 create files and write files.
1442 Frequency : Always on ia64 patchless client, and possibly others.
1444 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1445 Details : This allows the patchless client to be loaded on architectures
1446 without this export.
1450 Description: RedHat Update kernel for RHEL5
1451 Details : Add the kernel config file for RHEL5.
1455 Description: RedHat Update kernel for RHEL5
1456 Details : Modify the kernel config file more closer RHEL5.
1460 Description: Build failure against Centos5 (RHEL5)
1461 Details : Define PAGE_SIZE when it isn't present.
1465 Description: test_7 on sanity-quota.sh can't be used on separate machines
1466 Details : add facet to handle it
1470 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1471 Details : with NFS, the anon dentry's parent was set to itself in
1472 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1473 corresponding dentry other than use rec->ur_name.
1475 Severity : enhancement
1477 Description: lfs setstripe enhancement
1478 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1482 Description: enable data checksumming by default
1483 Details : enable checksum by default, allow --disable-checksum
1484 configure option and "-o nochecksum" mount option. Checksums
1485 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1489 Description: lustre support for RHEL5
1490 Details : Add support for RHEL5.
1494 Description: Client eviction due to failover config
1495 Details : after a connection loss, the lustre client should attempt to
1496 reconnect to the last active server first before trying the
1497 other potential connections.
1501 Description: when mds and osts use different quota unit(32bit and 64bit),
1502 quota will be released repeatly.
1503 Details : void sending multiple quota reqs to mds, which will keep the
1504 status between the reqs.
1508 Description: osts not allocated evenly to files
1509 Details : change the condition to increase offset_idx
1512 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1514 Description: Data corruption for OSTs that are formatted larger than 2TB
1516 Details : When generating the bio request for lustre file writes the
1517 sector number would overflow a temporary variable before being
1518 used for the IO. The data reads correctly from Lustre (which
1519 will overflow in a similar manner) but other file data or
1520 filesystem metadata may be corrupted in some cases.
1524 Description: Tune the kernel for good SCSI performance.
1525 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1526 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1531 Description: rewrite ext2-derived code in obdclass/uuid.c
1532 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1533 from scratch preserving functionality.
1537 Description: "lfs find -obd UUID" prints directories
1538 Details : "lfs find -obd UUID" will return all directory names instead
1539 of just file names. It is incorrect because the directories
1540 do not reside on the OSTs.
1544 Description: use i_size_read and i_size_write in 2.6 port
1545 Details : replace inode->i_size access with i_size_read/write()
1548 Frequency : when removing large files
1550 Description: scheduling issue during removal of large Lustre files
1551 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1552 It causes scheduling issues when removing large files (17TB in the
1556 Frequency : only with liblustre clients on XT3
1558 Description: evictions taking too long
1559 Details : allow llrd to evict clients directly on OSTs
1562 Frequency : only on ppc
1564 Description: /proc/fs/lustre/devices broken on ppc
1565 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1566 We should make sure the type of 'index' is loff_t and then cast
1567 explicitly as needed below. Do not assign an explicitly cast
1571 Frequency : only for rhel5
1573 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1574 Details : Modify the target file & which_kernel.
1576 Severity : enhancement
1578 Description: omit set fsid for export NFS
1579 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1580 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1584 Frequency : Always, for kernels after 2.6.16
1585 Description: Fix warning idr_remove called for id=.. which is not allocated.
1586 Details : Last kernels save old s_dev before kill super and not allow
1587 to restore from callback - restore it before call kill_anon_super.
1591 Description: buffer overruns could theoretically occur
1592 Details : llapi_semantic_traverse() modifies the "path" argument by
1593 appending values to the end of the origin string, and a
1594 overrun may occur. Adding buffer overrun check in liblustreapi.
1598 Description: Fix error on 'ls .' at the top of the Lustre mount.
1599 Details : Don't revalidate dentry if it is a root dentry.
1603 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1604 Details : Modify vm-tunables-rhel4.patch.
1608 Description: Kernel config for 2.6.18-vanilla.
1609 Details : Modify targets/2.6-vanilla.target.in.
1610 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1611 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1612 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1613 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1617 Description: improve handling recoverable errors
1618 Details : if request processig with error which can be recoverable on server
1619 request should be resend, otherwise page released from cache and
1624 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1625 Details : Modify target file & which_patch.
1626 A flaw was found in the IA32 system call emulation provided
1627 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1628 value could be stored in the %RAX register, which could trigger an
1629 out-of-bounds system call table access. An untrusted local user
1630 could exploit this flaw to run code in the kernel
1631 (ie a root privilege escalation). (CVE-2007-4573).
1635 Description: O_DIRECT bypasses client statistics.
1636 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1637 read_ahead_stats were updated. copied the stats section from
1638 osc_send_oap_rpc() into async_internal().
1642 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1646 Description: change order of libsysio includes
1647 Details : '#include sysio.h' should always come before '#include xtio.h'
1651 Description: handle "serious error: objid * already exists" more gracefully
1652 Details : If LAST_ID value on disk is smaller than the objects existing in
1653 the O/0/d* directories, it indicates disk corruption and causes an
1654 LBUG(). If the object is 0-length, then we should use the existing
1655 object. This will help to avoid a full fsck in most cases.
1657 Severity : enhancement
1659 Description: adapt the lustre_config script to support the upgrade case
1660 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1661 server targets to 1.6 in parallel.
1665 Description: To avoid grant space > avaible space when the disk is almost
1666 full. Without this patch you might see the error "grant XXXX >
1667 available" or some LBUG about grant, when the disk is almost
1669 Details : In filter_check_grant, for non_grant cache write, we should
1670 check the left space by if (*left > ungranted + bytes), instead
1671 of (*left > ungranted), because only we are sure the left space
1672 is enough for another "bytes", then the ungrant space should be
1673 increase. In client, we should update cl_avail_grant only there
1674 is OBD_MD_FLGRANT in the reply.
1678 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1679 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1680 described in CVE-2007-4573. This problem could allow untrusted
1681 local users to gain root access.
1684 Frequency : when using O_DIRECT and quotas
1686 Description: Incorrect file ownership on O_DIRECT output files
1687 Details : block usage reported by 'lfs quota' does not take into account
1688 files that have been written with O_DIRECT.
1693 Description: touch file failed when fs is not full
1694 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1695 otherwise we can get ENOSP while fs is not full.
1699 Description: parallel lock callbacks
1700 Details : Instead of sending blocking and completion callbacks as separated
1701 requests, adding them to a set and sending in parallel.
1704 Frequency : only for Cray XT3
1705 Bugzilla : 12829/13455
1706 Description: Changing primary group doesn't change the group lustre assigns to
1708 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1709 overridden with the primary group provided by the group upcall,
1710 whereas the supplied fsgid can be trusted if it is in the list of
1711 supplementary groups returned by the group upcall.
1713 Severity : enhancement
1715 Description: Allow masking D_WARNING, D_ERROR messages from console
1716 Details : Console messages can now be disabled via lnet.printk.
1720 Description: User code with malformed file open parameter crashes client node
1721 Details : Before packing join_file req, all the related reference should be
1722 checked carefully in case some malformed flags cause fake
1723 join_file req on client.
1727 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1728 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1729 mode is equal to the granted mode. However, it should also check
1730 that LDLM_FL_CP_REQD is not set.
1734 Description: Heavy nfs access might result in deadlocks
1735 Details : After ELC code landed, it is now improper to enqueue any mds
1736 locks under och_sem, because enqueue might want to decide to
1737 cancel open locks for same inode we are holding och_sem for.
1741 Description: Client eviction while running blogbench
1742 Details : A lot of unlink operations with concurrent I/O can lead to a
1743 deadlock causing evictions. To address the problem, the number of
1744 oustanding OST_DESTROY requests is now throttled to
1745 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1746 callbacks are processed in priority.
1750 Description: enable ACLs on MDS by default
1751 Details : ACLs must be enabled on MDS by default.
1754 Frequency : PPC/PPC64 only
1756 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1757 Details : fix duplicated definitions between asm-ppc64/types.h and
1758 lustre_types.h on PPC.
1761 Frequency : PPC/PPC64 only
1763 Description: asm-ppc/segment.h does not exist
1764 Details : fix compile issue on PPC.
1768 Description: better handle error messages in extents code
1771 Frequency : RHEL4 only
1773 Description: mkfs is very slow on IA64/RHEL4
1774 Details : A performance regression has been discovered in the MPT Fusion
1775 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1776 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1777 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1779 Severity : enhancement
1781 Description: SNMP support enhancement
1782 Details : Adding total number of sampled request for an MDS node in snmp
1785 Severity : enhancement
1787 Description: Optimize ldlm waiting list processing for PR extent locks
1788 Details : When processing waiting list for read extent lock and meeting read
1789 lock that is same or wider to it that is not contended, skip
1790 processing rest of the list and immediatelly return current
1791 status of conflictness, since we are guaranteed there are no
1792 conflicting locks in the rest of the list.
1796 Description: Time out and refuse to reconnect
1797 Details : When the failover node is the primary node, it is possible
1798 to have two identical connections in imp_conn_list. We must
1799 compare not conn's pointers but NIDs, otherwise we can defeat
1800 connection throttling.
1804 Description: port llog fixes from b1_6 into HEAD
1805 Details : Port llog reference couting and some llog cleanups from b1_6
1806 (bug 10800) into HEAD, for protect from panic and access to already
1807 free llog structures.
1811 Description: Detect stride IO mode in read-ahead
1812 Details : When a client does stride read, read-ahead should detect that and
1813 read-ahead pages according to the detected stride pattern.
1817 Description: data checksumming impacts single node performance
1818 Details : add support for several checksum algorithm. Currently, only CRC32
1819 and Adler-32 are supported. The checksum type can be changed on
1820 the fly via /proc/fs/lustre/osc/*/checksum_type.
1824 Description: use adler32 for page checksums
1825 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1830 Description: build for x2 fails
1831 Details : fix compile issue on Cray systems.
1835 Description: Properly match for duplicate locks
1836 Details : Due to different lock order from skiplists code, we need to
1837 traverse entire list for now
1840 Frequency : only on PPC/SLES10
1842 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1843 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1844 be defined. Add a hack in mkfs_lustre.c to work around this compile
1849 Description: LASSERT on MDS when client holding flock lock dies
1850 Details : ldlm pool logic depends on number of granted locks equal to
1851 number of released locks which is not true for flock locks, so
1852 just exclude such locks from consideration.
1856 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1857 Details : Use fsfilt_sync() for both the whole filesystem sync and
1858 individual file sync to eliminate dangerous inode locking
1859 with I_LOCK that can lead to a deadlock.
1863 Description: performance in 1.6.3
1864 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1865 sure that 1MB requests are not fragmented by the block layer.
1867 Severity : enhancement
1869 Description: organize the server-side client stats on per-nid basis
1870 Details : Change the structure of stats under obdfilter and mds to
1878 The "uuid"s file would list the uuids of _active_ exports.
1879 And the clear entry is to clear all stats and stale nids.
1881 Severity : enhancement
1883 Description: eliminate client locks in face of contention
1884 Details : file contention detection and lockless i/o implementation
1885 for contended files.
1889 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1892 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1894 Description: MDS may lose file striping (and hence file data) in some cases
1895 Details : If there are additional extended attributes stored on the MDS,
1896 in particular ACLs, SELinux, or user attributes (if user_xattr
1897 is specified for the client mount options) then there is a risk
1898 of attribute loss. Additionally, the Lustre file striping
1899 needs to be larger than default (e.g. striped over all OSTs),
1900 and an additional attribute must be stored initially in the
1901 inode and then increase in size enough to be moved to the
1902 external attribute block (e.g. ACL growing in size) for file
1907 Description: skiplist implementation simplification
1908 Details : skiplists are used to group compatible locks on granted list
1909 that was implemented as tracking first and last lock of each
1910 lock group the patch changes that to using doubly linked lists
1914 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1915 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1916 because LASSERT on both the data supplied by a client, and the
1917 data on disk is dangerous and incorrect.
1919 Severity : enhancement
1921 Description: Slow truncate/writes to huge files at high offsets.
1922 Details : Directly associate cached pages to lock that protect those pages,
1923 this allows us to quickly find what pages to write and remove
1924 once lock callback is received.
1928 Description: more ldlm soft lockups
1929 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1930 starve other threads from the resource lock for a long time in
1931 case of long waiting queue, so change the debug level from
1932 D_OTHER to the less frequently used D_INFO.
1934 Severity : enhancement
1936 Description: add -gid, -group, -uid, -user options to lfs find
1940 Description: Hung threads in invalidate_inode_pages2_range
1941 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1942 once one is completed. As a result, some RPCs are stuck in the
1943 queue and are never sent.
1947 Description: filter threads hungs on waiting journal commit
1948 Details : Cleanup filter group llog code, then only filter group llog will
1949 be only created in the MDS/OST syncing process.
1953 Description: Procfs and llog threads access destoryed import sometimes.
1954 Details : Sync the import destoryed process with procfs and llog threads by
1955 the import refcount and semaphore.
1957 Severity : enhancement
1959 Description: openlock cache of b1_6 port to HEAD
1964 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1965 Details : If insertion of an extent fails, then discard the inode
1966 preallocation and free data blocks else it can lead to duplicate
1971 Description: don't always update ctime in ext3_xattr_set_handle()
1972 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1973 In some cases the ctime should not be updated, for example for
1974 2.0->1.8 compatibility it is necessary to delete an xattr and it
1975 should not update the ctime.
1979 Bugzilla : 15713/16362
1980 Description: Assertion in iopen_connect_dentry in 1.6.3
1981 Details : looking up an inode via iopen with the wrong generation number can
1982 populate the dcache with a disconneced dentry while the inode
1983 number is in the process of being reallocated. This causes an
1984 assertion failure in iopen since the inode's dentry list contains
1985 both a connected and disconnected dentry.
1989 Description: assertion failure in ldlm_handle2lock()
1990 Details : fix a race between class_handle_unhash() and class_handle2object()
1991 introduced in lustre 1.6.5 by bug 13622.
1996 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1997 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2002 Description: oops in sd_iostats_seq_show()
2003 Details : unloading/reloading the scsi low level driver triggers a kernel
2004 bug when trying to access the sd iostat file.
2009 Description: Kernel panics during QLogic driver reload
2010 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2011 patch, causing memory corruption.
2016 Description: journal_dev option does not work in b1_6
2017 Details : pass mount option during pre-mount.
2019 Severity : enhancement
2021 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2022 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2023 extent information of a file. It can be used to map logical blocks
2024 in a file to physical blocks in the block device.
2028 Description: LDLM soft lockups - improvement
2029 Details : It is be possible to send the lock handle along with each read
2030 or write request because the client is already doing a lock match
2031 itself so there isn't any reason the OST should have to re-do that
2035 Frequency : only on Cray X2
2037 Description: X2 build failures
2038 Details : fix build failures on Cray X2.
2042 Description: xid & resent requests
2043 Details : Initialize RPC XID from clock at startup (randomly if clock is
2046 Severity : enhancement
2048 Description: Add lustre_start utility to start or stop multiple Lustre servers
2053 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2054 Details : In case of memory pressure, list_del() can be called twice on
2055 req->rq_history_list, causing a kernel oops.
2059 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2060 Details : fix stack overflow in the distributed lock manager by defering
2061 export eviction after a failed AST to the elt thread instead of
2062 handling it in the dlm interpret routine.
2066 Description: Convert some comments to new format.
2067 Details : Update documenting comments to match doxygen conventions.
2071 Description: Grammar fixes.
2072 Details : A couple of trivial spelling fixes.
2076 Description: OSD_COUNTERS-mandatory
2077 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2078 to simplify the code.
2082 Description: simplify cmm_device freeing logic.
2083 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2087 Description: Add lockdep support to dt_object_operations locking interface.
2088 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2089 parameter indicating lock ordering. Update mdd code to use new
2094 Description: Introduce failloc constants for lockless IO tests.
2095 Details : Add two new failloc constants to test lockless IO. Only one of
2096 them in implemented---another is checked in yet to be landed
2101 Description: Add lockdep support for inode mutex.
2102 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2103 in the situations where Lustre has to lock more than one inode
2108 Description: Add optional invariants checking support.
2109 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2110 configure switch. This macro is to be used for consistency and
2111 sanity checks that are too expensive to be left in `production'
2116 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2117 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2122 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2123 Details : Use spin_lock_nested() in (the only) situation where more than
2124 one ldlm_lock is locked simultaneously. Also, fix possible
2125 dead-lock in ldlm_lock_change_resource() by enforcing particular
2130 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2131 Details : Instead of specifying each ldlm_lock call-back through separate
2132 parameter, wrap them into struct ldlm_callback_suite.
2136 Description: Kill join_lru obd method and its callers.
2137 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2138 in memory---a requirement implemented through ->o_join_lru() obd
2139 method in HEAD. Remove this method and its users.
2143 Description: Add asynchronous ldlm ENQUEUE completion handler.
2144 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2145 case that stock ldlm_completion_ast() cannot handle as it waits
2146 until lock is granted. Introduce new ldlm_completion_ast_async()
2147 for this. Also comment ldlm_completion_ast().
2151 Description: ldlm_error <-> errno conversion.
2152 Details : Add functions to map (rather arbitrary) between LDLM error codes
2153 and standard errno values. CLIO needs this to prevent LDLM specific
2154 constants from escaping ldlm and osc.
2158 Description: Kill unused ldlm_handle2lock_ns() function.
2159 Details : Kill unused ldlm_handle2lock_ns() function.
2163 Description: Add lu_ref support to ldlm_lock
2164 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2165 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2166 and ldlm_resource. LDLM interface has to be changed, because code that
2167 releases a reference on a lock, has to "know" what reference this is.
2168 In the most frequent case
2170 lock = ldlm_handle2lock(handle);
2172 LDLM_LOCK_PUT(lock);
2174 no changes are required. When any other reference (received _not_ from
2175 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2176 instead of LDLM_LOCK_PUT().
2178 Arguably, changes are pervasive, and interface requires some discipline
2179 for proper use. On the other hand, it was very instrumental in finding
2180 a few leaked lock references.
2184 Description: Add ldlm_lock_addref_try().
2185 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2186 attempts to addref a lock that might be being canceled
2191 Description: Add ldlm_weigh_callback().
2192 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2193 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2194 hard-coded `number of pages' logic.
2198 Description: Add lockdep annotations to llog code.
2199 Details : Use appropriately tagged _nested() locking calls in the places
2200 where llog takes more than one ->lgh_lock lock.
