3 * Server support for kernels:
4 2.6.18-164.11.1.el5 (RHEL 5)
5 2.6.18-164.11.1.0.1.el5 (OEL 5)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-164.11.1.el5 (RHEL 5),
9 2.6.18-164.11.1.0.1.el5 (OEL 5)
10 2.6.16.60-0.42.8 (SLES 10),
12 2.6.29.4-167.fc11 (FC11)
13 * Recommended e2fsprogs version: 1.41.10-sun2
14 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
15 Lustre filesystem with 4K stack may cause a stack overflow. For more
16 information, please refer to bugzilla 17630.
18 --------------------------------------------------------------------------------
20 2010-07-15 Oracle, Inc.
22 * Server support for kernels:
23 2.6.18-164.11.1.el5 (RHEL 5)
24 2.6.18-164.11.1.0.1.el5 (OEL 5)
25 * Client support for unpatched kernels:
26 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
27 2.6.18-164.11.1.el5 (RHEL 5),
28 2.6.18-164.11.1.0.1.el5 (OEL 5)
29 2.6.16.60-0.42.8 (SLES 10),
31 2.6.29.4-167.fc11 (FC11)
32 * Recommended e2fsprogs version: 1.41.10-sun2
34 Severity : enhancement
36 Description: Distribute lfs_migrate script with lustre
37 Details : The lfs_migrate mentioned in the manual is now included in the
38 lustre tree in lustre/scripts and will distributed in lustre RPMs.
41 Severity : enhancement
43 Description: Failover nid list added to import proc file.
45 Severity : enhancement
47 Description: "som_preview" mount option required to use SOM
48 Details : The clients will need to use som_preview mount option
52 Description: Lustre does not do 1MB IOs to HW RAID
53 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
55 Severity : enhancement
57 Description: lfs quota output is not very convenient for awk/sed-parsing
58 Details : Some positions in lfs quota output table could be empty or
59 non-empty which made it hard to parse it with scripts, now
60 a dash is put instead of space where there is not supposed
65 Description: Update to OFED 1.5.1
66 Details : OFED 1.5.1 will be used in environments that don't use
71 Description: Remove lnet. prefix from lctl params display
72 Details : Also renamed lnet.memused to lnet_memused to
77 Description: Use conf_param -d to remove permanant settings.
81 Description: Add list_param -R
82 Details : lctl list_param now supports recursive mode.
85 Bugzilla : 21829, 22312
86 Description: Fix llobdstat.
87 Details : 1. Now updating the write_bytes counter
88 2. Port script fixes from 1.8.
92 Description: decrease the usage of memory on clients.
93 Details : 1. On clients, recycle dentries and inodes unused.
94 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
99 Description: llobdstat fix and enhancement
100 Details : add a counter to set a limit to how many samples will be returned
101 fix a wildcard in the path to limit to obdfilter stats only
105 Description: lfs check only allowed for root
106 Details : Code cleanups and added a regresion test of non-root use of
111 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
112 Details : Fix hr->hr_index overload.
116 Description: Fix FIEMAP functionality.
120 Description: FC11 patchless client support added
124 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
125 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
129 Description: New acc-sm script and various test-framework enhancments.
133 Description: Previously landed patch was reverted.
137 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
141 Description: Add ldiskfs maxdirsize mount option.
145 Description: Fix mkfs.lustre support for 16TB LUNs.
149 Description: Print mmp_check_interval
150 Details : print mmp_check_interval and make it possible to abort
151 mount operation in case it takes too long.
155 Description: Add cascading_rw to lustre mpi tests.
159 Description: Store SOM data in EA.
160 Details : Also several SOM related cleanups.
164 Description: Add refcount to md_open_data
168 Description: Use refcount on lov_request_set to prevent memory corruption.
172 Description: Statahead: use parent dir's i_mutex for synchronization.
176 Description: loadgen improvements.
180 Description: Enable peer health detection in ksocklnd and ko2iblnd.
184 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
188 Description: Fix race on obd_devcie:md_stats
192 Description: Fix SLES 11 client builds.
193 Details : Resolved uninitalized pointer gcc warnings.
197 Description: Allow non-root access for "lfs check".
198 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
202 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
204 Severity : enhancement
206 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
207 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
211 Description: Sanity test_27p fails.
212 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
213 Sanity tests 27 were cleaned up and improved.
217 Description: Unify req format on client/servers
218 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
219 buffers and swabbers approach..
224 Description: drop join file stuff.
226 Severity : enhancement
228 Description: Add 2.6.27 and SLES11 (patchless client) support
232 Description: Fix in ptlrpc_expire_one_request() to print the signed time
235 Severity : enhancement
237 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
240 Severity : enhancement
242 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
244 Severity : enhancement
246 Description: correctly handle big reply message.
247 Details : send LNet event if reply is bigger then buffer and adjust this buffer
252 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
254 Severity : enhancement
256 Description: Add support for OEL5.
260 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
262 Severity : enhancement
264 Description: "weak-modules" support
265 Details : Implement "weak-modules" support which enables kernel modules
266 to be used with any kernel that implements the same kABI. In
267 order to achieve this modules are now installed in
268 /lib/modules/$(uname -r)/updates/kernel on all distributions.
272 Description: client could not reconnect to OST because of an active request.
273 Details : The client now retries to reconnect to the same server, if a
274 connect request failed with EBUSY or -EAGAIN. Test case 35b
275 conf-sanity.sh has been created.
279 Description: Don't allow make backward step on assignin osc next id.
280 Details : race between allocation next id and ll_sync thread can be cause
281 of set wrong osc next id and can be kill valid ost objects.
285 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
289 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
293 Description: Error handling in osc_statfs_interpret() has been improved.
294 Details : Check in osc_statfs_interpret() for EBADR.
298 Description: Do not send statfs() requests to OSTs disabled by administrator.
299 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
301 Severity : enhancement
303 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
305 Severity : enhancement
307 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
309 Severity : enhancement
311 Description: Enhancement for lfs(1) command to use numeric uid/gid.
315 Description: Improved cleanup procedure for conf-sanity test_46a.
319 Description: Do not start bulk transfer, if client tried to reconnect.
320 Details : Do not start bulk transfer, if client tried to reconnect,
321 but failed due to active rpcs. Check for exp_abort_active_req
322 has been implemented.
326 Description: Prevent inconsistences between linux and lustre mount structures.
327 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
328 Make the sleep interruptible.
330 Severity : enhancement
332 Description: provide server to client comms path
333 Details : server to client communications path via new LDLM_SET_INFO rpc
337 Description: truncate starts GFP_FS allocation under transaction causing deadlock
338 Details : ldiskfs_truncate calls grab_cache_page which may start page
339 allocation under an open transaction. This may lead to
340 calling prune_icache with consequent lustre reentrance.
344 Description: send correctly lsm on open replay
345 Details : MDS is trust to LSM size on replay open, but client can set wrong size
350 Description: enable client interface failover
351 Details : When a child reconnects from another NID, properly update export
352 nid hash position and ldlm reverse import.
354 Severity : enhancement
356 Description: Communicate OST degraded/readonly state via statfs to MDS
357 Details : Flags in the statfs returned from OSTs indicate whether the
358 OST is in a degraded RAID state, or if the filesystem has
359 turned read-only after a filesystem error is detected.
364 Description: don't panic if EPROTO was hit when reading symlink
365 Details : correctly handling request reference in error cases.
367 Severity : enhancement
369 Description: Add LustreNetLink, a kernel-userspace communcation path.
371 Severity : enhancement
373 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
378 Description: access to llog context before init.
379 Details : move handling CATALOGS file at osc layer and forbid access to llog
383 Frequency : rare, only when client is writing to OST on local node
385 Description: Avoid deadlock for local client writes
386 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
387 memory freeing context. This allows OST threads to set the
388 PF_MEMALLOC flag on task structures in order to allocate memory
389 from reserved pools and complete IO.
390 Use GFP_HIGHUSER for OST allocations for non-local client writes,
391 so that the OST threads generate memory pressure and allow
392 inactive pages to be reclaimed.
394 Severity : enhancement
396 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
401 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
402 Details : move ldlm namespace creation in setup phase to avoid grab
403 _lprocfs_lock with cli_sem held.
407 Description: Temporarily disable grant shrink.
408 Details : Disable the feature for debugging.
411 Frequency : only when formatting test filesystems
413 Description: Unable to run several mkfs.lustre on loop devices at the same time
414 Details : mkfs.lustre returns error 256 on the concurrent loop devices
415 formatting. The solution is to proper handle the error.
417 Severity : enhancement
419 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
421 Severity : enhancement
423 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
428 Description: resolve race between obd_disconnect and class_disconnect_exports
429 Details : if obd_disconnect will be called to already disconnected export he
430 forget release one reference and osc module can't unloaded.
432 Severity : enhancement
434 Description: Allow tuning service thread via /proc
435 Details : For each service a new
436 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
437 created that can be used to set min/max thread counts, and get the
438 current number of running threads.
443 Descriptoin: don't return error if have partially created objects for file.
444 Details : lov_update_create_set uses set->set_success as index for created
445 objects, so if some requests will be failed, they will have hole
446 at end of array and qos_shrink_lsm for allocate correct lsm.
448 Severity : enhancement
450 Description: Update OFED support to 1.4.1
455 Description: Reduce small size read RPC
456 Details : Set read-ahead limit for every file and only do read-ahead when
457 available read-ahead pages are bigger than 1M to avoid small size
460 Severity : enhancement
462 Description: move AT tunable parameters for more consistent usage
463 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
465 Severity : enhancement
467 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
468 Details : allow skip disconnected ost for send statfs request and hide error
474 Description: fix racy locking of mballoc block bitmaps causing BUG
475 Details : The locking of the mballoc buddy bitmap and the in-memory
476 block bitmap was using two different spin locks in some
477 cases. This made it possible to incorrectly access the
478 mballoc bitmap while another process was modifying it,
479 causing a sanity assertion to fail. While no on-disk corruption
480 was reported, there was some risk of this happening.
483 Frequency : rare, on llog test 6
485 Descriptoin: don't allow connect to already connected import
486 Details : allowing connect to already connected import is hide connecting problem.
491 Descriptoin: allow kill process which wait statahead result
492 Details : for some reasons 'ls' can stick in waiting result from statahead,
493 in this case need way for kill this process.
495 Severity : enhancement
497 Description: Add state history info file, enhance import info file
498 Details : Track import connection state changes in a new osc/mdc proc file;
499 add overview-type data to the osc/mdc import proc file.
501 Severity : enhancement
503 Description: MDS create should not wait for statfs RPC while holding DLM lock.
505 Severity : enhancement
507 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
512 Descriptoin: OST grant too much space to client even there are not enough space.
513 Details : Client will shrink its grant cache to OST if there are no write
514 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
515 this grant cache if there are already not enough avaible space
516 (left_space < total_clients * 32M).