2204 Description: Add loi_kms_set().
2205 Details : Wrap kms updates into a helper function.
2209 Description: Constify instances of struct lsm_operations.
2210 Details : Constify instances of struct lsm_operations.
2214 Description: lu_conf support.
2215 Details : On a server, a file system object is uniquely identified
2216 by a fid, which is sufficient to locate and load all object
2217 state (inode). On a client, on the other hand, more data are
2218 necessary instantiate an object. Change lu_object_find() and
2219 friends to take additional `lu_conf' argument describing object.
2220 Typically this includes layout information.
2224 Description: lu_context fixes.
2225 Details : Introduce new lu_context functions that are needed on the client
2226 side, where some system threads (ptlrpcd) are shared by multiple
2227 modules, and so cannot be stopped during module shutdown.
2231 Description: Add start and stop methods to lu_device_type_operations.
2232 Details : Introduce two new methods in lu_device_type_operations, that are
2233 invoked when first instance of a given type is created and last one
2234 is destroyed respectively. This is need by CLIO.
2238 Description: Add lu_ref support to struct lu_device.
2239 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2240 track leaked references.
2244 Description: Introduce lu_kmem_descr.
2245 Details : lu_kmem_descr and its companion interface allow to create
2246 and destroy a number of kmem caches at once.
2250 Description: Fix lu_object finalization race.
2251 Details : Fix a race between lu_object_find() finding an object and its
2252 concurrent finalization. This race is (most likely) not possible
2253 on the server, but might happen on the client.
2257 Description: Introduce lu_ref interface.
2258 Details : lu_ref is a debugging module allowing to track references to
2259 a given object. It is quite cpu expensive, and has to be
2260 explicitly enabled with --enable-lu_ref. See usage description
2265 Description: Factor lu_site procfs stats into a separate function.
2266 Details : Separate lu_site stats printing code into a separate function
2267 that can be reused on a client.
2271 Description: Constify instances of struct {lu,dt,md}_device_operations.
2272 Details : Constify instances of struct {lu,dt,md}_device_operations.
2276 Description: Introduce struct md_site and move meta-data specific parts of
2277 struct lu_site here.
2278 Details : Move md-specific fields out of struct lu_site into special struct
2279 md_site, so that lu_site can be used on a client.
2283 Description: Kill mdd_lov_destroy().
2284 Details : Remove unused mdd code.
2288 Description: Add st_block checking to multistat.c.
2289 Details : Add st_block checking to multistat.c.
2293 Description: Add lu_ref support to struct obd_device.
2294 Details : Add lu_ref tracking to obd_device.
2298 Description: Kill obd_set_fail_loc().
2299 Details : Remove unused code.
2303 Description: Add special type for ptlrpc_request interpret functions.
2304 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2305 there. Actual usage will be in CLIO.
2309 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2310 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2311 doesn't work with lockdep.
2315 Description: Add rwv.c test program.
2316 Details : New testing program exercising readv(2) and writev(2) (Qian).
2320 Description: Add sendfile.c test program.
2321 Details : New testing program exercising sendfile(2) (Jay).
2325 Description: Ratelimit a message that can be very frequent.
2326 Details : Ratelimit a memory allocation failure message that can
2331 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2332 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2336 Description: Liblustre build fixes.
2337 Details : Liblustre build fixes.
2341 Description: libcfs: add cfs_{need,cond}_resched() interface.
2342 Details : libcfs: add cfs_{need,cond}_resched() definition and
2343 implementations for Linux, NT, and liblustre.
2345 Severity : enhancement
2347 Description: More exported tunables for mballoc
2348 Details : Add support for tunable preallocation window and new tunables for
2349 large/small requests
2353 Description: Detect corruption of block bitmap and checking for preallocations
2354 Details : Checks validity of on-disk block bitmap. Also it does better
2355 checking of number of applied preallocations. When corruption is
2356 found, it turns filesystem readonly to prevent further corruptions.
2360 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2361 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2362 window to 0. If it does hit cache miss, and read-pattern is still
2363 stride-io mode, does not reset the stride window, but also does not
2364 increase the stride window length in this case.
2368 Frequency : only for big-endian servers
2369 Description: Check if system is big-endian while mounting fs with extents feature
2370 Details : Mounting a filesystem with extents feature will fail on big-endian
2371 systems since ext3-based ldiskfs is not supported on big-endian
2372 systems. This can be over-riden with "bigendian_extents" mount option.
2374 Severity : enhancement
2376 Description: The root squash functionality
2377 Details : A security feature, which is to prevent users from being able
2378 to mount lustre on their desktop, run as root, and delete
2379 all of the files in the filesystem. The goal is accomplished by
2380 remapping user id (UID) and group id (GID) of the root user to
2381 a UID and GID specified by the system administartor via Lustre
2382 configuration management server (MGS). The functionality also
2383 allows to specify sets of clients for which the remapping does
2388 Description: Excessive recovery window
2389 Details : With AT enabled, the recovery window can be excessively long (6000+
2390 seconds). To address this problem, we no longer use
2391 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2392 timeout no longer depends on the service time, it is set to
2393 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2394 time via pb_service_time.
2398 Description: Watchdog triggered on MDS failover
2399 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2400 the OSTs know that the MDS "UUID" can be reused for the same export
2401 from a different NID, so we do not need to wait for the export to be
2405 Frequency : rare, only if using MMP with Linux RAID
2407 Description: MMP doesn't work with Linux RAID
2408 Details : While using HA for Lustre servers with Linux RAID, it is possible
2409 that MMP will not detect multiple mounts. To make this work we
2410 need to unplug the device queue in RAID when the MMP block is being
2411 written. Also while reading the MMP block, we should read it from
2412 disk and not the cached one.
2414 Severity : enhancement
2416 Description: open file using fid
2417 Details : A file can be opened using just its fid, like
2418 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2421 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2424 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2425 Details : add proper locking for creation of jbd_4k slab cache
2429 Description: MMP check in ext3_remount() fails without displaying any error
2430 Details : When multiple mount protection fails during remount, proper error
2433 Severity : enhancement
2435 Description: Allow stripe size to be up to 4G-64k
2436 Details : Fix math logic to allow large stripe sizes.
2440 Description: add check for >8TB ldiskfs filesystems
2441 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2442 Don't allow >8TB ldiskfs filesystems to be mounted without
2443 force_over_8tb mount option
2448 Descriptoin: OSS complains "too many blocks" and crashes
2449 Details : In filter_commitrw_write():
2450 1. lquota_chkquota();
2451 2. obdo_from_inode();
2452 3. lquota_pending_commit();
2453 "2." has changed uid and gid in this case, then caused
2454 lquota_chkquota() and lquota_pending_commit() to use
2455 different uid and gid. That is the root of the bug.
2459 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2460 Details : Detect and print out the 'stripe count = -1' case properly.
2464 Description: Out or order replies might be lost on replay
2465 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2466 request with tid smaller then one currently being added, add it
2467 to the start, not end of the list.
2469 --------------------------------------------------------------------------------
2471 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2473 * Support for kernels:
2474 2.6.5-7.283 (SLES 9),
2475 2.6.9-55.EL (RHEL 4),
2476 2.6.16.46-0.14 (SLES 10),
2477 2.6.18.8 vanilla (kernel.org)
2478 * Client support for unpatched kernels:
2479 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2480 2.6.16 - 2.6.22 vanilla (kernel.org)
2481 * Due to recently discovered recovery problems, we do not recommend
2482 using patchless RHEL 4 clients with this or any earlier release.
2483 * Recommended e2fsprogs version: 1.39.cfs8
2484 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2485 * Starting with this release, the ldiskfs backing filesystem required
2486 by Lustre is now in its own package, lustre-ldiskfs. This package
2487 should be installed. It is versioned separately from Lustre and
2488 may be released separately in future.
2492 Description: block reactivating mgc import until all deactivates complete
2493 Details : Fix race when failing back MDT/MGS to itself (testing)
2496 Frequency : at statup only
2498 Description: mds_lov_synchronize race leads to various problems
2499 Details : simultaneous MDT->OST connections at startup can cause the
2500 sync to abort, leaving the OSC in a bad state.
2502 Severity : enhancement
2504 Description: add optional extra BUILD_VERSION info
2505 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2506 to override the lustre version.
2509 Frequency : 2.6.18 servers only
2511 Description: ll_kern_mount() doesn't release the module reference
2512 Details : The ldiskfs module reference count never drops down to 0
2513 because ll_kern_mount() doesn't release the module reference.
2518 Description: server LBUG when using old ost_num_threads parameter
2519 Details : Accept the old ost_num_threads parameter but warn that it
2520 is deprecated, and fix an off-by-one error that caused an LBUG.
2525 Description: Transient SCSI error results in persistent IO issue
2526 Details : iobuf->dr_error is not reinitialized to 0 between two
2530 Frequency : sometimes when underlying device returns I/O errors
2532 Description: OSTs not going read-only during write failures
2533 Details : OSTs are not remounted read-only when the journal commit threads
2534 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2535 instead of the ext3 wrappers.
2539 Description: poor connect scaling with increasing client count
2540 Details : Don't run filter_grant_sanity_check for more than 100 exports
2541 to improve scaling for large numbers of clients.
2544 Frequency : SLES10 only
2546 Description: sanity-quota.sh quotacheck failed: rc = -22
2547 Details : Quotas cannot be enabled on SLES10.
2550 Frequency : liblustre clients only
2552 Description: getdirentries does not give error when run on compute nodes
2553 Details : getdirentries does not fail when the size specified as an argument
2554 is too small to contain at least one entry
2556 Severity : enhancement
2558 Description: Add LNET router traceability for debug purposes
2559 Details : If a checksum failure occurs with a router as part of the
2560 IO path, the NID of the last router that forwarded the bulk data
2561 is printed so it can be identified.
2566 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2567 Details : Due to a race condition, liblustre clients were occasionally
2568 evicted incorrectly.
2570 Severity : enhancement
2572 Description: lfs setstripe use optional parameters instead of postional
2575 Severity : enhancement
2577 Description: Nanosecond timestamp support for ldiskfs
2578 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2579 resolution timestamps. There is not yet support for this at
2580 the Lustre filesystem level.
2583 Frequency : during server recovery
2585 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2586 Details : request with rq_no_resend flag not awake l_wait_event if they get
2590 Frequency : nfs export on patchless client
2592 Description: connectathon hang when test nfs export over patchless client
2593 Details : Disconnected dentry cannot be found with lookup, so we do not need
2594 to unhash it or make it invalid
2597 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2598 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2599 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2604 Description: open req refcounting wrong on reconnect
2605 Details : If reconnect happened between getting open reply from server and
2606 call to mdc_set_replay_data in ll_file_open, we will schedule
2607 replay for unreferenced request that we are about to free.
2608 Subsequent close will crash in variety of ways.
2609 Check that request is still eligible for replay in
2610 mdc_set_replay_data().
2615 Description: disable writes to filesystem when reading health_check file
2616 Details : the default for reading the health_check proc file has changed
2617 to NOT do a journal transaction and write to disk, because this
2618 can cause reads of the /proc file to hang and block HA state
2619 checking on a healthy but otherwise heavily loaded system. It
2620 is possible to return to the previous behaviour during configure
2621 with --enable-health-write.
2623 Severity : enhancement
2625 Description: 64-bit inode version
2626 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2627 to the inode. This will be required for version-based recovery.
2632 Description: MDS fails to start if a duplicate client export is detected
2633 Details : in some rare cases it was possible for a client to connect to
2634 an MDS multiple times. Upon recovery the MDS would detect this
2635 and fail during startup. Handle this more gracefully.
2637 Severity : enhancement
2639 Description: Add -o localflock option to simulate old noflock
2641 Details : This will achieve local-only flock/fcntl locks
2647 Description: log_commit_thread vs filter_destroy race leads to crash
2648 Details : Take import reference before releasing llog record semaphore
2653 Description: Wrong request locking in request set processing
2654 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2655 imp_delayed_list, in this place should be used imp_lock.
2658 Frequency : when reconnection
2660 Description: Grant Leak when osc reconnect to OST
2661 Details : When osc reconnect ost, OST(filter) should check whether it
2662 should grant more space to client by comparing fed_grant and
2663 cl_avail_grant, and return the granted space to client instead
2664 of "new granted" space, because client will call osc_init_grant
2665 to update the client grant space info.
2668 Frequency : when client reconnect to OST
2670 Description: Grant Leak when osc do resend and replay bulk write
2671 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2672 bulk write request, because the grant info while be sync between
2673 OSC and OST when reconnect, and we should ignore the grant info
2674 these of resend/replay write req.
2679 Description: Grant space more than avaiable left space sometimes.
2680 Details : When then OST is about to be full, if two bulk writing from
2681 different clients came to OST. Accord the avaliable space of the
2682 OST, the first req should be permitted, and the second one
2683 should be denied by ENOSPC. But if the seconde arrived before
2684 the first one is commited. The OST might wrongly permit second
2685 writing, which will cause grant space > avaiable space.
2688 Frequency : when client is evicted
2690 Description: Grant might be wrongly erased when osc is evicted by OST
2691 Details : when the import is evicted by server, it will fork another
2692 thread ptlrpc_invalidate_import_thread to invalidate the
2693 import, where the grant will be set to 0. While the original
2694 thread will update the grant it got when connecting. So if
2695 the former happened latter, the grant will be wrongly errased
2696 because of this race.
2701 Description: Checking Stale with correct fid
2702 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2703 to check whether it is stale, because sometimes, we want the
2704 enqueue happened anyway, and op_data.fid2 will not be initialized.
2706 Severity : enhancement
2708 Description: update patchless client
2709 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2712 Frequency : only with 2.4 kernel
2714 Description: random memory corruption
2715 Details : size of struct ll_inode_info is to big for union inode.u and this
2716 can be cause of random memory corruption.
2721 Description: Memory leak in recovery
2722 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2723 It should also check obd_fail before fsfilt_start, otherwise if
2724 fsfilt_start return -EROFS,(failover mds during mds recovery).
2725 then the req will return with repmsg->transno = 0 and rc = EROFS.
2726 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2727 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2728 be freed no matter whether fsfilt_commit success or not.
2731 Frequency : only with huge count clients
2733 Description: Prevents from taking the superblock lock in llap_from_page for
2735 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2736 for llap_from_page call in ll_removepage prevents from taking the
2737 superblock lock for a soon died page.
2742 Description: Not check open intent error before release open handle
2743 Details : in some rare cases, the open intent error is not checked before
2744 release open handle, which may cause
2745 ASSERTION(open_req->rq_transno != 0), because it tries to release
2746 the failed open handle.
2751 Description: Set cat log bitmap only after create log success.
2752 Details : in some rare cases, the cat log bitmap is set too early. and it
2753 should be set only after create log success.
2757 Description: Accessing a block bevice can re-enable I/O when Lustre is
2758 tearing down a device.
2759 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2764 Description: service threads may hog cpus when there are a lot of requests
2765 Details : Insert cond_resched to give other threads a chance to use some CPU
2770 Description: the cat log was not initialized in recovery
2771 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2772 unlink log on mds will not be initialized until mds post
2773 recovery. And also in mds post recovery, the unlink log will
2774 initialization will be done asynchronausly, so there will be race
2775 between add unlink log and unlink log initialization.
2779 Description: brw_stats were being printed incorrectly
2780 Details : brw_stats were being printed as log2 but all of them were not
2781 recorded as log2. Also remove some code duplication arising from
2782 filter_tally_{read,write}.
2786 Frequency : rare, only in recovery.
2787 Description: ASSERTION(req->rq_type != LI_POISON) failed
2788 Details : imp_lock should be held while iterating over imp_sending_list for
2789 prevent destroy request after get timeout in ptlrpc_queue_wait.
2793 Description: replay-single.sh test 52 fails
2794 Details : A lock's skiplist need to be cleanup when it being unlinked
2795 from its resource list.
2799 Description: Short directio read returns full requested size rather than
2801 Details : Direct I/O operations should return actual amount of bytes
2802 transferred rather than requested size.
2804 Severity : enhancement
2806 Description: metadata RPC reduction (e.g. for rm performance)
2807 Details : decrease the amount of synchronous RPC between clients and servers
2808 by canceling conflicing lock before the operation on the client
2809 and packing thier handles into the main operation RPC to server.
2811 Severity : enhancement
2813 Description: Async OSC create to avoid the blocking unnecessarily.
2814 Details : If a OST has no remain object, system will block on the creating
2815 when need to create a new object on this OST. Now, ways use
2816 pre-created objects when available, instead of blocking on an
2817 empty osc while others are not empty. If we must block, we block
2818 for the shortest possible period of time.
2822 Description: improve handling recoverable errors
2823 Details : if request processig with error which can be recoverable on server
2824 request should be resend, otherwise page released from cache and
2827 Severity : enhancement
2829 Description: refine locking for avoid write wrong info into lov_objid file
2830 Details : fix possible races with add new target and write/update data in
2833 --------------------------------------------------------------------------------
2835 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2840 Frequency : on some architectures
2842 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2843 Details : Uninitialized flags sometimes cause configuration commands to
2847 Frequency : patchless clients only
2849 Description: missing __iget() symbol export
2850 Details : The __iget() symbol export is missing. To avoid the need for
2851 this on patchless clients the deathrow inode reaper is turned
2852 off, and we depend on the VM to clean up old inodes. This
2853 dependency was during via the fix for bug 12181.
2855 --------------------------------------------------------------------------------
2857 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2859 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2860 INTEROPERATE with older versions automatically. In many cases a
2861 special upgrade step is needed. Please read the
2862 user documentation before upgrading any part of a live system.
2863 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2864 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2865 * WARNING: Lustre configuration and startup changes are required with
2866 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2870 Severity : enhancement
2872 Description: MountConf
2873 Details : Lustre configuration is now managed via mkfs and mount
2874 commands instead of lmc and lconf. New obd types (MGS, MGC)
2875 are added for dynamic configuration management. See
2876 https://mail.clusterfs.com/wikis/lustre/MountConf for
2879 Severity : enhancement
2881 Description: dynamic OST addition
2882 Details : OSTs can now be added to a live filesystem
2884 Severity : enhancement
2886 Description: startup order invariance
2887 Details : MDTs and OSTs can be started in any order. Clients only
2888 require the MDT to complete startup.
2890 Severity : enhancement
2892 Description: parallel, asynchronous orphan cleanup
2893 Details : orphan cleanup is now performed in separate threads for each
2894 OST, allowing parallel non-blocking operation.
2896 Severity : enhancement
2898 Description: optimized stripe assignment
2899 Details : stripe assignments are now made based on ost space available,
2900 ost previous usage, and OSS previous usage, in order to try
2901 to optimize storage space and networking resources.
2903 Severity : enhancement
2905 Description: Permanently set tunables
2906 Details : All writable /proc/fs/lustre tunables can now be permanently
2907 set on a per-server basis, at mkfs time or on a live system.
2909 Severity : enhancement
2911 Description: Lustre message v2
2912 Details : Add lustre message format v2.
2914 Severity : enhancement
2916 Description: client OST exclusion list
2917 Details : Clients can be started with a list of OSTs that should be
2918 declared "inactive" for known non-responsive OSTs.
2922 Description: SPEC SFS validation failure on NFS v2 over lustre.
2923 Details : Changes the blocksize for regular files to be 2x RPC size,
2924 and not depend on stripe size.
2926 Severity : enhancement
2928 Description: Multiple MD RPCs in flight.
2929 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2930 To avoid overly-overloading MDS, introduce a limit on number of
2931 MDS RPCs in flight for a single client and add /proc controls
2932 to adjust this limit.
2934 Severity : enhancement
2936 Description: client read/write statistics
2937 Details : Add client read/write call usage stats for performance
2938 analysis of user processes.
2939 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2940 file access. extents_stats shows chunk size distribution.
2941 extents_stats_per_process show chunk size distribution per
2944 Severity : enhancement
2946 Description: mds statistics
2947 Details : Add detailed mds operations statistics in
2948 /proc/fs/lustre/mds/*/stats.
2952 Description: Failure of copying files with lustre special EAs.
2953 Details : Client side always return success for setxattr call for lustre
2954 special xattr (currently only "trusted.lov").
2959 Description: Refcount LNET uuids
2960 Details : The global LNET uuid list grew linearly with every startup;
2961 refcount repeated list entries instead of always adding to
2964 Severity : enhancement
2966 Description: Dynamic service threads
2967 Details : Within a small range, start extra service threads
2968 automatically when the request queue builds up.
2971 Frequency : mixed-endian client/server environments
2973 Description: mixed-endian crashes
2974 Details : The new msg_v2 system had some failures in mixed-endian
2977 Severity : enhancement
2979 Description: Easy OST removal
2980 Details : OSTs can be permanently deactivated with e.g. 'lctl
2981 conf_param lustre-OST0001.osc.active=0'
2983 Severity : enhancement
2985 Description: MGS proc entries
2986 Details : Added basic proc entries for the MGS showing what filesystems
2989 Severity : enhancement
2991 Description: provide MGS failover
2992 Details : Added config lock reacquisition after MGS server failover.
2994 Severity : enhancement
2996 Description: add Linux 2.4 support
2997 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3001 Description: a large application tries to do I/O to the same resource and dies
3002 in the middle of it.
3003 Details : Check the req->rq_arrival time after the call to
3004 ost_brw_lock_get(), but before we do anything about
3005 processing it & sending the BULK transfer request. This
3006 should help move old stale pending locks off the queue as
3007 quickly as obd_timeout.
3010 Frequency : when an incorrect nid is specified during startup
3012 Description: ptlrpc connect to non-existant node causes kernel crash
3013 Details : LNET can't be re-entered from an event callback, which
3014 happened when we expire a message after the export has been
3015 cleaned up. Instead, hand the zombie cleanup off to another
3018 Severity : enhancement
3020 Description: plain/inodebits lock performance improvement
3021 Details : Grouping plain/inodebits in granted list by their request modes
3022 and bits policy, thus improving the performance of search through
3026 Frequency : only if OST filesystem is corrupted
3028 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3029 Details : for a short time RPCs with bulk IO are in the replay list,
3030 but replay of bulk IOs is unimplemented. If the OST filesystem
3031 is corrupted due to disk cache incoherency and then replay is
3032 started it is possible to trip an assertion. Avoid putting
3033 committed RPCs into the replay list at all to avoid this issue.
3036 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3038 Description: System hang on startup
3039 Details : This bug allowed the liblustre (e.g. catamount) client to
3040 return to the app before handling all startup RPCs. This
3041 could leave the node unresponsive to lustre network traffic
3042 and manifested as a server ptllnd timeout.
3044 Severity : enhancement
3046 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3047 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3048 boolean to control whether to print peer debug info when a
3049 client's RPC times out.
3052 Frequency : only for kernels with patches from Lustre below 1.4.3
3054 Description: Remove old rdonly API
3055 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3058 Frequency : only for devices with external journals
3060 Description: Set external device read-only also
3061 Details : During a commanded failover stop, we set the disk device
3062 read-only while the server shuts down. We now also set any
3063 external journal device read-only at the same time.
3066 Frequency : when upgrading from 1.4 while trying to change parameters
3068 Description: The wrong (new) MDC name was used when setting parameters for
3069 upgraded MDT's. Also allows changing of OSC (and MDC)
3070 parameters if --writeconf is specified at tunefs upgrade time.
3073 Frequency : when setting specific ost indicies
3075 Description: QOS code breaks on skipped indicies
3076 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3077 created with --index need not be sequential.
3079 Severity : enhancement
3081 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3082 Details : The uninit_groups feature works in conjunction with the kernel
3083 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3084 up the pass1 processing of e2fsck. This is a read-only feature
3085 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3086 mount filesystems that have had this feature enabled.
3088 Severity : enhancement
3090 Description: Improve multi-block allocation algorithm to avoid fragmentation
3091 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3092 allocation locality and avoid filesystem fragmentation.
3094 ------------------------------------------------------------------------------
3096 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3098 * Support for kernels:
3099 2.4.21-47.0.1.EL (RHEL 3)
3100 2.6.5-7.283 (SLES 9)
3101 2.6.9-42.0.10.EL (RHEL 4)
3102 2.6.12.6 vanilla (kernel.org)
3103 2.6.16.27-0.9 (SLES 10)
3104 * Recommended e2fsprogs version: 1.39.cfs5
3106 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3110 Frequency : occasional, depends on client load and configuration
3111 Bugzilla : 12181, 12203
3112 Description: data loss for recently-modified files
3114 Details : In some cases it is possible that recently written or created
3115 files may not be written to disk in a timely manner (this should
3116 normally be within 30s unless client IO load is very high).
3117 The problem appears as zero-length files or files that are a
3118 multiple of 1MB in size after a client crash or client eviction
3119 that are missing data at the end of the file.
3121 This problem is more likely to be hit on clients where files are
3122 repeatedly created and unlinked in the same directory, clients
3123 have a large amount of RAM, have many CPUs, the filesystem has
3124 many OSTs, the clients are rebooted frequently, and/or the files
3125 are not accessed by other nodes after being written.
3127 The presence of the problem can be detected by looking at
3128 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3129 smaller than the second (nr_unused) then dirty files will not
3130 be flushed automatically to disk. "sync; sleep 10" should be
3131 run several times on the node before unmounting it to update
3132 Lustre (this is also safe to run on nodes without this problem).
3134 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3135 that can cause this same problem, so customers using that kernel
3136 also need to update the kernel in addition to Lustre. In order
3137 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3139 It is normal that files written just before a client crash (less
3140 than 30s) may not yet have been flushed to disk, even for local
3144 Frequency : frequent on thin XT3 nodes
3146 Description: UUID collision on thin XT3 Linux nodes
3147 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3148 randomly, since we relied on an insufficiently-seeded PRNG.
3153 Description: OSS hangs after "All ost request buffers busy"
3154 Details : A deadlock between quota and journal operations caused OSS
3155 hangs after printing "All ost request buffers busy."
3158 Frequency : always on liblustre builds
3160 Description: Cleanup compiler warnings on liblustre
3163 Frequency : always on liblustre builds on XT3
3165 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3169 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3170 > 32000 subdirectories
3171 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3172 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3173 subdirectory count crosses 32000. This will aid e2fsck to
3174 correctly handle more than 32000 subdirectories.
3177 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3179 Description: System hang on startup
3180 Details : This bug allowed the liblustre (e.g. catamount) client to
3181 return to the app before handling all startup RPCs. This
3182 could leave the node unresponsive to lustre network traffic
3183 and manifested as a server ptllnd timeout.
3185 Severity : enhancement
3187 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3188 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3189 boolean to control whether to print peer debug info when a
3190 client's RPC times out.
3195 Description: make O_SYNC working on 2.6 kernels
3196 Details : 2.6 kernels use different method for mark pages for write,
3197 so need add a code to lustre for O_SYNC work.
3202 Description: Failure to close file and release space on NFS
3203 Details : Put inode details into lock acquired in ll_intent_file_open.
3204 Use mdc_intent_lock in ll_intent_open to properly
3205 detect all kind of errors unhandled by mdc_enqueue.
3210 Description: proc file read during shutdown sometimes raced obd removal,
3212 Details : Add lock to prevent obd access after proc file removal.
3215 Frequency : Only for files larger than 4GB on 32-bit clients.
3217 Description: improperly doing page alignment of locks
3218 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3219 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3224 Description: RPCs being resent when they shouldn't be
3225 Details : Some RPCs that should not be resent are being resent. This
3226 can cause inconsistencies in the RPC state machine. Do not
3227 resend such requests.
3230 Frequency : rare, only with NFS export
3232 Description: Crash on NFS re-export node
3233 Details : under very unusual load conditions an assertion is hit in
3234 ll_intent_file_open()
3237 Frequency : only if OST filesystem is corrupted
3239 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3240 Details : for a short time RPCs with bulk IO are in the replay list,
3241 but replay of bulk IOs is unimplemented. If the OST filesystem
3242 is corrupted due to disk cache incoherency and then replay is
3243 started it is possible to trip an assertion. Avoid putting
3244 committed RPCs into the replay list at all to avoid this issue.
3249 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3250 Details : Large single O_DIRECT read and write calls can fail to allocate
3251 a sufficiently large buffer to process the request. In case of
3252 allocation failure the allocation is retried with a smaller
3253 buffer and broken into smaller requests.
3255 Severity : enhancement
3257 Description: Add -o localflock option to simulate old noflock behaviour.
3258 Details : This will achieve local-only flock/fcntl locks coherentness.
3263 Description: versioning check is incomplete
3264 Details : Checking the version difference of client vs. server, report
3265 error if the gap is too big.
3270 Description: add support PG_writeback bit
3271 Details : add support for PG_writeback bit for Lustre, for more carefull
3272 work with page cache in 2.6 kernel. This also fix some deadlocks
3273 and remove hack for work O_SYNC with 2.6 kernel.
3275 Severity : enhancement
3277 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3278 Details : The uninit_groups feature works in conjunction with the kernel
3279 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3280 up the pass1 processing of e2fsck. This is a read-only feature
3281 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3282 mount filesystems that have had this feature enabled.
3284 Severity : enhancement
3286 Description: Improve multi-block allocation algorithm to avoid fragmentation
3287 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3288 allocation locality and avoid filesystem fragmentation.
3290 ------------------------------------------------------------------------------
3292 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3294 * Support for kernels:
3295 2.6.9-42.0.3.EL (RHEL 4)
3296 2.6.5-7.276 (SLES 9)
3297 2.4.21-47.0.1.EL (RHEL 3)
3298 2.6.12.6 vanilla (kernel.org)
3299 2.6.16.21-0.8 (SLES10)
3300 * Recommended e2fsprogs version: 1.39.cfs2-0
3302 * The backwards-compatible /proc/sys/portals symlink has been removed
3303 in this release. Before upgrading, please ensure that you change
3304 any configuration scripts or /etc/sysctl.conf files that access
3305 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3306 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3307 in advance of the upgrade on any system running Lustre 1.4.6 or
3308 newer, since /proc/sys/lnet was added in that version.
3309 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3313 Frequency : only when quota is used
3315 Description: avoid scanning export list for quota master
3316 Details : Change the algorithms to avoid scanning export list in order
3317 to improve the efficiency.
3320 Frequency : MDS failover only, very rarely
3322 Description: "went back in time" messages on mds failover
3323 Details : The greatest transno may be lost when the current operation
3324 finishes with an error (transno==0) and the client's last_rcvd
3325 record is over-written. Save the greatest transno in the
3326 mds_last_transno for this case.
3329 Frequency : always for specific kernels and striping counts
3331 Description: client may get "Matching packet too big" without ACL support
3332 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3333 when trying to access files in certain configurations. The
3334 clients should in fact be denied when mounting because they do
3335 not understand ACLs.
3338 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3340 Description: many clients connecting with IO in progress causes connect timeouts
3341 Details : Avoid synchronous journal commits to avoid delays caused by many
3342 clients connecting/disconnecting when bulk IO is in progress.
3343 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3344 OST_IO_PORTAL to avoid delays behind potentially many pending
3348 Frequency : occasionally with multiple writers to a single file
3350 Description: shared writes to file may result in wrong size reported by stat()
3351 Details : Allow growing of kms when extent lock is cancelled
3354 Frequency : always with random mmap IO to multi-striped file
3356 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3357 Details : Only if the hole is at the end of OST object so that kms is too
3358 small. Fix is to increase kms accordingly in ll_nopage.
3361 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3363 Description: writes to a missing object would leak memory on the OST
3364 Details : If there is an inconsistency between the MDS and OST filesystems,
3365 such that the MDS references an object that doesn't exist, writes
3366 to that object will leak memory due to incorrect cleanup in the
3367 error handling path, eventually running out of memory on the OST.
3372 Description: Creating too long symlink causes lustre errors
3373 Details : Check symlink and name lengths before sending requests to MDS.
3376 Frequency : only if flock is enabled (not on by default)
3378 Description: posix locks not released on fd closure on 2.6.9+
3379 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3380 kernels, this caused such locks not to be released on fd close and
3381 then assertions on fs unmount about still used locks.
3384 Frequency : MDS failover only, very rarely
3386 Description: clients may get ASSERTION(granted_lock != NULL)
3387 Details : When request was taking a long time, and a client was resending
3388 a getattr by name lock request. The were multiple lock requests
3389 with the same client lock handle and
3390 mds_getattr_name->fixup_handle_for_resent_request found one of the
3391 lock handles but later failed with ASSERTION(granted_lock != NULL).
3396 Description: handle->h_buffer_credits > 0, assertion failure
3397 Details : h_buffer_credits is zero after truncate, causing assertion
3398 failure. This patch extends the transaction or creates a new
3402 Frequency : NFS re-export or patchless client
3403 Bugzilla : 11179, 10796
3404 Description: Crash on NFS re-export node (__d_move)
3405 Details : We do not want to hash the dentry if we don't have a lock.
3406 But if this dentry is later used in d_move, we'd hit uninitialised
3407 list head d_hash, so we just do this to init d_hash field but
3408 leave dentry unhashed.
3411 Frequency : NFS re-export or patchless client
3413 Description: NFS exports has problem with symbolic link
3414 Details : lustre client didn't properly install dentry when re-exported
3415 to NFS or running patchless client.
3418 Frequency : NFS re-export or patchless client
3420 Description: Various nfs/patchless fixes.
3421 Details : fixes reuse disconected alias for lookup process - this fixes
3422 warning "find_exported_dentry: npd != pd",
3423 fix permission error with open files at nfs.
3424 fix apply umask when do revalidate.
3427 Frequency : occasional
3429 Description: Crash on NFS re-export node
3430 Details : calling clear_page() on the wrong pointer triggered oops in
3431 generic_mapping_read().
3434 Frequency : rarely, using O_DIRECT IO
3436 Description: unaligned directio crashes client with LASSERT
3437 Details : check for unaligned buffers before trying any requests.
3440 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3442 Description: stale data returned from RAID cache
3443 Details : If only a small amount of IO is done to the RAID device before
3444 reading it again it is possible to get stale data from the RAID
3445 cache instead of reading it from disk.
3448 Frequency : always for sles10 kernel
3450 Description: sles10 support
3451 Details : ll_follow_link: compile fixes and using of nd_set_link
3452 under newer kernels.
3455 Frequency : depends on arch, kernel and compiler version, always on sles10
3458 Description: recursive or deep enough symlinks cause stack overflow
3459 Details : getting rid of large stack-allocated variable in
3463 Frequency : depends on hardware
3465 Description: lustre write performance loss in the SLES10 kernel
3466 Details : the performance loss is caused by using of write barriers in the
3467 ext3 code. The SLES10 kernel turns barrier support on by
3468 default. The fix is to undo that change for ldiskfs.
3470 ------------------------------------------------------------------------------
3472 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3474 * Support for kernels:
3475 2.6.9-42.0.3EL (RHEL 4)
3476 2.6.5-7.276 (SLES 9)
3477 2.4.21-47.0.1.EL (RHEL 3)
3478 2.6.12.6 vanilla (kernel.org)
3482 Frequency : quota enabled and large files being deleted
3484 Description: releasing more than 4GB of quota at once hangs OST
3485 Details : If a user deletes more than 4GB of files on a single OST it
3486 will cause the OST to spin in an infinite loop. Release
3487 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3492 Description: statfs data retrieved from /proc may be stale or zero
3493 Details : When reading per-device statfs data from /proc, in the
3494 {kbytes,files}_{total,free,avail} files, it may appear
3495 as zero or be out of date.
3498 Frequency : systems with MD RAID1 external journal devices
3500 Description: lconf's call to blkid is confused by RAID1 journal devices
3501 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3502 of returning all block devices that match the journal UUID.
3505 Frequency : always, for aggregate stripe size over 4GB
3507 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3508 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3509 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3512 Frequency : always if "lfs find" used on a local file/directory
3514 Description: "lfs find" segfaults if used on a local file/directory
3515 Details : The case where a directory component was not specified wasn't
3516 handled correctly. Handle this properly.
3519 Frequency : always on ppc64
3521 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3522 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3523 perform endianness conversion.
3526 Frequency : rarely (truncate to non-zero file size after write under load)
3527 Bugzilla : 10730, 10687
3528 Description: Files padded with zeros to next 4K multiple
3529 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3530 it is possible that files that are truncated to a non-zero size
3531 immediately after being written are filled with zero bytes beyond
3532 the truncated size. No file data is lost.