519 Frequency : start MDS on uncleanly shutdowned MDS device
521 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
522 Details : stay in waiting mds<>ost recovery finished produce random bugs
523 due race between two ll_sync thread for one lov target. send
524 ACTIVATE event only if connect realy finished and import have
528 Frequency : rare, connect and disconnect target at same time
530 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
531 Details : don't call obd_disconnect under lov_lock. this long time
532 operation and can block ptlrpcd which answer to connect request.
537 Descriptoin: don't lose wakeup for imp_recovery_waitq
538 Details : recover_import_no_retry or invalidate_import and import_close can
539 both sleep on imp_recovery_waitq, but we was send only one wakeup
543 Frequency : always with long access acl
545 Descriptoin: mds can't pack reply with long acl.
546 Details : mds don't control size of acl but they limited by reint/getattr
549 Severity : enhancement
551 Description: Update to SLES10 kernel-2.6.16.60-0.33.
553 Severity : enhancement
555 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
558 Frequency : start MDS on uncleanly shutdowned MDS device
560 Descriptoin: aborting recovery hang on MDS
561 Details : don't throttle destroy RPCs for the MDT.
564 Frequency : on remount
566 Description: external journal device not working after the remount
567 Details : clear dev_rdonly flag for external journal devices in
573 Description: shutdown vs evict race
574 Details : client_disconnect_export vs connect request race.
575 if client will evicted at this time - we start invalidate
576 thread without referece to import and import can be freed
580 Frequency : rare, need acl's on inode.
582 Description: client can't handle ost additional correctly
583 Details : if ost was added after client connected to mds client can have
584 hit lnet_try_match_md ... to big messages to wide striped files.
585 in this case need teach client to handle config events about add
586 lov target and update client max ea size at that event.
588 Severity : enhancement
590 Description: Changelogs
591 Details : Changelogs are a lightweight mechanism to track filesystem
592 metadata and namespace changes. The changelog is recorded
593 permanently on the MDTs, and is periodically "consumed" / purged
594 when records are no longer needed.
596 Severity : enhancement
598 Description: compact fld format with extents
599 Details : Store range of seq rather than every seq in FLD. Seq
600 controller update FLD rather than clients. In Case of CMD, mdt0
601 has FLD, all other metadata server act as non persistent proxy
602 for FLD queries and cache fld entries in fld cache.
607 Description: don't skip ost target if they assigned to file
608 Details : Drop slow OSCs if we can, but not for requested start idx.
609 This means "if OSC is slow and it is not the requested
610 start OST, then it can be skipped, otherwise skip it only
611 if it is inactive/recovering/out-of-space.
615 Description: more cleanup in mds_lov
616 Details : not send LOV EA under replay, we can't know about they size at this
617 time. Don't allow client connect to mds before any ost connected,
618 for avoid problems with LOV EA size and returning EIO to client.
620 Severity : enhancement
622 Description: Interoperability at server side (Disk interoperability)
624 Severity : enhancement
626 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
628 Severity : enhancement
630 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
632 Severity : enhancement
634 Description: New client IO stack (CLIO).
636 Severity : enhancement
638 Description: Commit on sharing. Eliminate inter-client dependencies between
639 uncommitted transactions by doing transaction commits.
640 Thereby clients may recovery independently.
643 Frequency : Create a symlink file with a very long name
645 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
646 Details : If there is no extra space in the request for early cancels,
647 ldlm_req_handles_avail() returns 0 instead of a negative value.
649 Severity : enhancement
651 Description: Add /proc entry for import status
652 Details : The mdc, osc, and mgc import directories now have
653 an import directory that contains useful import data for debugging
656 Severity : enhancement
658 Description: Re-disable certain /proc logging
659 Details : Enable and disable client's offset_stats, extents_stats and
660 extents_stats_per_process stats logging on the fly.
663 Frequency : Only on FC kernels 2.6.22+
665 Description: oops in statahead
666 Details : Do not drop reference count for the dentry from VFS when lookup,
667 VFS will do that by itself.
669 Severity : enhancement
671 Description: Generic /proc file permissions
672 Details : Set /Proc file permissions in a more generic way to enable non-
673 root users operate on some /proc files.
677 Description: Hitting mdc_commit_close() ASSERTION
678 Details : Properly handle request reference release in
679 ll_release_openhandle().
683 Description: quota recovery deadlock during mds failover
684 Details : This patch includes att18982, att18236, att18237 in bz14840.
686 1. fix osts hang when mds does failover with quotaon
687 2. prevent watchdog storm when osts threads wait for the
692 Frequency : only patchless client
693 Description: add workaround for race between add/remove dentry from hash
695 Severity : enhancement
697 Description: Allow OST glimpses to return PW locks
701 Description: LBUG when llog conf file is full
702 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
706 Description: Prevent import from entering FULL state when server in recovery
710 Description: service mount cannot take device name with ":"
711 Details : Only when device name contains ":/" will mount treat it as
717 Description: replace ptlrpcd with the statahead thread to interpret the async
718 statahead RPC callback
722 Frequency : on recovery
723 Description: I/O failures after umount during fail back
724 Details : if client reconnected to restarted server we need join to recovery
725 instead of find server handler is changed and process self
726 eviction with cancel all locks.
728 Severity : enhancement
730 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
732 Severity : enhancement
734 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
736 Severity : enhancement
738 Description: Upcall on Lustre log has been dumped
739 Details : Allow for a user mode script to be called once a Lustre log has
740 been dumped. It passes the filename of the dumped log to the
741 script, the location of the script can be specified via
742 /proc/sys/lnet/debug_log_upcall.
747 Description: avoid idr_remove called for id which is not allocated.
748 Details : Move assigment s_dev for clustered nfs to end of initialization,
749 to avoid problem with error handling.
754 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
755 Details : When connection is reused this not moved from CONN_UNUSED_HASH
756 into CONN_USED_HASH and this prodice warning when put connection
757 again in unused hash.
760 Severity : enhancement
762 Description: File striping can now be set to use an arbitrary pool of OSTs.
764 Severity : enhancement
766 Description: Export bytes_read/bytes_write count on OSC/OST.
770 Description: Early reply size mismatch, MGC loses connection
771 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
772 the connect flags are properly negotiated.
777 Description: quotas are not honored with O_DIRECT
778 Details : all writes with the flag O_DIRECT will use grants which leads to
779 this problem. Now using OBD_BRW_SYNC to guard this.
783 Description: add quota statistics
784 Details : 1. sort out quota proc entries and proc code.
785 2. add quota statistics
787 Severity : enhancement
789 Description: enable quota support for HEAD.
793 Description: Properly propagate oinfo flags from lov to osc for statfs
794 Details : restore missing copy oi_flags to lov requests.
796 Severity : enhancement
798 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
799 : l_getgroups(8), lst(8), routerstat(8)
800 Details : included man pages for llobdstat(8), llstat(8),
801 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
803 Severity : enhancement
805 Description: configure's --enable-quota should check the
806 : kernel .config for CONFIG_QUOTA
807 Details : configure is terminated if --enable-quota is passed but
808 : no quota support is in kernel
812 Description: Remove portals compatibility
813 Details : Remove portals compatibility, not interoperable with releases
818 Description: Resolve device initialization race
819 Details : Prevent proc handler from accessing devices added to the
820 obd_devs array but yet be intialized.
822 Severity : enhancement
824 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
826 Severity : enhancement
828 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
833 Description: Using wrong pointer in osc_brw_prep_request
834 Details : Access to array[-1] can produce panic if kernel compiled with
835 CONFIG_PAGE_ALLOC enabled
839 Description: Client runs out of low memory
840 Details : Consider only lowmem when counting initial number of llap pages
844 Description: *optional* service tags registration
845 Details : if the "service tags" package is installed on a Lustre node
846 When the filesystem is mounted, a local-node service tag will
847 be created. See http://inventory.sun.com/ for more information
848 about the Service Tags asset management system.
852 Description: Kernel BUG tries to release flock
853 Details : Lustre does not destroy flock lock before last reference goes
854 away. So always drop flock locks when client is evicted and
855 perform unlock regardless of successfulness of speaking to MDS.
859 Description: add refcount protection for osc callbacks, avoid panic on shutdown
863 Description: sanity test 65a fails if stripecount of -1 is set
864 Details : handle -1 striping on filesystem in ll_dirstripe_verify
869 Description: ASSERTION(CheckWriteback(page,cmd)) failed
870 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
873 Severity : enhancement
875 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
879 Description: do not process already freed flock
880 Details : flock can possibly be freed by another thread before it reaches
881 to ldlm_flock_completion_ast.
885 Description: LBUG during stress test
886 Details : Need properly lock accesses the flock deadlock detection list.
890 Description: oops in page fault handler
891 Details : kernel page fault handler can return two special 'pages' in error
892 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
896 Description: timeout with invalidate import.
897 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
898 should be handled by ptlrpcd. This produce long age waiting and
899 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
901 Severity : enhancement
903 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
907 Description: enable MGS and MDT services start separately
908 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
909 a co-located MGS without starting the MGS, which is a complement
910 to 'nosvc' mount option.
914 Frequency : after recovery
915 Description: precreate to many object's after del orphan.
916 Details : del orphan st in oscc last_id == next_id and this triger growing
917 count of precreated objects. Set flag LOW to skip increase count
918 of precreated objects.
922 Frequency : rare, on clear nid stats
923 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
924 Details : when clean nid stats sometimes try destroy live entry,
925 and this produce panic in free.
929 Description: Stack overflow during MDS log replay
930 ease stack pressure by using a thread dealing llog_process.
934 Description: wait until IO finished before start new when do lock cancel.
935 Details : VM protocol want old IO finished before start new, in this case
936 need wait until PG_writeback is cleared until check dirty flag and
937 call writepages in lock cancel callback.
939 Severity : enhancement
941 Description: using special macro for print time and cleanup in includes.
945 Description: mds_mfd_close() ASSERTION(rc == 0)
946 Details : In mds_mfd_close(), we need protect inode's writecount change
947 within its orphan write semaphore to prevent possible races.
951 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
955 Frequency : rare, on shutdown ost
956 Description: don't hit live lock with umount ost.
957 Details : shrink_dcache_parent can be in long loop with destroy dentries,
958 use shrink_dcache_sb instead.
962 Description: don't panic with use echo client
963 Details : echo client pass NULL as client nid pointer and this produce null
968 Description: fix build on ppc32
969 Details : compile code with -m64 flag produce wrong object file for ppc32.
973 Description: add message levels for liblustreapi
977 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
978 Details : Move the 'good_osts' check before the 'total_bavail' check. This
979 will result in an -EAGAIN and in the exit call path we call
980 alloc_rr() which will with increasing aggressiveness attempt to
981 aquire precreated objects on the minimum number of required OSCs.