3534 Severity : enhancement
3536 Description: Allow recovery/failover for liblustre clients.
3537 Details : liblustre clients were unaware of failover configurations until
3540 Severity : enhancement
3542 Description: user file locks should fail when not mounting with flock option
3543 Details : Set up an error-returning stub in ll_file_operations.lock field
3544 to prevent incorrect behaviour when client is mounted without
3545 flock option. Also, set up properly f_op->flock field for
3549 Frequency : always on ia64
3551 Description: "lfs df" loops on printing out MDS statfs information
3552 Details : The obd_ioctl_data was not initialized and in some systems
3553 this caused a failure during the ioctl that did not return
3554 an error. Initialize the struct and return an error on failure.
3557 Frequency : SLES 9 only
3559 Description: Error of copying files with lustre special EAs as root
3560 Details : Client side always return success for setxattr call for lustre
3561 special xattr (currently only "trusted.lov").
3564 Frequency : rarely on clusters with both ia64+i386 clients
3566 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3567 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3568 clients are doing concurrent writes to the same file it is
3569 possible that the ia64 clients may LASSERT because the OST
3570 extent locks are not PAGE_SIZE aligned. Ensure that grown
3571 locks are always aligned on the request boundary.
3574 Frequency : specific use, occasional
3576 Description: Overwriting in use executable truncates on-disk binary image
3577 Details : If one node attempts to overwrite an executable in use by
3578 another node, we now correctly return ETXTBSY instead of
3579 truncating the file.
3581 Severity : enhancement
3583 Description: Async OSC create to avoid the blocking unnecessarily.
3584 Details : If a OST has no remain object, system will block on the creating
3585 when need to create a new object on this OST. Now, ways use
3586 pre-created objects when available, instead of blocking on an
3587 empty osc while others are not empty. If we must block, we block
3588 for the shortest possible period of time.
3593 Description: chmod on Lustre root is propagated to other clients
3594 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3595 invalid by the follow_mount time.
3598 Frequency : liblustre clients only
3600 Description: Race in 'instant cancel' lock handling could lead to such locks
3601 never to be granted in case of SMP MDS
3602 Details : Do not destroy not yet granted but cbpending locks in
3606 Frequency : replay/resend of open
3608 Description: non null lock assetion failure in mds_intent_policy
3609 Details : Trying to replay/resend lockless open requests resulted in
3610 mds_open() returning 0 with no lock. Now it sets a flag if
3611 a lock is going to be returned.
3613 Severity : enhancement
3615 Description: Checksum enhancements
3616 Details : New checksum enhancements allow for resending RPCs that failed
3619 Severity : enhancement
3621 Description: Tunables on number of dirty pages in cacche
3622 Details : Allow to set limit on number of dirty pages cached.
3627 Description: client crash on unmount - lock still has references
3628 Details : In some error handling cases it was possible to leak a lock
3629 reference on a client while accessing a file. This was not
3630 harmful to the client during operation, but would cause the
3631 client to crash when the filesystem is unmounted.
3634 Frequency : specific case, rare
3636 Description: ETXTBSY on mds though file not in use by client
3637 Details : ETXTBSY is no longer incorrectly returned when attempting to
3638 chmod or chown a directory that the user previously tried to
3639 execute or a currently-executing binary.
3642 Frequency : extremely rare except on liblustre-based clients
3644 Description: Lustre space not freed when files are deleted
3645 Details : Clean up open-unlinked files after client eviction. Previously
3646 the unlink was skipped and the files remained as orphans.
3651 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3652 Details : In some cases it was possible to send a blocking callback to
3653 a client doing a glimpse, even though that client didn't get
3654 a lock granted. When the glimpse lock is cancelled on the OST
3655 the freed lock is left on the waiting list and corrupted the list.
3658 Frequency : all core dumps
3660 Description: Broke core dumps to lustre
3661 Details : Negative dentry may be unhashed if parent does not have UPDATE
3662 lock, but some callers, e.g. do_coredump, expect dentry to be
3663 hashed after successful create, hash it in ll_create_it.
3665 ------------------------------------------------------------------------------
3667 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3669 * Support for kernels:
3670 2.6.9-42.0.2.EL (RHEL 4)
3671 2.6.5-7.276 (SLES 9)
3672 2.4.21-40.EL (RHEL 3)
3673 2.6.12.6 vanilla (kernel.org)
3677 Frequency : always on RHEL 3
3679 Description: Number of open files grows over time
3680 Details : The number of open files grows over time, whether or not
3681 Lustre is started. This was due to a filp leak introduced
3682 by one of our kernel patches.
3684 ------------------------------------------------------------------------------
3686 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3688 * Support for kernels:
3689 2.6.9-42.EL (RHEL 4)
3690 2.6.5-7.276 (SLES 9)
3691 2.4.21-40.EL (RHEL 3)
3692 2.6.12.6 vanilla (kernel.org)
3697 Bugzilla : 5719, 9635, 9792, 9684
3698 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3699 Details : If a server is under heavy load and cannot reply to new
3700 connection requests before the client resends the (re)connect,
3701 the connection handling code can behave badly if two service
3702 threads are concurrently handing separate (re)connections from
3703 the same client. Add better locking to the connection handling
3704 code, and ensure that only a single connection will be processed
3705 for a given client UUID, even if the lock is dropped.
3707 Severity : enhancement
3709 Description: add TCP zero-copy support to kernel
3710 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3711 sends if the hardware supports scatter-gather and checksumming.
3712 This allows socklnd to do client-write and server-read more
3713 efficiently and reduce CPU utilization from skbuf copying.
3716 Frequency : only if NFS exporting from client
3718 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3719 Details : If setting attributes on a file created under NFS that had
3720 never been opened it would be possible to oops the client
3721 if the file had no objects.
3724 Frequency : always for liblustre
3726 Description: liblustre client does MDS+OSTs setattr RPC for each write
3727 Details : When doing a write from a liblustre client, the client
3728 incorrectly issued an RPC to the MDS and each OST the file was
3729 striped over in order to update the timestamps. When writing
3730 with small chunks and many clients this could overwhelm the MDS
3731 with RPCs. In all cases it would slow down the write because
3732 these RPCs are unnecessary.
3734 Severity : enhancement
3736 Description: allow number of MDS service threads to be changed at module load
3737 Details : It is now possible to change the number of MDS service threads
3738 running. Adding "options mds mds_num_threads={N}" to the MDS's
3739 /etc/modprobe.conf will set the number of threads for the next
3740 time Lustre is restarted (assuming the "mds" module is also
3741 reloaded at that time). The default number of threads will
3742 stay the same, 32 for most systems.
3747 Description: OST crash if filesystem is unformatted or corrupt
3748 Details : If an OST is started on a device that has never been formatted
3749 or if the filesystem is corrupt and cannot even mount then the
3750 error handling cleanup routines would dereference a NULL pointer.
3755 Description: NULL pointer deref in llap_from_page.
3756 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3757 incorrect) mapping if the page was truncated/reclaimed while it was
3758 searched for. Check for this condition and skip such pages when
3759 doing readahead. Introduce extra check to llap_from_page() to
3760 verify page->mapping->host is non-NULL (so page is not anonymous).
3763 Frequency : Sometimes when using sys_sendfile
3765 Description: "page not covered by a lock" warnings from ll_readpage
3766 Details : sendfile called ll_readpage without right page locks present.
3767 Now we introduced ll_file_sendfile that does necessary locking
3768 around call to generic_file_sendfile() much like we do in
3772 Frequency : with certain MDS communication failures at client mount time
3774 Description: NULL pointer deref after failed client mount
3775 Details : a client connection request may delayed by the network layer
3776 and not be sent until after the PTLRPC layer has timed out the
3777 request. If the client fails the mount immediately it will try
3778 to clean up before the network times out the request. Add a
3779 reference from the request import to the obd device and delay
3780 the cleanup until the network drops the request.
3783 Frequency : occasionally during client (re)connect
3785 Description: assertion failure during client (re)connect
3786 Details : processing a client connection request may be delayed by the
3787 client or server longer than the client connect timeout. This
3788 causes the client to resend the connection request. If the
3789 original connection request is replied in this interval, the
3790 client may trip an assertion failure in ptlrpc_connect_interpret()
3791 which thought it would be the only running connect process.
3794 Frequency : only with obd_echo servers and clients that are rebooted
3796 Description: kernel BUG accessing uninitialized data structure
3797 Details : When running an obd_echo server it did not start the ping_evictor
3798 thread, and when a client was evicted an uninitialized data
3799 structure was accessed. Start the ping_evictor in the RPC
3800 service startup instead of the OBD startup.
3802 Severity : enhancement
3803 Bugzilla : 10193 (patchless)
3804 Description: Remove dependency on various unexported kernel interfaces.
3805 Details : No longer need reparent_to_init, exit_mm, exit_files,
3806 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3809 Frequency : rare (only users of deprecated and unsupported LDAP config)
3811 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3812 Details : LDAP apparently contains 'lustreName' attributes instead of
3813 'name'. A simple remapping of the name is sufficient.
3816 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3818 Description: waiting_locks_callback trips kernel BUG if client is evicted
3819 Details : Running with the dump_on_timeout debug flag turned on makes
3820 it possible that the waiting_locks_callback() can try to dump
3821 the Lustre kernel debug logs from an interrupt handler. Defer
3822 this log dumping to the expired_lock_main() thread.
3824 Severity : enhancement
3826 Description: Support NFS exporting on 2.6 kernels.
3827 Details : Implement non-rawops metadata methods for NFS server to use without
3828 changing NFS server code.
3831 Frequency : very rare (synthetic metadata workload only)
3833 Description: two racing renames might cause an MDS thread to deadlock
3834 Details : Running the "racer" program may cause one MDS thread to rename
3835 a file from being the source of a rename to being the target of
3836 a rename at exactly the same time that another thread is doing
3837 so, and the second thread has already enqueued these locks after
3838 doing a lookup of the target and is trying to relock them in
3839 order. Ensure that we don't try to re-lock the same resource.
3842 Frequency : only very large systems with liblustre clients
3844 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3845 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3846 all OSTs in parallel. This allows the eviction of stale liblustre
3847 clients to proceed much faster than if they were done in series,
3848 and also offers similar improvements for other set_info RPCs.
3853 Description: excessive CPU usage during initial read phase on client
3854 Details : During the initial read phase on a client, it would agressively
3855 retry readahead on the file, consuming too much CPU and impacting
3856 performance (since 1.4.5.8). Improve the readahead algorithm
3857 to avoid this, and also improve some other common cases (read
3858 of small files in particular, where "small" is files smaller than
3859 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3864 Description: MDS crash when receiving packet with unknown intent.
3865 Details : Do not LBUG in unknown intent case, just return -EFAULT
3867 Severity : enhancement
3868 Bugzilla : 9293, 9385
3869 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3870 Details : Do not serialize getattr (non-intent version) and statfs.
3873 Frequency : occasional, when OST network is overloaded/intermittent
3875 Description: client evicted by OST after bulk IO timeout
3876 Details : If a client sends a bulk IO request (read or write) the OST
3877 may evict the client if it is unresposive to its data GET/PUT
3878 request. This is incorrect if the network is overloaded (takes
3879 too long to transfer the RPC data) or dropped the OST GET/PUT
3880 request. There is no need to evict the client at all, since
3881 the pinger and/or lock callbacks will handle this, and the
3882 client can restart the bulk request.
3885 Frequency : Always when mmapping file with no objects
3887 Description: client crashes when mmapping file with no objects
3888 Details : Check that we actually have objects in a file before doing any
3889 operations on objects in ll_vm_open, ll_vm_close and
3895 Description: Request leak when working with deleted CWD
3896 Details : Introduce advanced request refcount tracking for requests
3897 referenced from lustre intent.
3899 Severity : Enhancement
3901 Description: Cache open file handles on client.
3902 Details : MDS now will return special lock along with openhandle, if
3903 requested and client is allowed to hold openhandle, even if unused,
3904 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3905 closing files for every read/write openration.
3907 Severity : Enhancement
3909 Description: Cache open negative dentries on client when possible.
3910 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3911 negative dentries on lock revocation.
3916 Description: Remounting a client read-only wasn't possible with a zconf mount
3917 Details : It wasn't possible to remount a client read-only with llmount.
3919 Severity : enhancement
3920 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3921 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3922 MPICH is included to add Lustre-specific ADIO interfaces.
3923 This is based closely on the UFS ADIO layer and only differs
3924 in file creation, in order to allow the OST striping to be set.
3925 This is user-contributed code and not supported by CFS.
3930 Description: extended inode attributes (immutable, append-only) work improperly
3931 when 2.4 and 2.6 kernels are used on client/server or vice versa
3932 Details : Introduce kernel-independent values for these flags.
3934 Severity : enhancement
3937 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3938 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3939 granularity. Now, Fractional MB granularity is be supported,
3940 this is very useful for low memory system.
3942 Severity : enhancement
3944 Description: Getattr by fid
3945 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3946 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3947 have locks and avoids getattr rpc requests for it on every stat.
3950 Frequency : Always, for filesystems larger than 2TB
3952 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3953 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3954 filesystems larger than 2TB to be mounted with mballoc enabled.
3957 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3959 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3960 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3961 filesystems larger than 2TB is mounted then the kernel will
3962 silently corrupt the start of the filesystem. CONFIG_LBD is
3963 enabled for all CFS-supported kernels, but the possibility of
3964 this happening with a modified kernel config exists.
3966 Severity : enhancement
3968 Description: add client O_DIRECT support for 2.6 kernels
3969 Details : It is now possible to do O_DIRECT reads and writes to files
3970 in the Lustre client mountpoint on 2.6 kernel clients.
3972 Severity : enhancement
3974 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3975 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3976 parallel, not waiting for response from every OST before sending
3977 a rpc to the next OST.
3982 Description: setattr vs write race when updating file timestamps
3983 Details : Client processes that update a file timestamp into the past
3984 right after writing to the file (e.g. tar) it is possible that
3985 the updated file modification time can be reset to the current
3986 time due to a race between processing the setattr and write RPC.
3988 Severity : enhancement
3990 Description: Bring 'lfs find' closer in line with regular Linux find.
3991 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3992 -print0 options and obtains all the needed info through the lustre
3995 Severity : enhancement
3997 Description: support up to 1024 configured devices on one node
3998 Details : change obd_dev array from statically allocated to dynamically
3999 allocated structs as they are first used to reduce memory usage
4004 Description: Flush dirty partially truncated pages during truncate
4005 Details : Immediatelly flush partially truncated pages in filter_setattr,
4006 this way we completely avoid having any pages in page cache on OST
4007 and can retire ugly workarounds during writes to flush such pages.
4012 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4013 Details : protect inode from truncation within vfs_unlink() context
4014 just take a reference before calling vfs_unlink() and release it
4015 when parent's i_sem is free.
4020 Description: last_id value checked outside lock on OST caused LASSERT failure
4021 Details : If there were multiple MDS->OST object precreate requests in
4022 flight, it was possible that the OST's last object id was checked
4023 outside a lock and incorrectly tripped an assertion. Move checks
4024 inside locks, and discard old precreate requests.
4027 Frequency : always, if extents are used on OSTs
4029 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4030 Details : OSTs using the extents format would not zero the high 16 bits of
4031 the index physical block number. This is not a problem for any
4032 OST filesystems smaller than 16TB, and no kernels support ext3
4033 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4034 new/modified files) and can be fixed for existing filesystems
4035 with e2fsprogs-1.39-cfs1.
4040 Description: import connection selection may be incorrect if timer wraps
4041 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4042 import connections to be ignored if the 32-bit jiffies counter
4043 wraps. Use a 64-bit jiffies counter.
4046 Frequency : very large clusters immediately after boot
4048 Description: LNET request buffers exhausted under heavy short-term load
4049 Details : If a large number of client requests are generated on a service
4050 that has previously never seen so many requests it is possible
4051 that the request buffer growth cannot keep up with the spike in
4052 demand. Instead of dropping incoming requests, they are held in
4053 the LND until the RPC service can accept more requests.
4056 Frequency : Sometimes during replay
4058 Description: Assertion failure in ll_local_open after replay.
4059 Details : If replay happened on an open request reply before we were able
4060 to set replay handler, reply will become not swabbed tripping the
4061 assertion in ll_local_open. Now we set the handler right after
4062 recognising of open request
4065 Frequency : very rare
4067 Description: kernel reports "badness in vsnprintf"
4068 Details : Reading from the "recovery_status" /proc file in small chunks
4069 may cause a negative length in lprocfs_obd_rd_recovery_status()
4070 call to vsnprintf() (which is otherwise harmless). Exit early
4071 if there is no more space in the output buffer.
4073 Severity : enhancement
4075 Description: clear OBD RPC statistics by writing to them
4076 Details : It is now possible to clear the OBD RPC statistics by writing
4077 to the "stats" file.
4082 Description: Client mtime is not the same on different clients after utimes
4083 Details : In some cases, the client was using the utimes() syscall on
4084 a file cached on another node. The clients now validate the
4085 ctime from the MDS + OSTs to determine which one is right.
4090 Description: Inability to activate failout mode
4091 Details : lconf script incorrectly assumed that in pythong string's numeric
4092 value is used in comparisons.
4095 Frequency : always with multiple stripes per file
4097 Description: Inefficient object allocation for mutli-stripe files
4098 Details : When selecting which OSTs to stripe files over, for files with
4099 a stripe count that divides evenly into the number of OSTs,
4100 the MDS is always picking the same starting OST for each file.
4101 Return the OST selection heuristic to the original design.
4106 Description: mount failures may take full timeout to return an error
4107 Details : Under some heavy load conditions it is possible that a
4108 failed mount can wait for the full obd_timeout interval,
4109 possibly several minutes, before reporting an error.
4110 Instead return an error as soon as the status is known.
4112 Frequency : quota enabled and large files being deleted
4114 Description: releasing more than 4GB of quota at once hangs OST
4115 Details : If a user deletes more than 4GB of files on a single OST it
4116 will cause the OST to spin in an infinite loop. Release
4117 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4122 Description: statfs data retrieved from /proc may be stale or zero
4123 Details : When reading per-device statfs data from /proc, in the
4124 {kbytes,files}_{total,free,avail} files, it may appear
4125 as zero or be out of date.
4128 Frequency : systems with MD RAID1 external journal devices
4130 Description: lconf's call to blkid is confused by RAID1 journal devices
4131 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4132 of returning all block devices that match the journal UUID.
4135 Frequency : always, for aggregate stripe size over 4GB
4137 Description: assertion fails when trying to use 4GB stripe size
4138 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4139 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4142 Frequency : always on ppc64
4144 Description: the first write on an ext3 filesystem with mballoc got stuck
4145 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4146 perform endianness conversion.