985 Description: Use old size assignment to avoid deadlock
986 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
987 the scheduling while holding a spinlock. We do not need locking
988 for size in ll_update_inode() because size is only updated from
989 the MDS for directories or files without objects, so there is no
990 other place to do the update, and concurrent access to such inodes
991 are protected by the inode lock.
995 Description: resolve "_IOWR redefined" build error on SLES10
999 Description: dump the memory debugging after all modules are unloaded to
1000 suppress false negative in conf_sanity test 39
1002 Severity : enhancement
1004 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1009 Description: fix SLES kernel versioning
1010 Details : the kernel version for our SLES 10 kernel did not include a "-"
1011 before the "smp" at the end. while this was not a problem in
1012 general, it did mean that software trying to use the kernel
1013 version to try to detect a vendor specific kernel would fail.
1014 this was most evident by the OFED build scripts.
1018 Description: Don't update lov_desc members until making sure they are valid
1019 Details : When updating lov_desc members via proc fs, need fix their
1020 validities before doing the real update.
1024 Description: don't put request into delay list while invalidate in flight.
1025 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1026 import in flight. this produce timeout for invalidate and sometimes
1027 can cause stale data.
1031 Frequency : on ppc only
1032 Description: not convert ost objects for directory because it's not exist.
1033 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1037 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1039 Severity : enhancement
1041 Description: Update to support 2.6.22.14 vanilla kernel.
1045 Frequency : rare, on recovery
1046 Description: read procfs can produce deadlock in some situation
1047 Details : Holding lprocfs lock which send rpc can produce block for destroy
1048 obd objects and this also block reconnect with -EALREADY. This
1049 isn't fix all lprocfs bugs - but make it rare.
1051 Severity : enhancement
1053 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1056 Frequency : frequent on X2 node
1058 Description: mdc_set_open_replay_data LBUG
1059 Details : Set replay data for requests that are eligible for replay.
1063 Description: lustre_mgs: operation 101 on unconnected MGS
1064 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1065 MGC, and late on MGC cannot successfully connect to MGS and a lot
1066 of the error messages complaining that MGS is not connected.
1070 Frequency : on network error
1071 Description: panic with double free request if network error
1072 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1073 but it's true only for synchronus enqueue, for async enqueue
1074 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1077 Severity : enhancement
1079 Description: client-side metadata stat-ahead during readdir(directory readahead)
1080 Details : perform client-side metadata stat-ahead when the client detects
1081 readdir and sequential stat of dir entries therein
1084 Frequency : on start mds
1086 Description: Implement get_info(last_id) in obdfilter.
1089 Frequency : occasional
1091 Description: Correctly check stale fid, not start epoch if ost not support SOM
1092 Details : open with flag O_CREATE need set old fid in op_fid3 because
1093 op_fid2 was overwritten with new generated fid, but mds can answer
1094 with one of these two fids and both is not stale. Setattr
1095 incorrectly started an epoch and assume will be called
1096 done_writing, but without SOM done_writing ever being called.
1099 Frequency : rare, depends on device drivers and load
1101 Description: MDS or OSS nodes crash due to stack overflow
1102 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1103 In some cases, in conjunction with device drivers that use a lot
1104 of stack the MDS (or possibly OSS) service threads could overflow
1105 the stack. One change which was identified to consume additional
1106 stack has been reworked to avoid the extra stack usage.
1109 Frequency : occasional
1111 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1112 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1113 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1114 oscc_has_objects and see OSC has no objests and call
1115 oscc_internal_create to resend create request.
1117 Severity : enhancement
1119 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1121 Severity : enhancement
1123 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1126 Frequency : very rare
1128 Description: Fix replay if there is an un-replied request and open
1129 Details : In some cases, older replay request will revert the
1130 mcd->mcd_last_xid on MDS which is used to record the client's
1131 latest sent request.
1133 Severity : enhancement
1135 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1137 Severity : enhancement
1139 Description: Add rhel5 support to HEAD.
1141 Serverity : enhancement
1143 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1148 Description: Don't allow skipping OSTs if index has been specified.
1149 Details : Don't allow skipping OSTs if index has been specified, make
1150 locking in internal create lots better.
1154 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1155 Details : access to bitfield in structure is always rounded to long
1156 and this produce problem with not atomic change any bit.
1160 Description: Lustre make rpms failed.
1161 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1162 builds Lustre rpms from tarball.
1165 Frequency : rare on shutdown ost
1167 Description: If llog cancel was not send before clean_exports phase, this can
1168 produce deadlock in llog code.
1169 Details : If llog thread has last reference to obd and call class_import_put
1170 this produce deadlock because llog_cleanup_commit_master wait when
1171 last llog_commit_thread exited, but this never success because was
1172 called from llog_commit_thread.
1176 Description: allow userland application know is lost one of stripes.
1177 Details : fill lvb_blocks with error code on ost and return it to
1178 application if error flag found.
1182 Description: NULL lov_tgts causing MDS oops
1183 Details : more safe checks for NULL lov_tgts for avoid oops.
1185 Severity : enhancement
1187 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1191 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1192 Details : If an OST is down the MDS will hang indefinitely in
1193 obd_statfs_rqset() waiting for the statfs data. While for
1194 MDS QOS usage of statfs, it should not stuck in waiting.
1196 Severity : enhancement
1198 Description: remote_acl support
1199 Details : Support ACL-based permission check for remote user.
1200 Support setfacl/getfacl for remote user with the utils
1201 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1202 the system "{s,g}etfacl" utils.
1204 Severity : enhancement
1206 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1208 Severity : enhancement
1210 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1214 Description: make mgs_setparam() handle fsname containing dash
1215 Details : fsname containing a dash does not work with lctl conf_param
1217 Severity : enhancement
1219 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1221 Severity : enhancement
1223 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1227 Frequency : rare, at shutdown
1228 Description: access already free / zero obd_namespace.
1229 Details : if client_disconnect_export was called without force flag set,
1230 and exist connect request in flight, this can produce access to
1231 NULL pointer (or already free pointer) when connect_interpret
1232 store ocd flags in obd_namespace.
1236 Frequency : only at startup
1237 Description: not alloc memory with spinlock held.
1238 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1239 if any spinlock held.
1241 Severity : enhancement
1243 Description: make lustre randomly fail allocating memory
1244 Details : Make lustre randomly failed allocating memory for testing purpose.
1246 Severity : enhancement
1248 Description: lost problems with lov objid file
1249 Details : Fixes some scability and access to not inited memory problems
1250 in work with lov objdid file.
1255 Description: lfs find does not continue on file error
1256 Details : Continue other files processing when a file/dir is absent.
1260 Description: Inconsistent usage of lustre_pack_reply()
1261 Details : Standardize the usage of lustre_pack_reply() such that it
1262 always generate a CERROR on failure.
1265 Frequency : occasional
1267 Description: MDS hang or stay in waiting lock
1268 Details : If client receive lock with CBPENDING flag ldlm need send lock
1269 cancel as separate rpc, to avoid situation when cancel request
1270 can't processed due all i/o threads stay in wait lock.
1274 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1278 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1282 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1286 Description: Update to RHEL4 latest kernel.
1288 Severity : enhancement
1290 Description: Build SLES10 patchless client fails
1291 Details : The configure was broken by run ./configure with
1292 --with-linux-obj=.... argument for patchless client. When the
1293 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1294 can't search header adequately. Use absolute path such as
1295 -I($LINUX)/include instead.
1299 Description: interrupt oig_wait produce painc on resend.
1300 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1301 private set, and this produce situation when rq_ptlrpcd_data not
1302 copyed to new allocated request and triggered LBUG on assert
1303 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1304 ptlrpcd set if request is changed and can be safety changed to use
1309 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1310 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1311 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1312 these macros and replace them with their difinitions instead.
1316 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1317 Details : Update which_patch & target file for SLES10 latest kernel.
1320 Description: Remove client patches from SLES 10 kernel.
1321 Details : This causes SLES 10 clients to behave as patchless clients
1322 even on a Lustre-patched (server) kernel.
1324 Severity : enhancement
1326 Description: self-adjustable client's lru lists
1327 Details : use adaptive algorithm for managing client cached locks lru
1328 lists according to current server load, other client's work
1329 pattern, memory activities, etc. Both, server and client
1330 side namespaces provide number of proc tunables for controlling
1333 Severity : enhancement
1335 Description: light-weight GSS support
1336 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1337 and reduce performance overhead.
1339 Severity : enhancement
1341 Description: Linux keyring support
1342 Details : Support using service of Linux keyring for Lustre GSS internal
1343 context refresh/cache mechanism.
1347 Description: Fix errors in lfs documentation
1348 Details : Fixes man pages
1352 Description: don't use GFP_* in generic Lustre code.
1353 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1357 Description: obdclass is limited by single OBD_ALLOC(idarray)
1358 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1362 Description: warnings with build patchless client with vanila 2.6.19 and up
1363 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1364 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1368 Frequency : Only for SLES
1369 Description: sanity_quota fail test_1
1370 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1371 file, which makes TSTID[2] inunique.
1375 Frequency : Only for SLES9
1376 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1377 Details : Update target/ChangeLog/which_patch .
1381 Frequency : Only for test_10 in sanity-quota.sh
1382 Description: a bug in quota test code
1383 Details : checking a test flag in a wrong place
1387 Frequency : Only for test_13 in sanity-quota.sh
1388 Description: a bug in quota test script
1389 Details : shouldn't check overfull allocated quota
1393 Description: jbd statistics
1394 Details : Port older jbd statistics patch for sles10
1396 Severity : enhancement
1398 Description: Add printing inode info into message about error in writepage.
1403 Description: reply_lock_interpret crash due to race with it and lock cancel.
1404 Details : Do not replay locks that are being cancelled. Do not reference
1405 locks by their address during replay, just by their handle.
1409 Frequency : When flocks are used.
1410 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1411 Details : Flock locks might destroy just granted lock if it could be merged
1412 with another existing flock, this is done in completion handler,
1413 so teach ldlm_cli_enquque_fini that this is a valid case for
1419 Description: Oops in read and write path when failing to allocate lock.
1420 Details : Check if lock allocation failed and return error back.
1424 Description: lstripe command fails for valid OST index
1425 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1426 instead of lov->desc.ld_active_tgt_count.
1430 Description: sanity.sh failed test 103
1431 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1436 Description: df doesn't work properly if diskfs blocksize != 4K
1437 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1441 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1442 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1444 Severity : enhancement
1446 Description: Adaptive timeouts
1447 Details : RPC timeouts adapt to changing server load and network
1448 conditions to reduce resend attempts and improve recovery time.
1452 Description: llapi_file_create() does not allow some changes
1453 Details : add llapi_file_open() that allows specifying the mode and
1454 open flags, and also returns an open file handle.