4148 ------------------------------------------------------------------------------
4150 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4152 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4153 INTEROPERATE with older versions automatically. Please read the
4154 user documentation before upgrading any part of a live system.
4155 * WARNING: Lustre networking configuration changes are required with
4156 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4159 * Support for kernels:
4160 2.6.9-22.0.2.EL (RHEL 4)
4161 2.6.5-7.244 (SLES 9)
4162 2.6.12.6 vanilla (kernel.org)
4165 Severity : enhancement
4166 Bugzilla : 7981/8208
4167 Description: Introduced Lustre Networking (LNET)
4168 Details : LNET is new networking infrastructure for Lustre, it includes
4169 a reorganized network configuration mode (see the user
4170 documentation for full details) as well as support for routing
4171 between different network fabrics. Lustre Networking Devices
4172 (LNDs) for the supported network fabrics have also been
4173 created for this new infrastructure.
4175 Severity : enhancement
4176 Description: Introduced Access control lists
4177 Details : clients can set ACLs on files and directories in order to have
4178 more fine-grained permissions than the standard Unix UGO+RWX.
4179 The MDS must be started with the "-o acl" mount option.
4181 Severity : enhancement
4182 Description: Introduced filesystem quotas
4183 Details : Administrators may now establish per-user quotas on the
4186 Severity : enhancement
4188 Description: Configuration change for the XT3
4189 The PTLLND is now used to run Lustre over Portals on the XT3
4190 The configure option(s) --with-cray-portals are no longer used.
4191 Rather --with-portals=<path-to-portals-includes> is used to
4192 enable building on the XT3. In addition to enable XT3 specific
4193 features the option --enable-cray-xt3 must be used.
4198 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4199 Details : A race between allocating a new llcd and re-getting the llcd_lock
4200 allowed another thread to grab newly-allocated llcd.
4202 Severity : enhancement
4204 Description: 2.6 OST async journal commit and locking fix to improve performance
4205 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4206 2.6 kernels are now async as they already were in 2.4 kernels so
4207 that they can commit concurrently with the network bulk transfer.
4208 For block-allocated files the filter allocation semaphore is held
4209 to avoid filesystem fragmentation during allocation. BKL lock
4210 removed for 2.6 xattr operations where it is no longer needed.
4215 Description: lconf incorrectly determined whether two IP networks could talk
4216 Details : In some more complicated routing and multiple-network
4217 configurations, lconf will avoid trying to make a network
4218 connection to a disjoint part of the IP space. It was doing the
4219 math incorrectly for one set of cases.
4224 Description: Fix for potential infinite loop processing records in an llog.
4225 Details : If an llog record is corrupted/zeroed, it is possible to loop
4226 forever in llog_process(). Validate the llog record length
4227 and skip the remainder of the block on error.
4230 Frequency : occasional (liblustre only)
4232 Description: liblustre could not open files whose last component is a symlink
4233 Details : sysio_path_walk() would incorrectly pass the open intent to
4234 intermediate path components.
4237 Frequency : rare (liblustre only with non-standard tuning)
4238 Bugzilla : 7201 (7350)
4239 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4240 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4241 after it has been released, only do this for referenced locks
4243 Severity : enhancement
4245 Description: specifying an (invalid) directory default stripe_size of -1
4246 would reset the directory default striping
4247 Details : stripe_size -1 was used internally to signal directory stripe
4248 removal, now use "all default" to signal dir stripe removal
4249 as a directory striping of "all default" is not useful
4252 Frequency : common for large clusters running liblustre clients
4254 Description: doing an ls when liblustre clients are running is slow
4255 Details : sending a glimpse AST to a liblustre client waits for every AST
4256 to time out, as liblustre clients will not respond. Since they
4257 cannot cache data we refresh the OST lock LVB from disk instead.
4259 Severity : enhancement
4261 Description: doing an ls at the same time as file IO can be slow
4262 Details : enqueue and other "small" requests can be blocked behind many
4263 large IO requests. Create a new OST IO portal for non-IO
4264 requests so they can be processed faster.
4267 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4269 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4270 created file with mode 000
4271 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4272 allow an ftruncate() as a non-root user to a file with mode 000.
4273 Lustre now respects this flag to disable mode checks when
4274 truncating a file owned by the user
4277 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4280 Description: Revoking locks from clients that went dead or catatonic might take
4282 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4283 cancellation of such locks instant on servers without waiting for
4284 any reply from clients. Clients drops these locks when cancel
4285 notification from server is received without replying.
4288 Frequency : liblustre-only, when liblustre client dies or becomes busy
4290 Description: Doing ls on Linux clients can take a long time with active
4292 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4293 granting such locks to it in the first place if possible. Locks
4294 are taken by proxy on the OST during the read or write and
4295 dropped immediately afterward. Add connect flags handling, do
4296 not grant locks to liblustre clients for glimpse ASTs.
4298 Severity : enhancement
4300 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4301 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4302 sequential reads and behaves badly with applications doing random
4303 reads. Improve it by reading ahead at least read region, and
4304 avoiding excessive large RPC for small reads.
4306 Severity : enhancement
4308 Description: Creating more than 1000 files for a single job may cause a load
4309 imbalance on the OSTs if there are also a large number of OSTs.
4310 Details : qos_prep_create() uses an OST index reseed value that is an
4311 even multiple of the number of available OSTs so that if the
4312 reseed happens in the middle of the object allocation it will
4313 still utilize the OSTs as uniformly as possible.
4318 Description: OST or MDS may oops in ping_evictor_main()
4319 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4320 but doesn't restart at beginning of obd_exports_timed list
4323 Severity : enhancement
4325 Description: improve by-nid export eviction on the MDS and OST
4326 Details : allow multiple exports with the same NID to be evicted at one
4327 time without re-searching the exports list.
4330 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4332 Description: MDS may oops in groups_free()
4333 Details : in rare race conditions a newly allocated group_info struct is
4334 freed again, and this can be NULL. The 2.4 compatibility code
4335 for groups_free() checked for a NULL pointer, but 2.6 did not.
4338 Frequency : common for liblustre clients doing little filesystem IO
4339 Bugzilla : 9352, 7313
4340 Description: server may evict liblustre clients accessing contended locks
4341 Details : if a client is granted a lock or receives a completion AST
4342 with a blocking AST already set it would not reply to the AST
4343 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4346 Frequency : lfs setstripe, only systems with more than 160 OSTs
4348 Description: unable to set striping with a starting offset beyond OST 160
4349 Details : llapi_create_file() incorrectly limited the starting stripe
4350 index to the maximum single-file stripe count.
4353 Frequency : LDAP users only
4355 Description: lconf did not handle in-kernel recovery with LDAP properly
4356 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4358 Severity : enhancement
4360 Description: bind OST threads to NUMA nodes to improve performance
4361 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4362 node and do their allocations there to localize memory access
4364 Severity : enhancement
4366 Description: llmount can determine client NID directly from Myrinet (GM)
4367 Details : the client NID code from gmnalnid was moved directly into
4368 llmount, removing the need to use this or specifying the
4369 client NID explicitly when mounting GM clients with zeroconf
4372 Frequency : if client is started with down MDS
4374 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4375 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4376 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4377 ptlrpc_import_delayed_req() because we didn't check if the
4378 request was interrupted, and we also didn't break out of the
4379 event loop if there was no timeout
4384 Description: data loss during non-page-aligned writes to a single file from
4385 both multiple nodes and multiple threads on one node at same time
4386 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4387 inconsistency led to false short-reads, that were cached and later
4388 used by ->prepare_write() to fill in partially written page,
4389 leading to data loss.
4392 Frequency : always, if lconf --abort_recovery used
4394 Description: lconf --abort_recovery fails with 'Operation not supported'
4395 Details : lconf was attempting to abort recovery on the MDT device and not
4398 Severity : enhancement
4400 Description: remove cleanup logs
4401 Details : replace lconf-generated cleanup logs with lustre internal
4402 cleanup routines. Eliminates the need for client-cleanup and
4405 Severity : enhancement
4407 Description: add support for EAs (user and system) on lustre filesystems
4408 Details : it is now possible to store extended attributes in the Lustre
4409 client filesystem, and with the user_xattr mount option it
4410 is possible to allow users to store EAs on their files also
4412 Severity : enhancement
4414 Description: Add possibility (config option) to show minimal available OST free
4416 Details : When compiled with --enable-mindf configure option, statfs(2)
4417 (and so, df) will return least minimal free space available from
4418 all OSTs as amount of free space on FS, instead of summary of
4419 free spaces of all OSTs.
4421 Severity : enhancement
4423 Description: do not expand extent locks acquired on OST-side
4424 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4425 by server: they are not cached anyway.
4428 Frequency : when mmap is used/binaries executed from Lustre
4430 Description: Unmmap pages before throwing them away from read cache.
4431 Details : llap_shrink cache now attempts to unmap pages before discarding
4432 them (if unmapping failed - do not discard). SLES9 kernel has
4433 extra checks that trigger if this unmapping is not done first.
4438 Description: lconf didn't resolve symlinks before checking to see whether a
4439 given mountpoint was already in use
4442 Frequency : when migrating failover services
4443 Bugzilla : 6395, 9514
4444 Description: When migrating a subset of services from a node (e.g. failback
4445 from a failover service node) the remaining services would
4446 time out and evict clients.
4447 Details : lconf --force (implied by --failover) sets the global obd_timeout
4448 to 5 seconds in order to quickly disconnect, but this caused
4449 other RPCs to time out too quickly. Do not change the global
4450 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4452 Severity : enhancement
4453 Frequency : if MDS is started with down OST
4454 Bugzilla : 9439,5706
4455 Description: Allow startup/shutdown of an MDS without depending on the
4456 availability of the OSTs.
4457 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4458 Add appropriate locking and lov-osc refcounts for safe
4459 cleaning. Add osc abort_inflight calls in case the
4460 synchronize never started.
4463 Frequency : occasional (Cray XT3 only)
4465 Description: root not authorized to access files in CRAY_PORTALS environment
4466 Details : The client process capabilities were not honoured on the MDS in
4467 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4468 been accessed by an authorized user then root was able to access
4469 the file on the local client also. The root user capabilities
4470 are now allowed on the MDS, as this environment has secure UID.
4473 Frequency : occasional
4475 Description: ldiskfs "too long searching" message happens too often
4476 Details : A debugging message (otherwise harmless) prints too often on
4477 the OST console. This has been reduced to only happen when
4478 there are fragmentation problems on the filesystem.
4483 Description: Division by zero in statfs when all OSCs are inactive
4484 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4485 lov_statfs divides by value returned by lov_get_stripecnt().
4489 Bugzilla : 9489, 3273
4490 Description: First write from each client to each OST was only 4kB in size,
4491 to initialize client writeback cache, which caused sub-optimal
4492 RPCs and poor layout on disk for the first writen file.
4493 Details : Clients now request an initial cache grant at (re)connect time
4494 and so that they can start streaming writes to the cache right
4495 away and always do full-sized RPCs if there is enough data.
4496 If the OST is rebooted the client also re-establishes its grant
4497 so that client cached writes will be honoured under the grant.
4502 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4503 Details : Now I/O RPCs go to different portal number and (presumably) fast
4504 lock requests (and glimses) and other RPCs get their own service
4505 threads pool that should be able to service those RPCs
4508 Severity : enhancement
4510 Description: Ability to exchange lustre version between client and servers and
4511 issue warnings at client side if client is too old. Also for
4512 liblustre clients there is ability to refuse connection of too old
4514 Details : New 'version' field is added to connect data structure that is
4515 filled with version info. That info is later checked by server and
4519 Frequency : rare, liblustre only.
4520 Bugzilla : 9296, 9581
4521 Description: Two simultaneous writes from liblustre at offset within same page
4522 might proceed at the same time overwriting eachother with stale
4524 Details : I/O lock withing llu_file_prwv was released too early, before data
4525 actually was hitting the wire. Extended lock-holding time until
4526 server acknowledges receiving data.
4529 Frequency : extremely rare. Never observed in practice.
4531 Description: avoid generating lustre_handle cookie of 0.
4532 Details : class_handle_hash() generates handle cookies by incrementing
4533 global counter, and can hit 0 occasionaly (this is unlikely, but
4534 not impossible, because initial value of cookie counter is
4535 selected randonly). Value of 0 is used as a sentinel meaning
4536 "unassigned handle" --- avoid it. Also coalesce two critical
4537 sections in this function into one.
4539 Severity : enhancement
4541 Description: allow liblustre clients to delegate truncate locking to OST
4542 Details : To avoid overhead of locking, liblustre client instructs OST to
4543 take extent lock in ost_punch() on client's behalf. New connection
4544 flag is added to handle backward compatibility.
4546 Severity : enhancement
4547 Bugzilla : 4928, 7341, 9758
4548 Description: allow number of OST service threads to be specified
4549 Details : a module parameter allows the number of OST service threads
4550 to be specified via "options ost ost_num_threads={N}" in the
4551 OSS's /etc/modules.conf or /etc/modprobe.conf.
4555 Bugzilla : 6146, 9635, 9895
4556 Description: servers crash with bad pointer in target_handle_connect()
4557 Details : In rare cases when a client is reconnecting it was possible that
4558 the connection request was the last reference for that export.
4559 We would temporarily drop the export reference and get a new
4560 one, but this may have been the last reference and the export
4561 was just destroyed. Get new reference before dropping old one.
4563 Severity : enhancement
4564 Frequency : if client is started with failover MDS
4566 Description: Allow multiple MDS hostnames in the mount command
4567 Details : Try to read the configuration from all specified MDS
4568 hostnames during a client mount in case the "primary"
4571 Severity : enhancement
4573 Description: Stop sending data to evicted clients as soon as possible.
4574 Details : Check if the client we are about to send or are sending data to
4575 was evicted already. (Check is done every second of waiting,
4576 for which l_wait_event interface was extended to allow checking
4577 of exit condition at specified intervals).
4580 Frequency : rare, normally only when NFS exporting is done from client
4582 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4584 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4587 Frequency : always, if srand() is called before liblustre initialization
4589 Description: Liblustre uses system PRNG disturbing its usage by user application
4590 Details : Introduce internal to lustre fast and high-quality PRNG for
4591 lustre usage and make liblustre and some other places in generic
4592 lustre code to use it.
4594 Severity : enhancement
4595 Bugzilla : 9477, 9557, 9870
4596 Description: Verify that the MDS configuration logs are updated when xml is
4597 Details : Check if the .xml configuration logs are newer than the config
4598 logs stored on the MDS and report an error if this is the case.
4599 Request --write-conf, or allow starting with --old_conf.
4601 Severity : enhancement
4603 Description: Handle symlinks in the path when checking if Lustre is mounted.
4604 Details : Resolve intermediate symlinks when checking if a client has
4605 mounted a filesystem to avoid duplicate client mounts.
4610 Description: lconf can hit an error exception but still return success.
4611 Details : The lconf command catches the Command error exception at the top
4612 level script context and will exit with the associated exit
4613 status, but doesn't ensure that this exit status is non-zero.
4618 Description: failure of ptlrpc thread startup can cause oops
4619 Details : Starting a ptlrpc service thread can fail if there are a large
4620 number of threads or the server memory is very fragmented.
4621 Handle this without oopsing.
4624 Frequency : always, only if liblustre and non-default acceptor port was used
4626 Description: liblustre cannot connect to servers with non-default acceptor port
4627 Details : tcpnal_set_default_params() was not called and was therefore
4628 ignoring the environment varaible TCPNAL_PORT, as well as other
4629 TCPNAL_ environment variables
4634 Description: two objects could be created on the same OST for a single file
4635 Details : If an OST is down, in some cases it was possible to create two
4636 objects on a single OST for a single file. No problems other
4637 than potential performance impact and spurious error messages.
4641 Bugzilla : 5681, 9562
4642 Description: Client may oops in ll_unhash_aliases
4643 Details : Client dcache may become inconsistent in race condition.
4644 In some cases "getcwd" can fail if the current directory is
4650 Description: Inode refcounting problems in NFS export code
4651 Details : link_raw functions used to call d_instantiate without obtaining
4652 extra inode reference first.
4656 Bugzilla : 9942, 9903
4657 Description: Referencing freed requests leading to crash, memleaks with NFS.
4658 Details : We used to require that call to ll_revalidate_it was always
4659 followed by ll_lookup_it. Also with revalidate_special() it is
4660 possible to call ll_revalidate_it() twice for the same dentry
4661 even if first occurence returned success. This fix changes semantic
4662 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4663 reference on a request referred from the intent.
4664 ll_intent_release() then releases such a request.
4667 Frequency : rare, normally benchmark loads only
4669 Description: unlinked inodes were kept in memory on the client
4670 Details : If a client is repeatedly creating and unlinking files it
4671 can accumulate a lot of stale inodes in the inode slab cache.
4672 If there is no other client load running this can cause the
4673 client node to run out of memory. Instead flush old inodes
4674 from client cache that have the same inode number as a new inode.
4677 Frequency : SLES9 2.6.5 kernel and long filenames only
4678 Bugzilla : 9969, 10379
4679 Description: utime reports stale NFS file handle
4680 Details : SLES9 uses out-of-dentry names in some cases, which confused
4681 the lustre dentry revalidation. Change it to always use the
4685 Frequency : rare, unless heavy write-truncate concurrency is continuous
4686 Bugzilla : 4180, 6984, 7171, 9963, 9331
4687 Description: OST becomes very slow and/or deadlocked during object unlink
4688 Details : filter_destroy() was holding onto the parent directory lock
4689 while truncating+unlinking objects. For very large objects this
4690 may block other threads for a long time and slow overall OST
4691 responsiveness. It may also be possible to get a lock ordering
4692 deadlock in this case, or run out of journal credits because of
4693 the combined truncate+unlink. Solution is to do object truncate
4694 first in one transaction without parent lock, and then do the
4695 final unlink in a new transaction with the parent lock. This
4696 reduces the lock hold time dramatically.
4699 Frequency : rare, 2.4 kernels only
4701 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4702 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4703 before unmounting their filesystem in order to avoid deadlock.
4704 The kernel_locked() function in 2.4 kernels only checks whether
4705 the kernel lock is held, not whether it is this process that is
4706 holding it as 2.6 kernels do.
4711 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4712 Details : The client ptlrpc code may be trying to reconnect to a down
4713 server before a previous connection attempt has timed out.