1458 Description: merge and cleanup kernel patches.
1459 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1463 Description: Add journal checksum support.(Kernel part)
1464 Details : The journal checksum feature adds two new flags i.e
1465 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1466 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1467 indicates that the commit block contains the checksum for
1468 the blocks described by the descriptor blocks. Now commit
1469 record can be sent to disk without waiting for descriptor
1470 blocks to be written to disk. This behavior is controlled
1471 using JBD2_FEATURE_ASYNC_COMMIT flag.
1475 Description: OSS needs mutliple precreate threads
1476 Details : Add ability to start more than one create thread per OSS.
1480 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1481 Details : execute lfs setstripe on client
1485 Description: mds_obd_create error creating tmp object
1486 Details : When the user sets quota on root, llog will be affected and can't
1487 create files and write files.
1490 Frequency : Always on ia64 patchless client, and possibly others.
1492 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1493 Details : This allows the patchless client to be loaded on architectures
1494 without this export.
1498 Description: RedHat Update kernel for RHEL5
1499 Details : Add the kernel config file for RHEL5.
1503 Description: RedHat Update kernel for RHEL5
1504 Details : Modify the kernel config file more closer RHEL5.
1508 Description: Build failure against Centos5 (RHEL5)
1509 Details : Define PAGE_SIZE when it isn't present.
1513 Description: test_7 on sanity-quota.sh can't be used on separate machines
1514 Details : add facet to handle it
1518 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1519 Details : with NFS, the anon dentry's parent was set to itself in
1520 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1521 corresponding dentry other than use rec->ur_name.
1523 Severity : enhancement
1525 Description: lfs setstripe enhancement
1526 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1530 Description: enable data checksumming by default
1531 Details : enable checksum by default, allow --disable-checksum
1532 configure option and "-o nochecksum" mount option. Checksums
1533 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1537 Description: lustre support for RHEL5
1538 Details : Add support for RHEL5.
1542 Description: Client eviction due to failover config
1543 Details : after a connection loss, the lustre client should attempt to
1544 reconnect to the last active server first before trying the
1545 other potential connections.
1549 Description: when mds and osts use different quota unit(32bit and 64bit),
1550 quota will be released repeatly.
1551 Details : void sending multiple quota reqs to mds, which will keep the
1552 status between the reqs.
1556 Description: osts not allocated evenly to files
1557 Details : change the condition to increase offset_idx
1560 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1562 Description: Data corruption for OSTs that are formatted larger than 2TB
1564 Details : When generating the bio request for lustre file writes the
1565 sector number would overflow a temporary variable before being
1566 used for the IO. The data reads correctly from Lustre (which
1567 will overflow in a similar manner) but other file data or
1568 filesystem metadata may be corrupted in some cases.
1572 Description: Tune the kernel for good SCSI performance.
1573 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1574 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1579 Description: rewrite ext2-derived code in obdclass/uuid.c
1580 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1581 from scratch preserving functionality.
1585 Description: "lfs find -obd UUID" prints directories
1586 Details : "lfs find -obd UUID" will return all directory names instead
1587 of just file names. It is incorrect because the directories
1588 do not reside on the OSTs.
1592 Description: use i_size_read and i_size_write in 2.6 port
1593 Details : replace inode->i_size access with i_size_read/write()
1596 Frequency : when removing large files
1598 Description: scheduling issue during removal of large Lustre files
1599 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1600 It causes scheduling issues when removing large files (17TB in the
1604 Frequency : only with liblustre clients on XT3
1606 Description: evictions taking too long
1607 Details : allow llrd to evict clients directly on OSTs
1610 Frequency : only on ppc
1612 Description: /proc/fs/lustre/devices broken on ppc
1613 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1614 We should make sure the type of 'index' is loff_t and then cast
1615 explicitly as needed below. Do not assign an explicitly cast
1619 Frequency : only for rhel5
1621 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1622 Details : Modify the target file & which_kernel.
1624 Severity : enhancement
1626 Description: omit set fsid for export NFS
1627 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1628 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1632 Frequency : Always, for kernels after 2.6.16
1633 Description: Fix warning idr_remove called for id=.. which is not allocated.
1634 Details : Last kernels save old s_dev before kill super and not allow
1635 to restore from callback - restore it before call kill_anon_super.
1639 Description: buffer overruns could theoretically occur
1640 Details : llapi_semantic_traverse() modifies the "path" argument by
1641 appending values to the end of the origin string, and a
1642 overrun may occur. Adding buffer overrun check in liblustreapi.
1646 Description: Fix error on 'ls .' at the top of the Lustre mount.
1647 Details : Don't revalidate dentry if it is a root dentry.
1651 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1652 Details : Modify vm-tunables-rhel4.patch.
1656 Description: Kernel config for 2.6.18-vanilla.
1657 Details : Modify targets/2.6-vanilla.target.in.
1658 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1659 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1660 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1661 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1665 Description: improve handling recoverable errors
1666 Details : if request processig with error which can be recoverable on server
1667 request should be resend, otherwise page released from cache and
1672 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1673 Details : Modify target file & which_patch.
1674 A flaw was found in the IA32 system call emulation provided
1675 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1676 value could be stored in the %RAX register, which could trigger an
1677 out-of-bounds system call table access. An untrusted local user
1678 could exploit this flaw to run code in the kernel
1679 (ie a root privilege escalation). (CVE-2007-4573).
1683 Description: O_DIRECT bypasses client statistics.
1684 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1685 read_ahead_stats were updated. copied the stats section from
1686 osc_send_oap_rpc() into async_internal().
1690 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1694 Description: change order of libsysio includes
1695 Details : '#include sysio.h' should always come before '#include xtio.h'
1699 Description: handle "serious error: objid * already exists" more gracefully
1700 Details : If LAST_ID value on disk is smaller than the objects existing in
1701 the O/0/d* directories, it indicates disk corruption and causes an
1702 LBUG(). If the object is 0-length, then we should use the existing
1703 object. This will help to avoid a full fsck in most cases.
1705 Severity : enhancement
1707 Description: adapt the lustre_config script to support the upgrade case
1708 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1709 server targets to 1.6 in parallel.
1713 Description: To avoid grant space > avaible space when the disk is almost
1714 full. Without this patch you might see the error "grant XXXX >
1715 available" or some LBUG about grant, when the disk is almost
1717 Details : In filter_check_grant, for non_grant cache write, we should
1718 check the left space by if (*left > ungranted + bytes), instead
1719 of (*left > ungranted), because only we are sure the left space
1720 is enough for another "bytes", then the ungrant space should be
1721 increase. In client, we should update cl_avail_grant only there
1722 is OBD_MD_FLGRANT in the reply.
1726 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1727 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1728 described in CVE-2007-4573. This problem could allow untrusted
1729 local users to gain root access.
1732 Frequency : when using O_DIRECT and quotas
1734 Description: Incorrect file ownership on O_DIRECT output files
1735 Details : block usage reported by 'lfs quota' does not take into account
1736 files that have been written with O_DIRECT.
1741 Description: touch file failed when fs is not full
1742 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1743 otherwise we can get ENOSP while fs is not full.
1747 Description: parallel lock callbacks
1748 Details : Instead of sending blocking and completion callbacks as separated
1749 requests, adding them to a set and sending in parallel.
1752 Frequency : only for Cray XT3
1753 Bugzilla : 12829/13455
1754 Description: Changing primary group doesn't change the group lustre assigns to
1756 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1757 overridden with the primary group provided by the group upcall,
1758 whereas the supplied fsgid can be trusted if it is in the list of
1759 supplementary groups returned by the group upcall.
1761 Severity : enhancement
1763 Description: Allow masking D_WARNING, D_ERROR messages from console
1764 Details : Console messages can now be disabled via lnet.printk.
1768 Description: User code with malformed file open parameter crashes client node
1769 Details : Before packing join_file req, all the related reference should be
1770 checked carefully in case some malformed flags cause fake
1771 join_file req on client.
1775 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1776 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1777 mode is equal to the granted mode. However, it should also check
1778 that LDLM_FL_CP_REQD is not set.
1782 Description: Heavy nfs access might result in deadlocks
1783 Details : After ELC code landed, it is now improper to enqueue any mds
1784 locks under och_sem, because enqueue might want to decide to
1785 cancel open locks for same inode we are holding och_sem for.
1789 Description: Client eviction while running blogbench
1790 Details : A lot of unlink operations with concurrent I/O can lead to a
1791 deadlock causing evictions. To address the problem, the number of
1792 oustanding OST_DESTROY requests is now throttled to
1793 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1794 callbacks are processed in priority.
1798 Description: enable ACLs on MDS by default
1799 Details : ACLs must be enabled on MDS by default.
1802 Frequency : PPC/PPC64 only
1804 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1805 Details : fix duplicated definitions between asm-ppc64/types.h and
1806 lustre_types.h on PPC.
1809 Frequency : PPC/PPC64 only
1811 Description: asm-ppc/segment.h does not exist
1812 Details : fix compile issue on PPC.
1816 Description: better handle error messages in extents code
1819 Frequency : RHEL4 only
1821 Description: mkfs is very slow on IA64/RHEL4
1822 Details : A performance regression has been discovered in the MPT Fusion
1823 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1824 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1825 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1827 Severity : enhancement
1829 Description: SNMP support enhancement
1830 Details : Adding total number of sampled request for an MDS node in snmp
1833 Severity : enhancement
1835 Description: Optimize ldlm waiting list processing for PR extent locks
1836 Details : When processing waiting list for read extent lock and meeting read
1837 lock that is same or wider to it that is not contended, skip
1838 processing rest of the list and immediatelly return current
1839 status of conflictness, since we are guaranteed there are no
1840 conflicting locks in the rest of the list.
1844 Description: Time out and refuse to reconnect
1845 Details : When the failover node is the primary node, it is possible
1846 to have two identical connections in imp_conn_list. We must
1847 compare not conn's pointers but NIDs, otherwise we can defeat
1848 connection throttling.
1852 Description: port llog fixes from b1_6 into HEAD
1853 Details : Port llog reference couting and some llog cleanups from b1_6
1854 (bug 10800) into HEAD, for protect from panic and access to already
1855 free llog structures.
1859 Description: Detect stride IO mode in read-ahead
1860 Details : When a client does stride read, read-ahead should detect that and
1861 read-ahead pages according to the detected stride pattern.
1865 Description: data checksumming impacts single node performance
1866 Details : add support for several checksum algorithm. Currently, only CRC32
1867 and Adler-32 are supported. The checksum type can be changed on
1868 the fly via /proc/fs/lustre/osc/*/checksum_type.
1872 Description: use adler32 for page checksums
1873 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1878 Description: build for x2 fails
1879 Details : fix compile issue on Cray systems.
1883 Description: Properly match for duplicate locks
1884 Details : Due to different lock order from skiplists code, we need to
1885 traverse entire list for now
1888 Frequency : only on PPC/SLES10
1890 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1891 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1892 be defined. Add a hack in mkfs_lustre.c to work around this compile
1897 Description: LASSERT on MDS when client holding flock lock dies
1898 Details : ldlm pool logic depends on number of granted locks equal to
1899 number of released locks which is not true for flock locks, so
1900 just exclude such locks from consideration.
1904 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1905 Details : Use fsfilt_sync() for both the whole filesystem sync and
1906 individual file sync to eliminate dangerous inode locking
1907 with I_LOCK that can lead to a deadlock.
1911 Description: performance in 1.6.3
1912 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1913 sure that 1MB requests are not fragmented by the block layer.
1915 Severity : enhancement
1917 Description: organize the server-side client stats on per-nid basis
1918 Details : Change the structure of stats under obdfilter and mds to
1926 The "uuid"s file would list the uuids of _active_ exports.
1927 And the clear entry is to clear all stats and stale nids.
1929 Severity : enhancement
1931 Description: eliminate client locks in face of contention
1932 Details : file contention detection and lockless i/o implementation
1933 for contended files.
1937 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1940 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1942 Description: MDS may lose file striping (and hence file data) in some cases
1943 Details : If there are additional extended attributes stored on the MDS,
1944 in particular ACLs, SELinux, or user attributes (if user_xattr
1945 is specified for the client mount options) then there is a risk
1946 of attribute loss. Additionally, the Lustre file striping
1947 needs to be larger than default (e.g. striped over all OSTs),
1948 and an additional attribute must be stored initially in the
1949 inode and then increase in size enough to be moved to the
1950 external attribute block (e.g. ACL growing in size) for file
1955 Description: skiplist implementation simplification
1956 Details : skiplists are used to group compatible locks on granted list
1957 that was implemented as tracking first and last lock of each
1958 lock group the patch changes that to using doubly linked lists
1962 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1963 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1964 because LASSERT on both the data supplied by a client, and the
1965 data on disk is dangerous and incorrect.
1967 Severity : enhancement
1969 Description: Slow truncate/writes to huge files at high offsets.
1970 Details : Directly associate cached pages to lock that protect those pages,
1971 this allows us to quickly find what pages to write and remove
1972 once lock callback is received.
1976 Description: more ldlm soft lockups
1977 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1978 starve other threads from the resource lock for a long time in
1979 case of long waiting queue, so change the debug level from
1980 D_OTHER to the less frequently used D_INFO.
1982 Severity : enhancement
1984 Description: add -gid, -group, -uid, -user options to lfs find
1988 Description: Hung threads in invalidate_inode_pages2_range
1989 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1990 once one is completed. As a result, some RPCs are stuck in the
1991 queue and are never sent.
1995 Description: filter threads hungs on waiting journal commit
1996 Details : Cleanup filter group llog code, then only filter group llog will
1997 be only created in the MDS/OST syncing process.
2001 Description: Procfs and llog threads access destoryed import sometimes.
2002 Details : Sync the import destoryed process with procfs and llog threads by
2003 the import refcount and semaphore.
2005 Severity : enhancement
2007 Description: openlock cache of b1_6 port to HEAD
2012 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2013 Details : If insertion of an extent fails, then discard the inode
2014 preallocation and free data blocks else it can lead to duplicate
2019 Description: don't always update ctime in ext3_xattr_set_handle()
2020 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2021 In some cases the ctime should not be updated, for example for
2022 2.0->1.8 compatibility it is necessary to delete an xattr and it
2023 should not update the ctime.
2027 Bugzilla : 15713/16362
2028 Description: Assertion in iopen_connect_dentry in 1.6.3
2029 Details : looking up an inode via iopen with the wrong generation number can
2030 populate the dcache with a disconneced dentry while the inode
2031 number is in the process of being reallocated. This causes an
2032 assertion failure in iopen since the inode's dentry list contains
2033 both a connected and disconnected dentry.
2037 Description: assertion failure in ldlm_handle2lock()
2038 Details : fix a race between class_handle_unhash() and class_handle2object()
2039 introduced in lustre 1.6.5 by bug 13622.
2044 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2045 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2050 Description: oops in sd_iostats_seq_show()
2051 Details : unloading/reloading the scsi low level driver triggers a kernel
2052 bug when trying to access the sd iostat file.
2057 Description: Kernel panics during QLogic driver reload
2058 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2059 patch, causing memory corruption.
2064 Description: journal_dev option does not work in b1_6
2065 Details : pass mount option during pre-mount.
2067 Severity : enhancement
2069 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2070 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2071 extent information of a file. It can be used to map logical blocks
2072 in a file to physical blocks in the block device.
2076 Description: LDLM soft lockups - improvement
2077 Details : It is be possible to send the lock handle along with each read
2078 or write request because the client is already doing a lock match
2079 itself so there isn't any reason the OST should have to re-do that
2083 Frequency : only on Cray X2
2085 Description: X2 build failures
2086 Details : fix build failures on Cray X2.
2090 Description: xid & resent requests
2091 Details : Initialize RPC XID from clock at startup (randomly if clock is
2094 Severity : enhancement
2096 Description: Add lustre_start utility to start or stop multiple Lustre servers
2101 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2102 Details : In case of memory pressure, list_del() can be called twice on
2103 req->rq_history_list, causing a kernel oops.
2107 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2108 Details : fix stack overflow in the distributed lock manager by defering
2109 export eviction after a failed AST to the elt thread instead of
2110 handling it in the dlm interpret routine.
2114 Description: Convert some comments to new format.
2115 Details : Update documenting comments to match doxygen conventions.
2119 Description: Grammar fixes.
2120 Details : A couple of trivial spelling fixes.
2124 Description: OSD_COUNTERS-mandatory
2125 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2126 to simplify the code.
2130 Description: simplify cmm_device freeing logic.
2131 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2135 Description: Add lockdep support to dt_object_operations locking interface.
2136 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2137 parameter indicating lock ordering. Update mdd code to use new
2142 Description: Introduce failloc constants for lockless IO tests.
2143 Details : Add two new failloc constants to test lockless IO. Only one of
2144 them in implemented---another is checked in yet to be landed
2149 Description: Add lockdep support for inode mutex.
2150 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2151 in the situations where Lustre has to lock more than one inode
2156 Description: Add optional invariants checking support.
2157 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2158 configure switch. This macro is to be used for consistency and
2159 sanity checks that are too expensive to be left in `production'
2164 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2165 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2170 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2171 Details : Use spin_lock_nested() in (the only) situation where more than
2172 one ldlm_lock is locked simultaneously. Also, fix possible
2173 dead-lock in ldlm_lock_change_resource() by enforcing particular
2178 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2179 Details : Instead of specifying each ldlm_lock call-back through separate
2180 parameter, wrap them into struct ldlm_callback_suite.
2184 Description: Kill join_lru obd method and its callers.
2185 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2186 in memory---a requirement implemented through ->o_join_lru() obd
2187 method in HEAD. Remove this method and its users.
2191 Description: Add asynchronous ldlm ENQUEUE completion handler.
2192 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2193 case that stock ldlm_completion_ast() cannot handle as it waits
2194 until lock is granted. Introduce new ldlm_completion_ast_async()
2195 for this. Also comment ldlm_completion_ast().
2199 Description: ldlm_error <-> errno conversion.
2200 Details : Add functions to map (rather arbitrary) between LDLM error codes
2201 and standard errno values. CLIO needs this to prevent LDLM specific
2202 constants from escaping ldlm and osc.
2206 Description: Kill unused ldlm_handle2lock_ns() function.
2207 Details : Kill unused ldlm_handle2lock_ns() function.
2211 Description: Add lu_ref support to ldlm_lock
2212 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2213 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2214 and ldlm_resource. LDLM interface has to be changed, because code that
2215 releases a reference on a lock, has to "know" what reference this is.
2216 In the most frequent case
2218 lock = ldlm_handle2lock(handle);
2220 LDLM_LOCK_PUT(lock);
2222 no changes are required. When any other reference (received _not_ from
2223 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2224 instead of LDLM_LOCK_PUT().
2226 Arguably, changes are pervasive, and interface requires some discipline
2227 for proper use. On the other hand, it was very instrumental in finding
2228 a few leaked lock references.
2232 Description: Add ldlm_lock_addref_try().
2233 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2234 attempts to addref a lock that might be being canceled
2239 Description: Add ldlm_weigh_callback().
2240 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2241 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2242 hard-coded `number of pages' logic.
2246 Description: Add lockdep annotations to llog code.
2247 Details : Use appropriately tagged _nested() locking calls in the places
2248 where llog takes more than one ->lgh_lock lock.
2252 Description: Add loi_kms_set().
2253 Details : Wrap kms updates into a helper function.
2257 Description: Constify instances of struct lsm_operations.
2258 Details : Constify instances of struct lsm_operations.
2262 Description: lu_conf support.
2263 Details : On a server, a file system object is uniquely identified
2264 by a fid, which is sufficient to locate and load all object
2265 state (inode). On a client, on the other hand, more data are
2266 necessary instantiate an object. Change lu_object_find() and
2267 friends to take additional `lu_conf' argument describing object.
2268 Typically this includes layout information.
2272 Description: lu_context fixes.
2273 Details : Introduce new lu_context functions that are needed on the client
2274 side, where some system threads (ptlrpcd) are shared by multiple
2275 modules, and so cannot be stopped during module shutdown.
2279 Description: Add start and stop methods to lu_device_type_operations.
2280 Details : Introduce two new methods in lu_device_type_operations, that are
2281 invoked when first instance of a given type is created and last one
2282 is destroyed respectively. This is need by CLIO.
2286 Description: Add lu_ref support to struct lu_device.
2287 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2288 track leaked references.
2292 Description: Introduce lu_kmem_descr.
2293 Details : lu_kmem_descr and its companion interface allow to create
2294 and destroy a number of kmem caches at once.
2298 Description: Fix lu_object finalization race.
2299 Details : Fix a race between lu_object_find() finding an object and its
2300 concurrent finalization. This race is (most likely) not possible
2301 on the server, but might happen on the client.
2305 Description: Introduce lu_ref interface.
2306 Details : lu_ref is a debugging module allowing to track references to
2307 a given object. It is quite cpu expensive, and has to be
2308 explicitly enabled with --enable-lu_ref. See usage description
2313 Description: Factor lu_site procfs stats into a separate function.
2314 Details : Separate lu_site stats printing code into a separate function
2315 that can be reused on a client.
2319 Description: Constify instances of struct {lu,dt,md}_device_operations.
2320 Details : Constify instances of struct {lu,dt,md}_device_operations.
2324 Description: Introduce struct md_site and move meta-data specific parts of
2325 struct lu_site here.
2326 Details : Move md-specific fields out of struct lu_site into special struct
2327 md_site, so that lu_site can be used on a client.