4714 Increase the reconnect interval to be longer than the connection
4715 timeout interval to avoid sending duplicate connections to
4719 Frequency : echo_client brw_test command
4721 Description: fix echo_client to work with OST preallocated code
4722 Details : OST preallocation code (5137) didn't take echo_client IO path
4723 into account: echo_client calls filter methods outside of any
4724 OST thread and, hence, there is no per-thread preallocated
4725 pages and buffers to use. Solution: hijack pga pages for IO. As
4726 a byproduct, this avoids unnecessary data copying.
4730 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4731 Description: Client can oops in mdc_commit_close() after open replay
4732 Details : It was possible for the MDS to return an open request with no
4733 transaction number in mds_finish_transno() if the client was
4734 evicted, but without actually returning an error. Clients
4735 would later try to replay that open and may trip an assertion
4736 Simplify the client close codepath, and always return an error
4737 from the MDS in case the open is not successful.
4740 Frequency : rare, 2.6 OSTs only
4742 Description: OST may deadlock under high load on fragmented files
4743 Details : If there was a heavy load and highly-fragmented OST filesystems
4744 it was possible to have all the OST threads deadlock waiting on
4745 allocation of biovecs, because the biovecs were not released
4746 until the entire RPC IO was completed. Instead, release biovecs
4747 as soon as they are complete to ensure forward IO progress.
4749 Severity : enhancement
4751 Description: Support for specifying external journal device at mount
4752 Details : If an OST or MDS device is formatted with an external journal
4753 device, this device major/minor is stored in the ext3 superblock
4754 and may not be valid for failover. Allow detecting and
4755 specifying the external journal at mount time.
4760 Description: Mounting an MDS with pending unlinked files may cause oops
4761 Details : target_finish_recovery() calls mds_postrecov() which returned
4762 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4763 considers this an error and immediately begins cleaning up the
4764 lov, just after starting the mds_lov process
4766 Severity : enhancement
4768 Description: Implement 'lfs df' to report actual free space on per-OST basis
4769 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4770 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4771 usage of objects; '-h' to report in human readable format.
4773 ------------------------------------------------------------------------------
4775 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4782 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4783 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4784 at startup during orphan unlinking. Instead, skip buddy update
4785 before setup, as it will be regenerated after recovery is complete.
4790 Description: If an OST is inactive, its locks might reference stale inodes.
4791 Details : lov_change_cbdata() must iterate over all namespaces, even if
4792 they are inactive to clear inode references from the lock.
4794 Severity : enhancement
4795 Frequency : occasional, if non-standard max_dirty_mb used
4797 Description: Client will block write RPCs if not enough grant
4798 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4799 then the client will block writes while waiting for another RPC
4800 to complete instead of consuming its dirty limit. With change
4801 we get improved performance when max_dirty_mb is small.
4803 Severity : enhancement
4804 Bugzilla : 3389, 6253
4805 Description: Add support for supplementary groups on the MDS.
4806 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4807 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4808 lookups for user supplementary groups into a cache.
4812 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4813 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4814 for liblustre O_EXCL is enforced by MDS.
4818 Description: readdir never returns NULL in liblustre.
4819 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4820 dentry in struct dirent.
4824 Frequency : liblustre only, depends on application IO pattern
4825 Description: liblustre clients evicted if not contacting servers
4826 Details : Don't put liblustre clients into the ping_evictor list, so
4827 they will not be evicted by the pinger ever.
4829 Severity : enhancement
4831 Description: Add ability to evict clients by NID from MDS.
4832 Details : By echoing "nid:$NID" string into
4833 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4834 $NID would be instantly evicted from this MDS and from all active
4835 OSTs connected to it.
4839 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4840 Details : lookup_it_finish() used to query file size from OSTs that was not
4845 Description: service threads change working directory to that of init
4846 Details : Starting lustre service threads may pin the working directory
4847 of the parent thread, making that filesystem busy. Threads
4848 now change to the working directory of init to avoid this.
4852 Frequency : during shutdown only
4853 Description: shutdown with a failed MDS or OST can cause unmount to hang
4854 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4859 Frequency : 2.6 only
4860 Description: chmod/chown may include an extra supplementary group
4861 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4862 supplementary group and if none is specified this is used.
4865 Bugzilla : 5479 (6816)
4867 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4868 Details : If lookup is in progress on a file that is unlinked we might try
4869 to revalidate the inode and fail in revalidate after lookup is
4870 complete and ll_file_open() enqueues the open again but
4871 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4874 Frequency : always, if lconf --abort_recovery used
4876 Description: lconf --abort_recovery fails with 'Operation not supported'
4877 Details : lconf was attempting to abort recovery on the MDT device and not
4880 ------------------------------------------------------------------------------
4882 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4887 Frequency : rare (only unsupported configurations with a node running as an
4889 Bugzilla : 6514, 5137
4890 Description: Mounting a Lustre file system on a node running as an OST could
4892 Details : OSTs now preallocates memory needed to write out data at
4893 startup, instead of when needed, to avoid having to
4894 allocate memory in possibly low memory situations.
4895 Specifically, if the file system is mounted on on OST,
4896 memory pressure could force it to try to write out data,
4897 which it needed to allocate memory to do. Due to the low
4898 memory, it would be unable to do so and the node would
4899 become unresponsive.
4901 Severity : enhancement
4903 Description: Addition of lconf --service command line option
4904 Details : lconf now accepts a '--service <arg>' option, which is
4905 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4907 Severity : enhancement
4909 Description: Failover mode is now the default for OSTs.
4910 Details : By default, OSTs will now run in failover mode. To return to
4911 the old behaviour, add '--failout' to the lmc line for OSTs.
4913 Severity : enhancement
4915 Description: Health checks are now provided for MDS and OSTs
4916 Details : Additional detailed health check information on MSD and OSTs
4917 is now provided through the procfs health_check value.
4920 Frequency : occasional, depends on IO load
4922 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4923 after numerous create/delete cycles
4924 Details : The ext3 inode allocation policy would not allocate new inodes
4925 very well on the OSTs because there are no new directories
4926 being created. Instead we look for groups with free space if
4927 the parent directories are nearly full.
4932 Description: Network or server problems during mount may cause partially
4933 mounted clients instead of returning an error.
4934 Details : The config llog parsing code may overwrite the error return
4935 code during mount error handling, returning success instead
4941 Description: MDS can fail to allocate large reply buffers
4942 Details : After long uptimes the MDS can fail to allocate large reply
4943 buffers (e.g. zconf client mount config records) due to memory
4944 fragmentation or consumption by the buffer cache. Preallocate
4945 some large reply buffers so that these replies can be sent even
4946 under memory pressure.
4950 Frequency : rare (liblustre)
4951 Description: fsx running with liblustre complained that using truncate() to
4952 extend the file doesn't work. This patch corrects that issue.
4953 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4954 ATTR_SIZE and lsm use in llu_setattr_raw.
4958 Frequency : rare, only 2.6 kernels
4959 Description: Unusual file access patterns on the MDS may result in inode
4960 data being lost in very rare circumstances.
4961 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4962 code in the 2.6 kernel caused the inode and/or EA data not to
4963 be read from disk, causing single-file corruption.
4967 Frequency : rare, only 2.6 filesystems using extents
4968 Description: Heavy concurrent write and delete load may cause data corruption.
4969 Details : It was possible under high-load situations to have an extent
4970 metadata block in the block device cache from a just-unlinked
4971 file overwrite a newly-allocated data block. We now unmap any
4972 metadata buffers that alias just-allocated data blocks.
4976 Frequency : filesystems with default stripe_count larger than 77
4977 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4978 Details : lconf specifies an inode size of 4096 bytes when the default
4979 stripe_count is larger than 77. This conflicts with the default
4980 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4981 this case to avoid pinning too much memory for large EAs.
4983 ------------------------------------------------------------------------------
4985 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4990 Frequency : rare (extremely heavy IO load with hundreds of clients)
4992 Description: Client is evicted, gets IO error writing to file
4993 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4994 caused clients to be evicted for AST timeouts. The fixes in
4995 bug 5192 mean we no longer need to have such short AST timeouts
4996 so ldlm_timeout has been increased.
4999 Frequency : occasional during --force or --failover shutdown under load
5000 Bugzilla : 5949, 4834
5001 Description: Server oops/LBUG if stopped with --force or --failover under load
5002 Details : a collection of import/export refcount and cleanup ordering
5003 issues fixed for safer force cleanup
5006 Frequency : only filesystems larger than 120 OSTs
5007 Bugzilla : 5990, 6223
5008 Description: lfs getstripe would oops on a very large filesystem
5009 Details : lov_getconfig used kfree on vmalloc'd memory
5012 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5013 Bugzilla : 6242, 6243
5014 Description: reading from files that had been truncated to a non-zero size
5015 but never opened returned no data
5016 Details : ll_file_read() reads zeros from no-object files to EOF
5021 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5022 Details : There's little harm in aborting MDS/OSS recovery and letting it
5023 try again, so I removed the LASSERT and return an error instead.
5025 Severity : enhancement
5027 Description: New debugging infrastructure for tracking down data corruption
5028 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5029 (b) cover more of the client-side code path, and (c) try to narrow
5030 down where problems occurred
5034 Bugzilla : 3819, 4364, 4397, 6313
5035 Description: Racing close and eviction MDS could cause assertion in mds_close
5036 Details : It was possible to get multiple mfd references during close and
5037 client eviction, leading to one thread referencing a freed mfd.
5039 Severity: : enhancement
5040 Bugzilla : 3262, 6359
5041 Description: Attempts to reconnect to servers are now more aggressive.
5042 Details : This builds on the enhanced upcall-less recovery that was added
5043 in 1.4.2. When trying to reconnect to servers, clients will
5044 now try each server in the failover group every 10 seconds. By
5045 default, clients would previously try one server every 25 seconds.
5050 Description: After recovery, certain operations trigger a failed
5051 assertion on a client.
5052 Details : Failing over an mds, using lconf -d --failover, while a
5053 client was doing a readdir() call would cause the client to
5054 LBUG after recovery completed and the readdir() was resent.
5056 Severity : enhancement
5058 Description: Default groups are now added by lconf
5059 Details : You can now run lconf --group <servicename> without having to
5060 manually add groups with lmc.
5063 Frequency : occasional
5065 Description: Nodes with an elan id of 0 trigger a failed assertion
5068 Frequency : always when accessing e.g. tty/console device nodes
5070 Description: tty and some other devices nodes cannot be used on lustre
5071 Details : file's private_data field is used by device data and lustre
5072 values in there got lost. New field was added to struct file to
5073 store fs-specific private data.
5076 Frequency : when exporting Lustre via NFS
5078 Description: NFSD failed occasionally when looking up a path component
5079 Details : NFSD is looking up ".." which was broken in ext3 directories
5080 that had grown large enough to become hashed.
5083 Frequency : Clusters with multiple interfaces not on the same subnet
5085 Description: Nodes will repeatedly try to reconnect to an interface which it
5086 cannot reach and report an error to the log.
5087 Details : Extra peer list entries will be created by lconf with some peers
5088 unreachable. lconf now validates the peer before adding it.
5091 Frequency : Only if a default stripe is set on the filesystem root.
5093 Description: Setting a default stripe on the filesystem root prevented the
5094 filesystem from being remounted.
5095 Details : The client was sending extra request flags in the root getattr
5096 request and did not allocate a reply buffer for the dir EA.
5099 Frequency : occasional, higher if lots of files are accessed by one client
5100 Bugzilla : 6159, 6097
5101 Description: Client trips assertion regarding lsm mismatch/magic
5102 Details : While revalidating inodes the VFS looks up inodes with ifind()
5103 and in rare cases can find an inode that is being freed.
5104 The ll_test_inode() code will free the lsm during ifind()
5105 when it finds an existing inode and then the VFS later attaches
5106 this free lsm to a new inode.
5110 Bugzilla : 6422, 7030
5111 Description: MDS deadlock between mkdir and client eviction
5112 Details : Creating a new file via mkdir or mknod (starting a transaction
5113 and getting the ns lock) can deadlock with client eviction
5114 (gets ns lock and trying to finish a synchronous transaction).
5117 Frequency : occasional
5118 Description: While starting a server, the fsfilt_ext3 module could not be
5120 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5121 kernels. Previously, lconf would still use the ext3 name
5122 when trying to load modules. Now, it will correctly use
5123 ext3 on 2.4 and ldiskfs on 2.6.
5125 Severity : enhancement
5126 Description: The default stripe count has been changed to 1
5127 Details : The interpretation of the default stripe count (0, to lfs
5128 or lmc) has been changed to mean striping across a single
5129 OST, rather than all available. For general usage we have
5130 found a stripe count of 1 or 2 works best.
5132 Severity : enhancement
5133 Description: Add support for compiling against Cray portals.
5134 Details : Conditional compiling for some areas that are different
5138 Frequency : occasional
5139 Bugzilla : 6409, 6834
5140 Description: Creating files with an explicit stripe count may lead to
5141 a failed assertion on the MDS
5142 Details : If some OSTs are full or unavailable, creating files may
5143 trigger a failed assertion on the MDS. Now, Lustre will
5144 try to use other servers or return an error to the
5148 Frequency : occasional
5150 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5151 caused lock timeout during readahead (since 1.4.2).
5152 Details : Processes doing readahead might match a lock that hasn't been
5153 granted yet if there are overlapping and conflicting lock
5154 requests. The readahead process waits on ungranted lock
5155 (original lock is CBPENDING), while OST waits for that process
5156 to cancel CBPENDING read lock and eventually evicts client.
5158 Severity : enhancement
5160 Description: Initial enabling of flock support for clients
5161 Details : Implements fcntl advisory locking and file status functions.
5162 This feature is provided as an optional mount flag (default
5163 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5164 locking are implemented yet, and those that are are not guaranteed
5165 to be completely correct in production environments.
5166 mount -t lustre -o [flock|noflock] ...
5169 Frequency : occasional
5171 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5172 assertion in the ext3 JBD (journaling) layer.
5173 Details : The b_committed_data struct is protected by the big kernel lock
5174 in 2.4 kernels, serializing journal_commit_transaction() and
5175 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5176 access to this struct. In 2.6 kernels there is finer grained
5177 locking to improve SMP performance of the JBD layer.
5181 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5186 Description: merge and cleanup kernel patches.
5189 -----------------------------------------------------------------------------
5191 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5193 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5194 versions of Lustre. Please update all servers and clients to
5195 version 1.4.2 or later at the same time. You must also run
5196 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5197 to update the configuration logs.
5199 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5200 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5201 - swab llog records of type '0' so we get proper header size/idx (5861)
5202 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5203 - fix rename of one directory over another leaking an inode (5953)
5204 - avoid SetPageDirty on 2.6 (5981)
5205 - don't re-add just-being-destroyed locks to the waiting list (5653)
5206 - when creating new directories, inherit the parent's custom
5207 striping settings if present parent (3048)
5208 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5209 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5210 - don't hold client locks on temporary worklist from l_lru (5666)
5211 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5212 - automatically evict dead clients (5921)
5213 - Update file size properly in create+truncate+fstat case (6196)
5214 - Do not unhash mountpoint dentries, do not allow removal of
5216 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5217 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5218 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5219 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5220 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5221 - Correctly handle reads of files with no objects (6243)
5222 - lctl recover will also mark a device active if deactivate used (5933)
5224 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5225 - allow --write-conf on an MDS with different nettype than client (5619)
5226 - don't write config llogs to MDS for mounts not from that MDS (5617)
5227 - lconf should create multiple TCP connections from a client (5201)
5228 - init scripts are now turned off by default; run chkconfig --on
5229 lustre and chkconfig --on lustrefs to use them
5230 - upcalls are no longer needed for clients to recover to failover
5232 - add --abort-recovery option to lconf to abort recovery on device
5234 - add support for an arbitrary number of OSTs (3026)
5235 - Quota support protocol changes.