2331 Description: Kill mdd_lov_destroy().
2332 Details : Remove unused mdd code.
2336 Description: Add st_block checking to multistat.c.
2337 Details : Add st_block checking to multistat.c.
2341 Description: Add lu_ref support to struct obd_device.
2342 Details : Add lu_ref tracking to obd_device.
2346 Description: Kill obd_set_fail_loc().
2347 Details : Remove unused code.
2351 Description: Add special type for ptlrpc_request interpret functions.
2352 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2353 there. Actual usage will be in CLIO.
2357 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2358 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2359 doesn't work with lockdep.
2363 Description: Add rwv.c test program.
2364 Details : New testing program exercising readv(2) and writev(2) (Qian).
2368 Description: Add sendfile.c test program.
2369 Details : New testing program exercising sendfile(2) (Jay).
2373 Description: Ratelimit a message that can be very frequent.
2374 Details : Ratelimit a memory allocation failure message that can
2379 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2380 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2384 Description: Liblustre build fixes.
2385 Details : Liblustre build fixes.
2389 Description: libcfs: add cfs_{need,cond}_resched() interface.
2390 Details : libcfs: add cfs_{need,cond}_resched() definition and
2391 implementations for Linux, NT, and liblustre.
2393 Severity : enhancement
2395 Description: More exported tunables for mballoc
2396 Details : Add support for tunable preallocation window and new tunables for
2397 large/small requests
2401 Description: Detect corruption of block bitmap and checking for preallocations
2402 Details : Checks validity of on-disk block bitmap. Also it does better
2403 checking of number of applied preallocations. When corruption is
2404 found, it turns filesystem readonly to prevent further corruptions.
2408 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2409 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2410 window to 0. If it does hit cache miss, and read-pattern is still
2411 stride-io mode, does not reset the stride window, but also does not
2412 increase the stride window length in this case.
2416 Frequency : only for big-endian servers
2417 Description: Check if system is big-endian while mounting fs with extents feature
2418 Details : Mounting a filesystem with extents feature will fail on big-endian
2419 systems since ext3-based ldiskfs is not supported on big-endian
2420 systems. This can be over-riden with "bigendian_extents" mount option.
2422 Severity : enhancement
2424 Description: The root squash functionality
2425 Details : A security feature, which is to prevent users from being able
2426 to mount lustre on their desktop, run as root, and delete
2427 all of the files in the filesystem. The goal is accomplished by
2428 remapping user id (UID) and group id (GID) of the root user to
2429 a UID and GID specified by the system administartor via Lustre
2430 configuration management server (MGS). The functionality also
2431 allows to specify sets of clients for which the remapping does
2436 Description: Excessive recovery window
2437 Details : With AT enabled, the recovery window can be excessively long (6000+
2438 seconds). To address this problem, we no longer use
2439 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2440 timeout no longer depends on the service time, it is set to
2441 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2442 time via pb_service_time.
2446 Description: Watchdog triggered on MDS failover
2447 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2448 the OSTs know that the MDS "UUID" can be reused for the same export
2449 from a different NID, so we do not need to wait for the export to be
2453 Frequency : rare, only if using MMP with Linux RAID
2455 Description: MMP doesn't work with Linux RAID
2456 Details : While using HA for Lustre servers with Linux RAID, it is possible
2457 that MMP will not detect multiple mounts. To make this work we
2458 need to unplug the device queue in RAID when the MMP block is being
2459 written. Also while reading the MMP block, we should read it from
2460 disk and not the cached one.
2462 Severity : enhancement
2464 Description: open file using fid
2465 Details : A file can be opened using just its fid, like
2466 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2469 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2472 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2473 Details : add proper locking for creation of jbd_4k slab cache
2477 Description: MMP check in ext3_remount() fails without displaying any error
2478 Details : When multiple mount protection fails during remount, proper error
2481 Severity : enhancement
2483 Description: Allow stripe size to be up to 4G-64k
2484 Details : Fix math logic to allow large stripe sizes.
2488 Description: add check for >8TB ldiskfs filesystems
2489 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2490 Don't allow >8TB ldiskfs filesystems to be mounted without
2491 force_over_8tb mount option
2496 Descriptoin: OSS complains "too many blocks" and crashes
2497 Details : In filter_commitrw_write():
2498 1. lquota_chkquota();
2499 2. obdo_from_inode();
2500 3. lquota_pending_commit();
2501 "2." has changed uid and gid in this case, then caused
2502 lquota_chkquota() and lquota_pending_commit() to use
2503 different uid and gid. That is the root of the bug.
2507 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2508 Details : Detect and print out the 'stripe count = -1' case properly.
2512 Description: Out or order replies might be lost on replay
2513 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2514 request with tid smaller then one currently being added, add it
2515 to the start, not end of the list.
2517 --------------------------------------------------------------------------------
2519 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2521 * Support for kernels:
2522 2.6.5-7.283 (SLES 9),
2523 2.6.9-55.EL (RHEL 4),
2524 2.6.16.46-0.14 (SLES 10),
2525 2.6.18.8 vanilla (kernel.org)
2526 * Client support for unpatched kernels:
2527 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2528 2.6.16 - 2.6.22 vanilla (kernel.org)
2529 * Due to recently discovered recovery problems, we do not recommend
2530 using patchless RHEL 4 clients with this or any earlier release.
2531 * Recommended e2fsprogs version: 1.39.cfs8
2532 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2533 * Starting with this release, the ldiskfs backing filesystem required
2534 by Lustre is now in its own package, lustre-ldiskfs. This package
2535 should be installed. It is versioned separately from Lustre and
2536 may be released separately in future.
2540 Description: block reactivating mgc import until all deactivates complete
2541 Details : Fix race when failing back MDT/MGS to itself (testing)
2544 Frequency : at statup only
2546 Description: mds_lov_synchronize race leads to various problems
2547 Details : simultaneous MDT->OST connections at startup can cause the
2548 sync to abort, leaving the OSC in a bad state.
2550 Severity : enhancement
2552 Description: add optional extra BUILD_VERSION info
2553 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2554 to override the lustre version.
2557 Frequency : 2.6.18 servers only
2559 Description: ll_kern_mount() doesn't release the module reference
2560 Details : The ldiskfs module reference count never drops down to 0
2561 because ll_kern_mount() doesn't release the module reference.
2566 Description: server LBUG when using old ost_num_threads parameter
2567 Details : Accept the old ost_num_threads parameter but warn that it
2568 is deprecated, and fix an off-by-one error that caused an LBUG.
2573 Description: Transient SCSI error results in persistent IO issue
2574 Details : iobuf->dr_error is not reinitialized to 0 between two
2578 Frequency : sometimes when underlying device returns I/O errors
2580 Description: OSTs not going read-only during write failures
2581 Details : OSTs are not remounted read-only when the journal commit threads
2582 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2583 instead of the ext3 wrappers.
2587 Description: poor connect scaling with increasing client count
2588 Details : Don't run filter_grant_sanity_check for more than 100 exports
2589 to improve scaling for large numbers of clients.
2592 Frequency : SLES10 only
2594 Description: sanity-quota.sh quotacheck failed: rc = -22
2595 Details : Quotas cannot be enabled on SLES10.
2598 Frequency : liblustre clients only
2600 Description: getdirentries does not give error when run on compute nodes
2601 Details : getdirentries does not fail when the size specified as an argument
2602 is too small to contain at least one entry
2604 Severity : enhancement
2606 Description: Add LNET router traceability for debug purposes
2607 Details : If a checksum failure occurs with a router as part of the
2608 IO path, the NID of the last router that forwarded the bulk data
2609 is printed so it can be identified.
2614 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2615 Details : Due to a race condition, liblustre clients were occasionally
2616 evicted incorrectly.
2618 Severity : enhancement
2620 Description: lfs setstripe use optional parameters instead of postional
2623 Severity : enhancement
2625 Description: Nanosecond timestamp support for ldiskfs
2626 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2627 resolution timestamps. There is not yet support for this at
2628 the Lustre filesystem level.
2631 Frequency : during server recovery
2633 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2634 Details : request with rq_no_resend flag not awake l_wait_event if they get
2638 Frequency : nfs export on patchless client
2640 Description: connectathon hang when test nfs export over patchless client
2641 Details : Disconnected dentry cannot be found with lookup, so we do not need
2642 to unhash it or make it invalid
2645 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2646 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2647 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2652 Description: open req refcounting wrong on reconnect
2653 Details : If reconnect happened between getting open reply from server and
2654 call to mdc_set_replay_data in ll_file_open, we will schedule
2655 replay for unreferenced request that we are about to free.
2656 Subsequent close will crash in variety of ways.
2657 Check that request is still eligible for replay in
2658 mdc_set_replay_data().
2663 Description: disable writes to filesystem when reading health_check file
2664 Details : the default for reading the health_check proc file has changed
2665 to NOT do a journal transaction and write to disk, because this
2666 can cause reads of the /proc file to hang and block HA state
2667 checking on a healthy but otherwise heavily loaded system. It
2668 is possible to return to the previous behaviour during configure
2669 with --enable-health-write.
2671 Severity : enhancement
2673 Description: 64-bit inode version
2674 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2675 to the inode. This will be required for version-based recovery.
2680 Description: MDS fails to start if a duplicate client export is detected
2681 Details : in some rare cases it was possible for a client to connect to
2682 an MDS multiple times. Upon recovery the MDS would detect this
2683 and fail during startup. Handle this more gracefully.
2685 Severity : enhancement
2687 Description: Add -o localflock option to simulate old noflock
2689 Details : This will achieve local-only flock/fcntl locks
2695 Description: log_commit_thread vs filter_destroy race leads to crash
2696 Details : Take import reference before releasing llog record semaphore
2701 Description: Wrong request locking in request set processing
2702 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2703 imp_delayed_list, in this place should be used imp_lock.
2706 Frequency : when reconnection
2708 Description: Grant Leak when osc reconnect to OST
2709 Details : When osc reconnect ost, OST(filter) should check whether it
2710 should grant more space to client by comparing fed_grant and
2711 cl_avail_grant, and return the granted space to client instead
2712 of "new granted" space, because client will call osc_init_grant
2713 to update the client grant space info.
2716 Frequency : when client reconnect to OST
2718 Description: Grant Leak when osc do resend and replay bulk write
2719 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2720 bulk write request, because the grant info while be sync between
2721 OSC and OST when reconnect, and we should ignore the grant info
2722 these of resend/replay write req.
2727 Description: Grant space more than avaiable left space sometimes.
2728 Details : When then OST is about to be full, if two bulk writing from
2729 different clients came to OST. Accord the avaliable space of the
2730 OST, the first req should be permitted, and the second one
2731 should be denied by ENOSPC. But if the seconde arrived before
2732 the first one is commited. The OST might wrongly permit second
2733 writing, which will cause grant space > avaiable space.
2736 Frequency : when client is evicted
2738 Description: Grant might be wrongly erased when osc is evicted by OST
2739 Details : when the import is evicted by server, it will fork another
2740 thread ptlrpc_invalidate_import_thread to invalidate the
2741 import, where the grant will be set to 0. While the original
2742 thread will update the grant it got when connecting. So if
2743 the former happened latter, the grant will be wrongly errased
2744 because of this race.
2749 Description: Checking Stale with correct fid
2750 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2751 to check whether it is stale, because sometimes, we want the
2752 enqueue happened anyway, and op_data.fid2 will not be initialized.
2754 Severity : enhancement
2756 Description: update patchless client
2757 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2760 Frequency : only with 2.4 kernel
2762 Description: random memory corruption
2763 Details : size of struct ll_inode_info is to big for union inode.u and this
2764 can be cause of random memory corruption.
2769 Description: Memory leak in recovery
2770 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2771 It should also check obd_fail before fsfilt_start, otherwise if
2772 fsfilt_start return -EROFS,(failover mds during mds recovery).
2773 then the req will return with repmsg->transno = 0 and rc = EROFS.
2774 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2775 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2776 be freed no matter whether fsfilt_commit success or not.
2779 Frequency : only with huge count clients
2781 Description: Prevents from taking the superblock lock in llap_from_page for
2783 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2784 for llap_from_page call in ll_removepage prevents from taking the
2785 superblock lock for a soon died page.
2790 Description: Not check open intent error before release open handle
2791 Details : in some rare cases, the open intent error is not checked before
2792 release open handle, which may cause
2793 ASSERTION(open_req->rq_transno != 0), because it tries to release
2794 the failed open handle.
2799 Description: Set cat log bitmap only after create log success.
2800 Details : in some rare cases, the cat log bitmap is set too early. and it
2801 should be set only after create log success.
2805 Description: Accessing a block bevice can re-enable I/O when Lustre is
2806 tearing down a device.
2807 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2812 Description: service threads may hog cpus when there are a lot of requests
2813 Details : Insert cond_resched to give other threads a chance to use some CPU
2818 Description: the cat log was not initialized in recovery
2819 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2820 unlink log on mds will not be initialized until mds post
2821 recovery. And also in mds post recovery, the unlink log will
2822 initialization will be done asynchronausly, so there will be race
2823 between add unlink log and unlink log initialization.
2827 Description: brw_stats were being printed incorrectly
2828 Details : brw_stats were being printed as log2 but all of them were not
2829 recorded as log2. Also remove some code duplication arising from
2830 filter_tally_{read,write}.
2834 Frequency : rare, only in recovery.
2835 Description: ASSERTION(req->rq_type != LI_POISON) failed
2836 Details : imp_lock should be held while iterating over imp_sending_list for
2837 prevent destroy request after get timeout in ptlrpc_queue_wait.
2841 Description: replay-single.sh test 52 fails
2842 Details : A lock's skiplist need to be cleanup when it being unlinked
2843 from its resource list.
2847 Description: Short directio read returns full requested size rather than
2849 Details : Direct I/O operations should return actual amount of bytes
2850 transferred rather than requested size.
2852 Severity : enhancement
2854 Description: metadata RPC reduction (e.g. for rm performance)
2855 Details : decrease the amount of synchronous RPC between clients and servers
2856 by canceling conflicing lock before the operation on the client
2857 and packing thier handles into the main operation RPC to server.
2859 Severity : enhancement
2861 Description: Async OSC create to avoid the blocking unnecessarily.
2862 Details : If a OST has no remain object, system will block on the creating
2863 when need to create a new object on this OST. Now, ways use
2864 pre-created objects when available, instead of blocking on an
2865 empty osc while others are not empty. If we must block, we block
2866 for the shortest possible period of time.
2870 Description: improve handling recoverable errors
2871 Details : if request processig with error which can be recoverable on server
2872 request should be resend, otherwise page released from cache and
2875 Severity : enhancement
2877 Description: refine locking for avoid write wrong info into lov_objid file
2878 Details : fix possible races with add new target and write/update data in
2881 --------------------------------------------------------------------------------
2883 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2888 Frequency : on some architectures
2890 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2891 Details : Uninitialized flags sometimes cause configuration commands to
2895 Frequency : patchless clients only
2897 Description: missing __iget() symbol export
2898 Details : The __iget() symbol export is missing. To avoid the need for
2899 this on patchless clients the deathrow inode reaper is turned
2900 off, and we depend on the VM to clean up old inodes. This
2901 dependency was during via the fix for bug 12181.
2903 --------------------------------------------------------------------------------
2905 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2907 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2908 INTEROPERATE with older versions automatically. In many cases a
2909 special upgrade step is needed. Please read the
2910 user documentation before upgrading any part of a live system.
2911 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2912 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2913 * WARNING: Lustre configuration and startup changes are required with
2914 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2918 Severity : enhancement
2920 Description: MountConf
2921 Details : Lustre configuration is now managed via mkfs and mount
2922 commands instead of lmc and lconf. New obd types (MGS, MGC)
2923 are added for dynamic configuration management. See
2924 https://mail.clusterfs.com/wikis/lustre/MountConf for
2927 Severity : enhancement
2929 Description: dynamic OST addition
2930 Details : OSTs can now be added to a live filesystem
2932 Severity : enhancement
2934 Description: startup order invariance
2935 Details : MDTs and OSTs can be started in any order. Clients only
2936 require the MDT to complete startup.
2938 Severity : enhancement
2940 Description: parallel, asynchronous orphan cleanup
2941 Details : orphan cleanup is now performed in separate threads for each
2942 OST, allowing parallel non-blocking operation.
2944 Severity : enhancement
2946 Description: optimized stripe assignment
2947 Details : stripe assignments are now made based on ost space available,
2948 ost previous usage, and OSS previous usage, in order to try
2949 to optimize storage space and networking resources.
2951 Severity : enhancement
2953 Description: Permanently set tunables
2954 Details : All writable /proc/fs/lustre tunables can now be permanently
2955 set on a per-server basis, at mkfs time or on a live system.
2957 Severity : enhancement
2959 Description: Lustre message v2
2960 Details : Add lustre message format v2.
2962 Severity : enhancement
2964 Description: client OST exclusion list
2965 Details : Clients can be started with a list of OSTs that should be
2966 declared "inactive" for known non-responsive OSTs.
2970 Description: SPEC SFS validation failure on NFS v2 over lustre.
2971 Details : Changes the blocksize for regular files to be 2x RPC size,
2972 and not depend on stripe size.
2974 Severity : enhancement
2976 Description: Multiple MD RPCs in flight.
2977 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2978 To avoid overly-overloading MDS, introduce a limit on number of
2979 MDS RPCs in flight for a single client and add /proc controls
2980 to adjust this limit.
2982 Severity : enhancement
2984 Description: client read/write statistics
2985 Details : Add client read/write call usage stats for performance
2986 analysis of user processes.
2987 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2988 file access. extents_stats shows chunk size distribution.
2989 extents_stats_per_process show chunk size distribution per
2992 Severity : enhancement
2994 Description: mds statistics
2995 Details : Add detailed mds operations statistics in
2996 /proc/fs/lustre/mds/*/stats.
3000 Description: Failure of copying files with lustre special EAs.
3001 Details : Client side always return success for setxattr call for lustre
3002 special xattr (currently only "trusted.lov").
3007 Description: Refcount LNET uuids
3008 Details : The global LNET uuid list grew linearly with every startup;
3009 refcount repeated list entries instead of always adding to
3012 Severity : enhancement
3014 Description: Dynamic service threads
3015 Details : Within a small range, start extra service threads
3016 automatically when the request queue builds up.
3019 Frequency : mixed-endian client/server environments
3021 Description: mixed-endian crashes
3022 Details : The new msg_v2 system had some failures in mixed-endian
3025 Severity : enhancement
3027 Description: Easy OST removal
3028 Details : OSTs can be permanently deactivated with e.g. 'lctl
3029 conf_param lustre-OST0001.osc.active=0'
3031 Severity : enhancement
3033 Description: MGS proc entries
3034 Details : Added basic proc entries for the MGS showing what filesystems
3037 Severity : enhancement
3039 Description: provide MGS failover
3040 Details : Added config lock reacquisition after MGS server failover.
3042 Severity : enhancement
3044 Description: add Linux 2.4 support
3045 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3049 Description: a large application tries to do I/O to the same resource and dies
3050 in the middle of it.
3051 Details : Check the req->rq_arrival time after the call to
3052 ost_brw_lock_get(), but before we do anything about
3053 processing it & sending the BULK transfer request. This
3054 should help move old stale pending locks off the queue as
3055 quickly as obd_timeout.
3058 Frequency : when an incorrect nid is specified during startup
3060 Description: ptlrpc connect to non-existant node causes kernel crash
3061 Details : LNET can't be re-entered from an event callback, which
3062 happened when we expire a message after the export has been
3063 cleaned up. Instead, hand the zombie cleanup off to another
3066 Severity : enhancement
3068 Description: plain/inodebits lock performance improvement
3069 Details : Grouping plain/inodebits in granted list by their request modes
3070 and bits policy, thus improving the performance of search through
3074 Frequency : only if OST filesystem is corrupted
3076 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3077 Details : for a short time RPCs with bulk IO are in the replay list,
3078 but replay of bulk IOs is unimplemented. If the OST filesystem
3079 is corrupted due to disk cache incoherency and then replay is
3080 started it is possible to trip an assertion. Avoid putting
3081 committed RPCs into the replay list at all to avoid this issue.
3084 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3086 Description: System hang on startup
3087 Details : This bug allowed the liblustre (e.g. catamount) client to
3088 return to the app before handling all startup RPCs. This
3089 could leave the node unresponsive to lustre network traffic
3090 and manifested as a server ptllnd timeout.
3092 Severity : enhancement
3094 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3095 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3096 boolean to control whether to print peer debug info when a
3097 client's RPC times out.
3100 Frequency : only for kernels with patches from Lustre below 1.4.3
3102 Description: Remove old rdonly API
3103 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3106 Frequency : only for devices with external journals
3108 Description: Set external device read-only also
3109 Details : During a commanded failover stop, we set the disk device
3110 read-only while the server shuts down. We now also set any
3111 external journal device read-only at the same time.
3114 Frequency : when upgrading from 1.4 while trying to change parameters
3116 Description: The wrong (new) MDC name was used when setting parameters for
3117 upgraded MDT's. Also allows changing of OSC (and MDC)
3118 parameters if --writeconf is specified at tunefs upgrade time.
3121 Frequency : when setting specific ost indicies
3123 Description: QOS code breaks on skipped indicies
3124 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3125 created with --index need not be sequential.
3127 Severity : enhancement
3129 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3130 Details : The uninit_groups feature works in conjunction with the kernel
3131 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3132 up the pass1 processing of e2fsck. This is a read-only feature
3133 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3134 mount filesystems that have had this feature enabled.