5236 - forward compatibility changes to wire structs (6007)
5237 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5238 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5239 - improved "lustre status" script
5240 - initialize blocksize for non-regular files (6062)
5241 - added --disable-server and --disable-client configure options (5782)
5242 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5243 - Vanilla 2.4.29 support
5244 - increase maximum number of obd devices to 520 (6242)
5245 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5246 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5247 SLES 9 2.6.5 kernels
5248 - limit stripes per file to 160 (the maximum EA size) (6093)
5250 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5253 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5254 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5255 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5256 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5257 - small ext3 extents cleanups and fixes (5733)
5258 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5259 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5260 - avoid needless client->OST connect, fix handle mismatch (5317)
5261 - fix DLM error path that led to out-of-sync client, long delays (5779)
5262 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5263 - fix several locking issues related to i_size (5492,5624,5654,5672)
5264 - don't move pending lock onto export if it is already evicted (5683)
5265 - fix kernel oops when creating .foo in unlinked directory (5548)
5266 - fix deadlock in obdfilter statistics vs. object create (5811)
5267 - use time_{before,after} to avoid timer jiffies wrap (5882)
5268 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5269 - Do not leak request if server was not able to process it (5154)
5270 - If mds_open unable to find parent dir, make that negative lookup(5154)
5271 - don't create new directories with extent-mapping (5909, 5936)
5273 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5274 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5275 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5276 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5277 - support for RHEL4 kernel on i686 (5773)
5278 - provide error messages when incompatible logs are encountered (5898)
5280 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5281 * version 1.4.0.10 (1.4.1 release candidate 1)
5283 - don't keep a lock reference when lock is not granted (4238)
5284 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5285 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5286 - drop import inflight refcount on signal_completed_replay error (5255)
5287 - unlock page after async write error during send (3677)
5288 - handle missing objects in filter_preprw_read properly (5265)
5289 - no transno return for symlink open, don't save no-trasno open (3440)
5290 - don't try to complete elan receive that already failed (4012)
5291 - free RPC server reply state on error (5406)
5292 - clean up thread from ptlrpc_start_thread() on error (5160)
5293 - readahead could read extra page into cache that wasn't ejected (5388)
5294 - prevent races in class_attach/setup/cleanup/detach (5260)
5295 - don't dereference de->d_inode after l_dput of de (5458)
5296 - use "int" for stripe value returned from lock_to_stripe (5544)
5297 - mballoc allocation and error-checking fixes in 2.6 (5504)
5298 - block device patches to fix I/O request sizes in 2.6 (5482)
5299 - look up hostnames for IB nals (5602)
5300 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5301 - don't start multiple acceptors for the same port (5277)
5302 - fix incorrect LASSERT in mds_getattr_name (5635)
5303 - export a proc file for general "ping" checking (5628)
5304 - fix "lfs check" to not block when the MDS is down (5628)
5306 - service request history (4965)
5307 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5308 - create an "evict_client" /proc entry on OSTs, like the MDS has
5309 - fix mount usage message, return errors per mount(8) (5168)
5310 - change grep [] to grep "[]" in tests so they work in more UMLs
5311 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5312 - remove ancient LOV_MAGIC_V0 EA support (5047)
5313 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5314 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5315 - allow readahead allocations to fail when low on memory (5383)
5316 - mmap locking landed again, after considerable improvement (2828)
5317 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5319 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5322 - send OST transaction number in read/write reply to free req (4966)
5323 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5324 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5325 - the watchdog thread now runs as interruptible (5246)
5326 - flock/lockf fixes (but it's still disabled, pending 5135)
5327 - don't use EXT3 constants in llite code (5094)
5328 - memory shortage at startup could cause assertion (5176)
5330 - reorganization of lov code
5331 - single portals codebase
5333 - add extents/mballoc support (5025)
5334 - direct I/O reads in the obdfilter (4048)
5335 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5337 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5340 - send OST transaction number in read/write reply to free req (4966)
5341 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5342 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5343 - the watchdog thread now runs as interruptible (5246)
5344 - handle missing objects in filter_preprw_read properly (5265)
5345 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5346 - drop import inflight refcount on signal_completed_replay error (5255)
5347 - unlock page after async write error during send (3677)
5348 - return original error code on reconstructed replies (3761)
5349 - no transno return for symlink open, don't save no-trasno open (3440)
5351 - add pid to ldlm debugging output (4922)
5352 - bump the watchdog timeouts -- we can't handle 30sec yet
5353 - extra debugging for orphan dentry/inode bug (5259)
5355 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5358 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5359 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5360 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5361 - fix NULL dereference in /proc/sys/portals/routes (4827)
5362 - allow failed mdc_close() operations to be interrupted (4561)
5363 - stop precreate on OST before MDS would time out on it (4778)
5364 - don't send partial-page writes before EOF from client (4410)
5365 - discard client grant for sub-page writes on large-page clients (4520)
5366 - don't free dentries not owned by NFS code, check generation (4806)
5367 - fix lsm leak if mds_create_objects() fails (4801)
5368 - limit debug_daemon file size, always print CERROR messages (4789)
5369 - use transno after validating reply (3892)
5370 - process timed out requests if import state changes (3754)
5371 - update mtime on OST during writes, return in glimpse (4829)
5372 - add mkfsoptions to LDAP (4679)
5373 - use ->max_readahead method instead of zapping global ra (5039)
5374 - don't interrupt __l_wait_event() during strace
5376 - add software watchdogs to catch hung threads quickly (4941)
5377 - make lustrefs init script start after nfs is mounted
5378 - fix CWARN/ERROR duplication (4930)
5379 - return async write errors to application if possible (2248)
5380 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5381 - print NAL number in %x format (4645)
5382 - update barely-supported suse-2.4.21-171 series (4842)
5383 - support for sles 9 %post scripts
5384 - support for building 2.6 kernel-source packages
5385 - support for sles km_* packages
5387 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5390 - ignore -ENOENT errors in osc_destroy (3639)
5391 - notify osc create thread that OSC is being cleaned up (4600)
5392 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5393 - reconstruct ost_handle() like mds_handle() (4657)
5394 - create a new thread to do import eviction to avoid deadlock (3969)
5395 - let lconf resolve symlinked-to devices (4629)
5396 - don't unlink "objects" from directory with default EA (4554)
5397 - hold socknal file ref over connect in case target is down (4394)
5398 - allow more than 32000 subdirectories in a single directory (3244)
5399 - fix blocks count for O_DIRECT writes (3751)
5400 - OST returns ENOSPC from object create when no space left (4539)
5401 - don't send truncate RPC if file size isn't changing (4410)
5402 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5403 - bind to privileged port in socknal and tcpnal (3689)
5405 - rate limit CERROR/CWARN console message to avoid overload (4519)
5406 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5407 - basic mmap support (3918)
5408 - kernel patch series update from b1_4 (4711)
5410 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5413 - avoid crash during MDS cleanup with OST shut down (2775)
5414 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5415 - don't use bad inodes on the MDS (3744)
5416 - dynamic object preallocation to improve recovery speed (4236)
5417 - don't hold spinlock over lock dumping or change debug flags (4401)
5418 - don't zero obd_dev when it is force cleaned (3651)
5419 - print grants to console if they go negative (4431)
5420 - "lctl deactivate" will stop automatic recovery attempts (3406)
5421 - look for existing locks in ldlm_handle_enqueue() (3764)
5422 - don't resolve lock handle twice in recovery avoiding race (4401)
5423 - revalidate should check working dir is a directory (4134)
5425 - don't always mark "slow" obdfilter messages as errors (4418)
5427 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5430 - don't close LustreDB during write_conf until it is done (3860)
5431 - fix typo in lconf for_each_profile (3821)
5432 - allow dumping logs from multiple threads at one time (3820)
5433 - don't allow multiple threads in OSC recovery (3812)
5434 - fix debug_size parameters (3864)
5435 - fix mds_postrecov to initialize import for llog ctxt (3121)
5436 - replace config semaphore with spinlock (3306)
5437 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5438 - don't allow enqueue to complete on a destroyed export (3822)
5439 - down write_lock before checking llog header bitmap (3825)
5440 - recover from lock replay timeout (3764)
5441 - up llog sem before sending rpc (3652)
5442 - reduce ns lock hold times when setting kms (3267)
5443 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5444 - fix NULL deref and obd_dev leak on setup error (3312)
5445 - replace some LBUG about llog ops with error handling (3841)
5446 - don't match INVALID dentries from d_lookup and spin (3784)
5447 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5448 - fix invalid assertion in ptlrpc_set_wait (3880)
5450 - add libwrap support for the TCP acceptor (3996)
5451 - add /proc/sys/portals/routes for non-root route listing (3994)
5452 - allow setting MDS UUID in .xml (2580)
5453 - print the stack of a process that LBUGs (4228)
5455 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5458 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5459 - make sure to unset replay flag from failed open requests (3440)
5460 - if default stripe count is 0, use OST count for inode size (3636)
5461 - update parent mtime/ctime on client for create/unlink (2611)
5462 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5463 - free recovery state on server during a forced cleanup (3571)
5464 - unregister_reply for resent reqs (3063)
5465 - loop back devices mounting and status check on 2.6 (3563)
5466 - fix resource-creation race that can provoke i_size == 0 (3513)
5467 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5468 - more debugging for page-accounting assertion (3746)
5469 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5470 - avoid deadlock after precreation failure (3758)
5471 - fix race and lock order deadlock in orphan handling (3450, 3750)
5472 - add validity checks when grabbing inodes from l_ast_data (3599)
5474 - add /proc/.../recovery_status to obdfilter (3428)
5475 - lightweight CDEBUG infrastructure, debug daemon (3668)
5476 - change default OSC RPC parameters to be better on small clusters
5477 - turn off OST read cache for files smaller than 32MB
5478 - install man pages and include them in rpms (3100)
5479 - add new init script for (un)mounting lustre filesystems (2593)
5480 - run chkconfig in %post for init scripts (3701)
5481 - drop scimac NAL (unmaintained)
5483 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5486 - clean kiobufs before and after use (3485)
5487 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5488 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5489 - change init script to fail more clearly if not run as root (1528)
5490 - allow clients to reconnect during replay (1742)
5491 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5492 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5493 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5494 - fix cond_resched (3554)
5495 - only evict unfinished clients after recovery (3515)
5496 - allow bulk resend, prevent data loss (3570)
5497 - dynamic ptlrpc request buffer allocation (2102)
5498 - don't allow unlinking open directory if it isn't empty (2904)
5499 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5500 - remove extraneous obd dereference causing LASSERT failure (3334)
5501 - don't use get_cycles() when creating temp. files on the mds (3156)
5502 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5503 - handle EEXIST for set-stripe, set proper directory name (3336)
5505 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5506 - fix ksocknal_fmb_callback() error messages (2918)
5508 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5511 - don't copy lvb into (possibly NULL) reply on error (2983)
5512 - don't deref dentry after dput, don't free lvb on error (2922)
5513 - use the kms to determine writeback rpc length (2947)
5514 - increment oti_logcookies when osc is inactive (2948)
5515 - update client's i_blocks count via lvb messages (2543)
5516 - handle intent open/close of special files properly (1557)
5517 - mount MDS with errors=remount-ro, like obdfilter (2009)
5518 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5519 - don't use cancelling-locks' kms values (2947)
5520 - use highest lock extent for kms, not last one (2925)
5521 - don't dereference ERR_PTR() dentry in error handling path (3107)
5522 - fix thread race in portals_debug_dumplog() (3122)
5523 - create lprocfs device entries at setup instead of at attach (1519)
5524 - common AST error handler, don't evict client on completion race (3145)
5525 - zero nameidata in detach_mnt in 2.6 (3118)
5526 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5527 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5528 - initialize RPC timeout timer earlier for 2.6 (3219)
5529 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5530 - print nal/nid for unknown nid (3258)
5531 - additional checks for oscc recovery before doing precreate (3284)
5532 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5533 - don't crash in mdc_close for bad permissions on open (3285)
5534 - zero i_rdev for non-device files (3147)
5535 - clear page->private before handing to FS, better assertion (3119)
5536 - tune the read pipeline (3236)
5537 - fix incorrect decref of invalidated dentry (2350)
5538 - provide read-ahead stats and refine rpc in flight stats (3328)
5539 - don't hold journal transaction open across create RPC (3313)
5540 - update atime on MDS at close time (3265)
5541 - close LDAP connection when recovering to avoid server load (3315)
5542 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5543 - don't leak open file on MDS after open resend (3325)
5544 - serialize filter_precreate and filter_destroy_precreated (3329)
5545 - loop device shouldn't call sync_dev() for nul device (3092)
5546 - clear page cache after eviction (2766)
5547 - resynchronize MDS->OST in background (2824)
5548 - refuse to mount the same filesystem twice on same mountpoint (3394)
5549 - allow llmount to create routes for mounting behind routers (3320)
5550 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5551 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5552 - fix rare problem with rename on htree directories (3417)
5554 - allow default OST striping configuration per directory (1414)
5555 - fix compilation for qswnal for 2.6 kernels (3125)
5556 - increase maximum number of MDS request buffers for large systems
5557 - change liblustreapi to be useful for external progs like lfsck (3098)
5558 - increase local configuration timeout for slow disks (3353)
5559 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5561 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5564 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5565 - don't overwrite extent policy data in reply if lock was blocked (2901)
5566 - drop filter export grants atomically with removal from device (2663)
5567 - del obd_self_export from work_list in class_disconnect_exports (2908)
5568 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5569 - swab reply message in mdc_close, other PPC fixes (2464)
5570 - fix destroying of named logs (2325)
5571 - overwrite old logs when running lconf --write_conf (2264)
5572 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5573 - fix race in target_handle_connect (2898)
5574 - mds_reint_create() should take same inode create lock (2926)
5575 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5576 - don't close files for self_export to avoid uninitialized obd (2936)
5577 - allow MDS with the same name as client node (2939)
5578 - hold dentry reference for closed log files for unlink (2325)
5579 - reserve space for all logs during transactions (2059)
5580 - don't evict page beyond end of stripe extent (2925)
5581 - don't oops on a deleted current working directory (2399)
5582 - handle hard links to targets without a parent properly (2517)
5583 - don't dereference NULL lock when racing during eviction (2867)
5584 - don't grow lock extents when lots of conflicting locks (2919)
5586 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5589 - account for cache space usage on clients to avoid data loss (974)
5590 - lfsck support in lustre kernel code (2349)
5591 - reduce journal credits needed for BRW writes (2370)
5592 - orphan handling to avoid losing space on client/server crashes
5593 - ptlrpcd can be blocked, stopping ALL progress (2477)
5594 - use lock value blocks to assist in proper KMS, faster stat (1021)
5595 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5596 - recovery for initial connections (2355)
5597 - fixes for mds_cleanup_orphans (1934)
5598 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5599 - block all file creations until orphan recovery completes (1901)
5600 - client remove rq_connection from request struct (2423)
5601 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5602 - recovery timer race (2670)
5603 - mdc_close recovey bug (2532)
5604 - ptlrpc cleanup bug (2710)
5605 - mds timeout on local locks (2588)
5606 - namespace lock held during RPCs (2431)
5607 - handle interrupted sync write properly (2503)
5608 - don't try to handle a message that hasn't been replied to (2699)
5609 - client assert failure during cleanup after abort recovery (2701)
5610 - leak mdc device after failed mount (2712)
5611 - ptlrpc_check_set allows timedout requests to complete (2714)
5612 - wait for inflight reqs when ptlrpcd finishes (2710)
5613 - make sure unregistered services are removed from the srv_list
5614 - reset bulk XID's when resending them (caught by 1138 test)
5615 - unregister_bulk after timeout
5616 - fix lconf error (2694)
5617 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5618 - readahead locks pages, leaves pending causing memory pressure (2673)
5619 - increase OST request buffers to 4096 on large machines (2729)
5620 - fix up permission of existing directories in simple_mkdir (2661)
5621 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5622 - don't assign transno to errored transactions (2742)
5623 - don't delete objects on OST if given a bogus objid from MDS (2751)
5624 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5625 - if rq_no_resend, then timeout request after recovery (2432)
5626 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5627 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5628 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5629 - MDS recovery shouldn't delete objects that it has given out (2730)
5630 - if enqueue arrives after completion, don't clobber LVB (2819)
5631 - don't unlock pages twice when trigger_group_io returns error (2814)
5632 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5633 - don't write pages to disk if there was an error (1450)
5634 - don't ping imports that have recovery disabled (2676)
5635 - take buffered bytes into account when balancing socknal conn (2817)
5636 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5637 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5638 - remove little-endian swabbing of llog records (1987)
5639 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5640 - retry reposting request buffers if they fail (1191)
5641 - grow extent at grant time to avoid granting a revoked lock (2809)
5642 - lock revoke doesn't evict page if covered by a second lock (2765)
5643 - disable VM readahead to avoid reading outside lock extents (2805)
5645 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5646 - updated kernel patches for hp-2.4.20 kernel (2681)
5648 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5651 - fix truncated write corruption (2366)
5652 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5654 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5655 - ensure len(uuid) < 37 in lmc (1171)
5656 - fix ia64 OOPS in llog_test (2255)
5657 - zero end of page at obdfilter for partial page writes (2648)
5658 - don't leave stale dentries around after renames (bug 2428)
5659 - fix timeouts when evicting a client with a single lock held (2642)
5660 - set deadline for the initial HELLO message to drain (2634)
5661 - print out dotted-quad IP addresses in the socknal (2302)
5663 - additional debugging for MDS client eviction problem (2443)
5664 - fix mkfsoptions support for osts (2603, 2604)
5666 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5669 - add series for the vanilla 2.6.0 kernel
5670 - add series for the vanilla 2.4.24 kernel
5671 - add series for a cray x86/64 UL kernel drop
5672 - fix xattr patches for the vanilla 2.4.19 series
5674 - generate true UUIDs in lmc (1171)
5675 - have portals stack dumping break in UML (2466)
5676 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5677 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5678 - wake callback waiting threads on client eviction (2460)
5679 - Add --ptldebug and --subsystem to lmc (1719)
5680 - update assertion to allow safe interrupt allocation
5681 - set rq_no_resend for cancel requests (2432)
5682 - recalculate ptlrpcd timeout after resend (2494)
5683 - call vfs_rmdir when removing pending directories (2368)
5684 - fix renaming a file to itself (2429)
5685 - lmc creates a default one-stripe lov (2454)
5686 - expand procfs space to handle large clusters (2326)
5687 - increase UML stack to avoid overflow
5688 - update lconf's list of debug and subsystem masks
5689 - fix lfs find --obd (2510)
5690 - /proc tunable for disabling filter read caching (2591)
5691 - stop rpm packages from altering slapd.conf (2301)
5692 - disable nagle in the socknal under 0conf (2578)
5693 - choose mds inode size based on stripe count (2572)
5694 - fix kernel-source rpm problems (2516)
5696 - add --disable-doc to avoid pdf generation (2421)
5697 - update documentation, tests, type-os, comments
5698 - avoid format warnings on ia64
5699 - remove the TOE NAL
5700 - tiny code cleanups by removing unused fields
5702 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5705 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5706 - remove the most blatant lies from BUILDING file (2371)
5707 - change default debug level to reasonable production setting
5708 - reduce client side cache size to reduce cache flush time
5709 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5710 - make TCP zerocopy and pinger support enabled by default (2476)
5711 - sync writes completed after process exits caused crashes (2319)
5712 - maintain correct mount count on the MDS (2356)
5713 - backout 1557, because 2316 wasn't really fixed
5714 - better file I/O statistics gathering in /proc
5715 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5716 - another kernel patch to fix zero-copy TCP function export
5717 - don't take duplicate lock when processing re-sent getattr (2420)
5718 - lctl uses obd_self_export instead of creating new conn (2353)
5719 - MDS/OST recovery case which requires object creation asserted (2425)
5720 - move lfs from /usr/sbin to /usr/bin in packages
5721 - fix race between mds_client_add and mds_client_free (2417)
5722 - use kmalloc instead of slabs in portals (2430)
5723 - don't create duplicate records when a failover MDS is present (2442)
5724 - remove unnecessary mount age check (2332)
5725 - don't remove directory inodes from locks prematurely (2451)
5726 - don't break if MDS service name is the same as hostname (2103)
5727 - fix races in client write RPC generation when cache full (2482)
5729 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5732 - remove now-unused request->rq_obd (278)
5733 - if an allocation fails, print out how much memory we've used (1933)
5734 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5735 - add the "configurable stack size" patch to most series files (1256)
5736 - ability to write large log records, for 100+ OST configs (2306)
5737 - fix NULL deref when filter_prep fails (2314)
5738 - fix operator precedence error in filter_sync
5739 - dynamic allocation of socknal TX descriptors (2315)
5740 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5741 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5742 - ensure that CWARN messages in Portals always get to the syslog
5743 - __init/__exit are not for prototype decls (ldlm_init/exit)
5744 - x86-64 compile warning fixes
5745 - fix gateway LMC keyword conflict (2318)
5746 - fix MDS lock inversions in getattr/reint paths (1844)
5747 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5748 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5749 - DLM race condition prevented some lock evictions (2328)
5750 - ENOMEM detection and retry on socknal sends (2230)
5751 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5752 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5753 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5754 - better error messages when a client is rejected during recovery (1505)
5755 - avoid cancelling locks which were never granted, after failure (2330)
5756 - fix i_sem/journal inversion in mds_client_add (2333)
5757 - fix truncate/getattr lock cycle deadlock (2334)
5758 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5759 - fix two rare exit paths which could leak an l_lock() ref (2321)
5760 - fencepost error in MDS/OST orphan recovery (2226)
5761 - make log record alignment 8 bytes (1988)
5762 - lstripe now fails when requested offset > ost_count (2237)
5763 - ensure that all kernel series have a complete list.h (1607)
5764 - fix crashes in special-file operations (2316)
5765 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5767 - allow configurable automake binary, for testing new versions
5768 - small update to the lfs documentation
5770 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5772 * fix negative export reference count in fsfilt_sync (2312)
5774 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5775 * release candidate 0.9.1
5777 - orphans are moved into the PENDING directory for possible recovery
5778 - replayed opens now open by fid for orphan/rename safety (1042)
5779 - last close of an orphan inode generates a transno (683)
5780 - chdir() and mount() now pin the directory entry (1020)
5781 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5782 - discard very old requests without processing them (1502)
5783 - remove some common, well-understood CERRORs (1505)
5784 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5785 - clear "grant" flags in OST replies until OST grant code lands (1644)
5786 - fix read performance by not clobbering i_blksize on client (1598)
5787 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5788 - make LRU size a /proc tunable, clears locks when reduced (707)
5789 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5790 - server locks take references on exports now (1558)
5791 - build fixes for 2.4.20-rh trees (1663)
5792 - return an error from lov_create if all OSCs are inactive (1751)
5793 - fix import levels when a reconnect happens without a timeout (1597)
5794 - exit early from mds_open if we get a lookup error (1749)
5795 - partial page read at EOF wouldn't wait for disk before sending (1642)
5796 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5797 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5798 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5799 - write/truncate lock inversion (1639)
5800 - Don't auto-load obdclass, portals modules during cleanup (1495)
5801 - fix timestamps from jumping to "now" (1763)
5802 - extra journal assertions (1648)
5803 - add an extra multiunlink test (1771)
5804 - fix read_record/write_record API (1776)
5805 - fix leak of offset_extent, possible incorrect i_size later (1772)
5806 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5807 - Debugging for the kqswnal_get_idle_tx problems (1820)
5808 - Allow recovery to be attempted multiple times (1536)
5809 - Write out MDS last_rcvd file after it is first created (1600)
5810 - Fix tx_descriptor leak in failed transmit situations (1827)
5811 - ext3 journaling fixes for assertion failure after IO error (1871)
5812 - class_export_put() on freed export after completion AST error (1896)
5813 - Fix revalidate looping in VFS (1322)
5814 - Don't access a freed export during MDS_REINT timeout (1521)
5815 - Add open-unlink recovery support on the MDS (1673,1764)
5816 - Return an error if no MDS data was read from last_rcvd (1946)
5817 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5818 - Don't setup a disk device that doesn't match exported UUID (317)
5819 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5820 - avoid committing NULL handle in force close
5821 - local.sh is now a one-stripe LOV configuration
5822 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5823 - fix ext3 htree duplicate directory entry corruption (1516)
5824 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5825 - update to new LOV EA format (2097)
5826 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5828 2003-06-15 Phil Schwan <phil@clusterfs.com>
5831 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5832 - per-import/export recovery handling (958, 931, 959)
5833 - multiple last-rcvd slots, for serving multiple FSes (949)
5834 - connections are again shared between multiple imp/exports (963, 964)
5835 - "umount -f" would hang if any requests needed to be sent (393, 978)
5836 - avoid pinning large req buffer by copying for queued messages (989)
5837 - add "uuid" to "lctl device" command to help upcalls (991)
5838 - "open" RPCs with transnos would confuse recovery counters (1037)
5839 - do proper endian conversion of all wire messages (288, 340, 891)
5840 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5841 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5842 - fix ptlrpc_connection leak in target_handle_connect (1174)
5843 - fix import refcounting bug in OST and MDS cleanup (1134)
5844 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5845 - fix very unlikely obd_types race condition (501)
5846 - remove osc_open hack for echo_client (1187)
5847 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5848 - a failure in read_inode2 leads to deadlock (1139)
5849 - cancel ack-locks as soon as transaction is committed (1072)
5850 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5851 - make sure to commitrw after any preprw to avoid deadlock (1162)
5852 - failing to execute a file in a lustre FS would lock inode (1203)
5853 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5854 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5855 - obd_uuid2tgt would walk off the end of the list (1255)
5856 - on IA64 the osc would give portals incorrect bulk size (1258)
5857 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5858 - fix lock inversion caused by new llite matching code (1282)
5859 - limit the number of dirty pages on a client to 10MB (1286)
5860 - timed out locks were not being corrected cancelled (1289)
5861 - fix O_DIRECT above 4GB on IA-32 (1292)
5862 * major user-visible changes
5863 - fail out/fail over policy now controlled by the upcall (993)
5865 - add OBD_PING to check server availability and failure (954)
5866 - lustre messages are now sent in sending host order (288, 340, 891)
5867 - add eadatalen to MDS getattr reply (340)
5868 - OST read replies may contain second buffer, with per-page status (593)
5870 2003-03-11 Phil Schwan <phil@clusterfs.com>
5873 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5874 - failed open()s could cause deadlock; fixed (867, 869)
5875 - stop cancelling OST locks when files are closed (481)
5876 - overlapping XID spaces caused network corruption (851, 853)
5877 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5878 - setattr_raw added, to do single-RPC, server-side setattrs
5879 - lmc/lconf syntax change for OST UUIDs
5880 - fix crashy race condition between ptlrpc_free_req and osc_close
5881 - don't use request in mdc_enqueue if we hit a timeout (889)
5882 - don't set the inode i_size for regular files from the MDS (896)
5883 - handle out of order completion AST (842)
5884 - don't LBUG if a lock request times out after receiving AST (913)
5885 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5886 - if a bad lock AST arrives, send an error instead of dropping entirely
5887 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5888 - fix leak in bulk IO when only partially completed (899, 900, 926)
5889 - fix O_DIRECT for ia64 (55)
5890 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5891 - C-z after timeout could hang a process forever; fixed (977)
5893 - client-side I/O cache (678, 924, 929, 941, 970)
5895 - READPAGE and SETATTRs which don't take server-side locks get
5898 2003-02-11 Phil Schwan <phil@clusterfs.com>
5901 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5902 - Fix fsfilt_extN_readpage() to read a full page of directory
5903 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5904 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5905 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5906 - Most intent operations were removed, in favour of a new RPC mode
5907 that does a single RPC to the server and bypasses most of the VFS
5908 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5909 - Aggressively cancel local locks on DLM servers
5910 - mds_reint_unlink sends EA to the client if it's the last nlink.
5911 client uses that EA to unlink OST objects.
5912 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5913 - recursive symlinks were fixed (439)
5914 - fixed NULL deref in DEBUG_REQ
5915 - filter_update_lastobjid no longer calls sync, which annoyed extN
5916 - fixed multi-client small-writes to a single file problem (445)
5917 - fixed mtime updates during file writes (607)
5918 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5919 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5920 - fixed Linux OST size reporting problem (444, 656)
5921 - OST now updates object mtime with writes or setattr (607, 619)
5922 - client verifies file size before zeroing page past EOF (445)
5923 - OST now writes last allocated objid to disk with allocation (108)
5924 - LOV on echo now works (409)
5926 - mds_reint_unlink sends a new buffer, with the EA included. this
5927 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5928 set if a regular file was being unlinked, and it was the last link
5929 - use PtlGet from the target for bulk writes (315)
5930 - OST now updates object mtime with writes or setattr (607, 619)
5931 - LDLM now has a grant-time callback to revalidate locked items, if
5933 - Many MDS operations were reorganized to combat race conditions
5935 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5936 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5938 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5941 - Fully reactivate OST imports after reconnection (512, others)
5942 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5943 - More graceful error handling for truncating on dead OST (515)
5944 - Don't error out unless we're actually accessing dead stripes (474)
5945 - Fix garbage sizes when stripes are missing (410)
5946 - LRU counters were broken, causing constant lock purge (433, 432)
5947 - garbage on read from stripes with failed OSTs (441)
5948 - mark OSCs as active before reconnecting during recovery (438)
5949 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5950 - lfind did not preserve OST order in output (443)
5951 - symlinks cause hung clients, incorrect data (439)
5952 - stop dereferencing request after dropping refcount (457)
5953 - don't LASSERT(spin_is_locked) on non-SMP (455)
5954 - fixes for many rename() bugs
5955 - fstat didn't correctly synchronize attributes (399)
5956 - server must handle lock cancellation during blocking AST prep (487)
5957 - bulk descriptors were free()d too soon (511)
5958 - fix paths in lconf, which would load incorrect modules (451, 507)
5959 - fix confusing lconf 'host not found' error message (386)
5960 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5961 - fix race condition in mdc_blocking_ast() for inode access (526)
5962 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5963 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5964 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5965 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5966 - fix lstripe in several ways (kernel side) (527)
5967 - fix request leak in ldlm_cli_enqueue (262)
5968 - incorrect OSC was marked inactive after OST failure
5969 - call mds_fs_cleanup before unmounting filesystem (524)
5970 - fix races between taking ns_lock and ldlm_lock_change_resource
5971 - fix races updating LOV export open file list
5972 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5973 - fix recovery NULL deref in ldlm_cli_cancel_unused
5974 - fix some DLM races by using new hash table for lock handles (419)
5975 - permit the client to specify desired inodes, at replay
5976 - duplicate requests when we queue them for replay reintegration
5977 - fix last_rcvd offset calculation
5978 - sync after each recovered transaction, so we always make progress
5979 - never, not always, ERESTART requests without transnos
5980 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5982 - skip replay if the MDS says that the client is already connected
5983 - don't check for a recovery-enabled export to match lctl's UUID
5984 - don't INC_USE_COUNT for phantom exports
5985 - don't crash when cleaning up phantom exports (567)
5986 - don't double-finish or set replay data for errored mdc_open requests
5987 - abort requests when they time out, so we don't get old replies
5988 - send/receive replies for AST messages again
5989 - if the client says that it doesn't have the lock, cancel it on the
5991 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5992 mark it as destroyed, it will all work out when decref is called
5993 - fix module use counts (22, 581)
5995 - ASTs now expect a reply (server cancels lock on error reply)
5997 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6000 - fix many simultaneous client startup (392)
6001 - fix dentry->d_it clobbering
6002 - credentials weren't being shipped for readdir/getattr operations
6003 - remove invalid assertions triggered during some concurrent MD
6005 - proper Lustre versions added (336, 389)
6006 - fix memory leak for create error case (398)
6007 - fix LOV locking bug that would get cli/srv out of sync
6008 - fix echo client over LOV (409)
6009 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6010 - fix double-O_EXCL intent crash (424)
6011 - avoid sending multiple lock CANCELs (352)
6013 - MDS can do multi-client recovery (modulo bugs in new code)
6015 - many updates, edits, cleanups
6017 2002-11-18 Phil Schwan <phil@clusterfs.com>
6020 - fix null d_it dereference (346)
6021 - fix full OST/dbench hang (333)
6022 - fix permission problem with file removal (286)
6023 - fix removal of OSCs from LOV when they fail
6024 - fix NULL deref during bulk timeout (214)
6025 - fix problems related to multiple filesystems on one MDS (241)
6026 - fixed serious subtle metadata locking bugs
6027 - free locks on clients when inodes are removed due to memory
6029 - fix inode pointer in lock data (285)
6030 - partial support for multiple MDS on a single host (241)
6031 - data locks weren't cancelled at clear_inode time (290, 311)
6032 - intent locks could lead to unbounded lock growth (205)
6033 - added a maximum lock count, an LRU list, and a flusher
6034 - fix multiple rename (365)
6035 - properly abstracted the echo client
6036 - OSC locked 1 byte too many; fixed
6037 - rewrote brw callback code:
6038 - fixed recovery bugs related to LOVs (306)
6039 - fixed too-many-pages-in-one-write crash (191)
6040 - fixed (again) crash in sync_io_timeout (214)
6041 - probably fixed callback-related race (385)
6043 - Add capability to MDS protocol
6044 - LDLM cancellations and callbacks on different portals
6046 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6049 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6050 - defer open object destruction to close time (601981, 138)
6051 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6052 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6053 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6054 - don't call obd_setattr in ll_file_release for destroyed objects
6055 * protocol change to lustre_msg: move |version| and add |flags|
6056 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6057 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6058 * added replay of create, unlink, link and rename operations during
6059 MDS failover; recovery should be much more robust now
6060 * remove failed OSCs from LOVs (only lov_create uses this so far)
6061 * the lustre-HOWTO was brought (more) up to date (582544)
6063 2002-10-23 Phil Schwan <phil@clusterfs.com>
6066 - in-use dentries weren't being reused properly (617851)
6067 - prevent multiple LDLM setup (599178)
6068 - fix LOV size calculations for truncate (617853)
6069 - fix client handling of MDS intent errors (POSIX)
6070 - fix permission bug in lovstripe.c test (624321)
6071 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6072 - truncate past end of file could corrupt data
6073 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6074 - a race in recovery could return ETIMEDOUT to apps (623947)
6075 - building outside the source directory was fixed
6076 * the lustre-HOWTO was brought (more) up to date (582544)
6077 * major progress was made on recovery functionality
6079 2002-10-10 Phil Schwan <phil@clusterfs.com>
6082 - recovery deadlock fix
6083 - rm -rf causes LBUG fix (617817)
6084 - file open by multiple tasks fix (618962)
6085 - directory permissions bugs (602707 and 620007)
6086 - journal_stop fixed with locking (611313)
6087 - O_APPEND failures resolved (618273, perhaps 614459)
6088 - lconf PATH fix (619770)
6089 - IA64 build fix (621450)
6090 - RPC buffer sizes scale with amount of memory
6092 2002-10-01 Phil Schwan <phil@clusterfs.com>
6095 - locks would be cancelled without throwing away data pages,
6096 resulting in inconsistent data (605627)
6097 - inode attributes were not always being refreshed (605627, 612449)
6098 - lconf now continues to cleanup after lctl reports an error
6099 - MDS now enforces user permissions (602707)
6100 - lprocfs cleanup fixed, but not yet enabled (614157)
6101 - fixed infinite server hang, should a client not respond to an AST
6102 - avoid going into recovery if user calls readlink() with a buffer
6103 that's too small (613941)
6104 - AST RPCs no longer require replies (614867) -- this may be changed
6105 - don't crash server if client sends an IOV that's too big (611336)
6106 - fixed lock conversion deadlock (611892)
6107 - fixed the following of symlinks (614622)
6108 * recovery: the server can remove locks from a client that dies, other
6109 clients can make progress
6110 * more extN patch fixes
6111 * compile-time configurable ptlrpc buffer allocations
6113 - collaborative read cache document
6114 - Lustre Lite Performance CDR document-in-progress
6116 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6119 - fix typo in patch-2.4.18
6121 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6124 - clear ptlrpc request each time in handle_incoming_request()
6125 - unlink of files now destroys the object on the OST
6127 2002-09-19 Peter Braam <braam@clusterfs.com>
6129 * add hard link support
6130 * change obdfile creation method
6131 * kernel patch changed
6133 2002-09-19 Peter Braam <braam@clusterfs.com>
6136 - stack overflow bug in extN fixed
6138 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6140 * documentation updates
6141 - add man pages for config tools
6142 - update tests/README to describe testing with new config tools
6143 - finish metadata API descriptions
6144 * bug fixes and cleanups
6145 - statfs workaround for 16TB limit
6146 - LOV stripe allocation improved, can stripe on subset of OSTs
6147 - LOV file size/IO offset was wrong for files > 4GB in size
6148 - object EA data was being dropped, caused files to be unreadable
6149 - memory overflow with non-LOV OST caused memory corruption
6150 - fixed regression tests to work with new config tools, obdfilter
6151 - fixed bug when directory size became larger than 1 block
6152 - fixed bug (for single client case) when PWD was deleted
6153 - invalidate local directory pages when doing intent-based ops
6154 - avoid LDLM oops when lock callback contained bad data
6156 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6158 * documentation updates
6159 * bug fixes and cleanups
6160 - configuration tools
6163 - 64-bit compile warnings
6164 - 64-bit internal statfs data
6166 * test_brw on persistent OST devices
6168 * lprocfs (disabled)
6170 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6172 * documentation updates
6173 * bug fixes and cleanups
6174 * configuration tools
6176 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6179 * many small fixes to 0_5_4
6180 * io/network handling
6181 * thinkos in MDS operations
6183 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6186 * crucial basic fixes to 0.5.3
6187 * IOR, Iozone work over Elan
6190 2002-08-07 Phil Schwan <phil@clusterfs.com>
6191 * version 0_5_3, our first alpha
6192 * we use the new Portals iovs
6193 * documentation updates
6194 * bug fixes and cleanups
6195 * small changes in the DLM wire protocol
6197 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6198 * version 0_5_1 with some initial stability,
6199 * locking on MD and file I/O.
6200 * documentation updates
6201 * several bug fixes since 0.5.0
6202 * small changes in wire protocol
6204 2002-07-18 Phil Schwan <phil@clusterfs.com>
6206 * delivered as Lustre Light Alpha
6207 * fixed a crash after handling invalid MDS requests
6208 * fixed directory pages for architectures with non-4k pages sizes
6210 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6211 * release version v0_4_4
6212 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6214 * Fixes a number of interruption problems with OST operations.
6215 * Update documentation for portals header changes
6216 * Move all wire protocol structs/defines to lustre_idl.h
6217 * Fixes symlink length bug.
6218 * Add tcpdump to repository.
6220 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6221 * release version v0_4_3
6222 * Fixes statfs for inodes on extN.
6223 * Fixes bug in runtests which would delete /etc/hosts.
6224 * Use 64-bit object IDs wherever possible (not into VFS though)
6225 Remove ost_get_info, which is unused by lustre, and out of date.
6227 2002-07-03 Peter Braam <braam@clusterfs.com>
6228 * release version v0_4_2 Fixes a lookup error (type not passed)
6229 * move forward to head of Portals
6230 * move forward to latest Lustre kernel
6232 2002-06-25 Peter Braam <braam@clusterfs.com>
6233 * release version v0_4_1. Hopefully stable on single node use.