3136 Severity : enhancement
3138 Description: Improve multi-block allocation algorithm to avoid fragmentation
3139 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3140 allocation locality and avoid filesystem fragmentation.
3142 ------------------------------------------------------------------------------
3144 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3146 * Support for kernels:
3147 2.4.21-47.0.1.EL (RHEL 3)
3148 2.6.5-7.283 (SLES 9)
3149 2.6.9-42.0.10.EL (RHEL 4)
3150 2.6.12.6 vanilla (kernel.org)
3151 2.6.16.27-0.9 (SLES 10)
3152 * Recommended e2fsprogs version: 1.39.cfs5
3154 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3158 Frequency : occasional, depends on client load and configuration
3159 Bugzilla : 12181, 12203
3160 Description: data loss for recently-modified files
3162 Details : In some cases it is possible that recently written or created
3163 files may not be written to disk in a timely manner (this should
3164 normally be within 30s unless client IO load is very high).
3165 The problem appears as zero-length files or files that are a
3166 multiple of 1MB in size after a client crash or client eviction
3167 that are missing data at the end of the file.
3169 This problem is more likely to be hit on clients where files are
3170 repeatedly created and unlinked in the same directory, clients
3171 have a large amount of RAM, have many CPUs, the filesystem has
3172 many OSTs, the clients are rebooted frequently, and/or the files
3173 are not accessed by other nodes after being written.
3175 The presence of the problem can be detected by looking at
3176 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3177 smaller than the second (nr_unused) then dirty files will not
3178 be flushed automatically to disk. "sync; sleep 10" should be
3179 run several times on the node before unmounting it to update
3180 Lustre (this is also safe to run on nodes without this problem).
3182 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3183 that can cause this same problem, so customers using that kernel
3184 also need to update the kernel in addition to Lustre. In order
3185 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3187 It is normal that files written just before a client crash (less
3188 than 30s) may not yet have been flushed to disk, even for local
3192 Frequency : frequent on thin XT3 nodes
3194 Description: UUID collision on thin XT3 Linux nodes
3195 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3196 randomly, since we relied on an insufficiently-seeded PRNG.
3201 Description: OSS hangs after "All ost request buffers busy"
3202 Details : A deadlock between quota and journal operations caused OSS
3203 hangs after printing "All ost request buffers busy."
3206 Frequency : always on liblustre builds
3208 Description: Cleanup compiler warnings on liblustre
3211 Frequency : always on liblustre builds on XT3
3213 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3217 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3218 > 32000 subdirectories
3219 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3220 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3221 subdirectory count crosses 32000. This will aid e2fsck to
3222 correctly handle more than 32000 subdirectories.
3225 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3227 Description: System hang on startup
3228 Details : This bug allowed the liblustre (e.g. catamount) client to
3229 return to the app before handling all startup RPCs. This
3230 could leave the node unresponsive to lustre network traffic
3231 and manifested as a server ptllnd timeout.
3233 Severity : enhancement
3235 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3236 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3237 boolean to control whether to print peer debug info when a
3238 client's RPC times out.
3243 Description: make O_SYNC working on 2.6 kernels
3244 Details : 2.6 kernels use different method for mark pages for write,
3245 so need add a code to lustre for O_SYNC work.
3250 Description: Failure to close file and release space on NFS
3251 Details : Put inode details into lock acquired in ll_intent_file_open.
3252 Use mdc_intent_lock in ll_intent_open to properly
3253 detect all kind of errors unhandled by mdc_enqueue.
3258 Description: proc file read during shutdown sometimes raced obd removal,
3260 Details : Add lock to prevent obd access after proc file removal.
3263 Frequency : Only for files larger than 4GB on 32-bit clients.
3265 Description: improperly doing page alignment of locks
3266 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3267 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3272 Description: RPCs being resent when they shouldn't be
3273 Details : Some RPCs that should not be resent are being resent. This
3274 can cause inconsistencies in the RPC state machine. Do not
3275 resend such requests.
3278 Frequency : rare, only with NFS export
3280 Description: Crash on NFS re-export node
3281 Details : under very unusual load conditions an assertion is hit in
3282 ll_intent_file_open()
3285 Frequency : only if OST filesystem is corrupted
3287 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3288 Details : for a short time RPCs with bulk IO are in the replay list,
3289 but replay of bulk IOs is unimplemented. If the OST filesystem
3290 is corrupted due to disk cache incoherency and then replay is
3291 started it is possible to trip an assertion. Avoid putting
3292 committed RPCs into the replay list at all to avoid this issue.
3297 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3298 Details : Large single O_DIRECT read and write calls can fail to allocate
3299 a sufficiently large buffer to process the request. In case of
3300 allocation failure the allocation is retried with a smaller
3301 buffer and broken into smaller requests.
3303 Severity : enhancement
3305 Description: Add -o localflock option to simulate old noflock behaviour.
3306 Details : This will achieve local-only flock/fcntl locks coherentness.
3311 Description: versioning check is incomplete
3312 Details : Checking the version difference of client vs. server, report
3313 error if the gap is too big.
3318 Description: add support PG_writeback bit
3319 Details : add support for PG_writeback bit for Lustre, for more carefull
3320 work with page cache in 2.6 kernel. This also fix some deadlocks
3321 and remove hack for work O_SYNC with 2.6 kernel.
3323 Severity : enhancement
3325 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3326 Details : The uninit_groups feature works in conjunction with the kernel
3327 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3328 up the pass1 processing of e2fsck. This is a read-only feature
3329 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3330 mount filesystems that have had this feature enabled.
3332 Severity : enhancement
3334 Description: Improve multi-block allocation algorithm to avoid fragmentation
3335 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3336 allocation locality and avoid filesystem fragmentation.
3338 ------------------------------------------------------------------------------
3340 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3342 * Support for kernels:
3343 2.6.9-42.0.3.EL (RHEL 4)
3344 2.6.5-7.276 (SLES 9)
3345 2.4.21-47.0.1.EL (RHEL 3)
3346 2.6.12.6 vanilla (kernel.org)
3347 2.6.16.21-0.8 (SLES10)
3348 * Recommended e2fsprogs version: 1.39.cfs2-0
3350 * The backwards-compatible /proc/sys/portals symlink has been removed
3351 in this release. Before upgrading, please ensure that you change
3352 any configuration scripts or /etc/sysctl.conf files that access
3353 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3354 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3355 in advance of the upgrade on any system running Lustre 1.4.6 or
3356 newer, since /proc/sys/lnet was added in that version.
3357 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3361 Frequency : only when quota is used
3363 Description: avoid scanning export list for quota master
3364 Details : Change the algorithms to avoid scanning export list in order
3365 to improve the efficiency.
3368 Frequency : MDS failover only, very rarely
3370 Description: "went back in time" messages on mds failover
3371 Details : The greatest transno may be lost when the current operation
3372 finishes with an error (transno==0) and the client's last_rcvd
3373 record is over-written. Save the greatest transno in the
3374 mds_last_transno for this case.
3377 Frequency : always for specific kernels and striping counts
3379 Description: client may get "Matching packet too big" without ACL support
3380 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3381 when trying to access files in certain configurations. The
3382 clients should in fact be denied when mounting because they do
3383 not understand ACLs.
3386 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3388 Description: many clients connecting with IO in progress causes connect timeouts
3389 Details : Avoid synchronous journal commits to avoid delays caused by many
3390 clients connecting/disconnecting when bulk IO is in progress.
3391 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3392 OST_IO_PORTAL to avoid delays behind potentially many pending
3396 Frequency : occasionally with multiple writers to a single file
3398 Description: shared writes to file may result in wrong size reported by stat()
3399 Details : Allow growing of kms when extent lock is cancelled
3402 Frequency : always with random mmap IO to multi-striped file
3404 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3405 Details : Only if the hole is at the end of OST object so that kms is too
3406 small. Fix is to increase kms accordingly in ll_nopage.
3409 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3411 Description: writes to a missing object would leak memory on the OST
3412 Details : If there is an inconsistency between the MDS and OST filesystems,
3413 such that the MDS references an object that doesn't exist, writes
3414 to that object will leak memory due to incorrect cleanup in the
3415 error handling path, eventually running out of memory on the OST.
3420 Description: Creating too long symlink causes lustre errors
3421 Details : Check symlink and name lengths before sending requests to MDS.
3424 Frequency : only if flock is enabled (not on by default)
3426 Description: posix locks not released on fd closure on 2.6.9+
3427 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3428 kernels, this caused such locks not to be released on fd close and
3429 then assertions on fs unmount about still used locks.
3432 Frequency : MDS failover only, very rarely
3434 Description: clients may get ASSERTION(granted_lock != NULL)
3435 Details : When request was taking a long time, and a client was resending
3436 a getattr by name lock request. The were multiple lock requests
3437 with the same client lock handle and
3438 mds_getattr_name->fixup_handle_for_resent_request found one of the
3439 lock handles but later failed with ASSERTION(granted_lock != NULL).
3444 Description: handle->h_buffer_credits > 0, assertion failure
3445 Details : h_buffer_credits is zero after truncate, causing assertion
3446 failure. This patch extends the transaction or creates a new
3450 Frequency : NFS re-export or patchless client
3451 Bugzilla : 11179, 10796
3452 Description: Crash on NFS re-export node (__d_move)
3453 Details : We do not want to hash the dentry if we don't have a lock.
3454 But if this dentry is later used in d_move, we'd hit uninitialised
3455 list head d_hash, so we just do this to init d_hash field but
3456 leave dentry unhashed.
3459 Frequency : NFS re-export or patchless client
3461 Description: NFS exports has problem with symbolic link
3462 Details : lustre client didn't properly install dentry when re-exported
3463 to NFS or running patchless client.
3466 Frequency : NFS re-export or patchless client
3468 Description: Various nfs/patchless fixes.
3469 Details : fixes reuse disconected alias for lookup process - this fixes
3470 warning "find_exported_dentry: npd != pd",
3471 fix permission error with open files at nfs.
3472 fix apply umask when do revalidate.
3475 Frequency : occasional
3477 Description: Crash on NFS re-export node
3478 Details : calling clear_page() on the wrong pointer triggered oops in
3479 generic_mapping_read().
3482 Frequency : rarely, using O_DIRECT IO
3484 Description: unaligned directio crashes client with LASSERT
3485 Details : check for unaligned buffers before trying any requests.
3488 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3490 Description: stale data returned from RAID cache
3491 Details : If only a small amount of IO is done to the RAID device before
3492 reading it again it is possible to get stale data from the RAID
3493 cache instead of reading it from disk.
3496 Frequency : always for sles10 kernel
3498 Description: sles10 support
3499 Details : ll_follow_link: compile fixes and using of nd_set_link
3500 under newer kernels.
3503 Frequency : depends on arch, kernel and compiler version, always on sles10
3506 Description: recursive or deep enough symlinks cause stack overflow
3507 Details : getting rid of large stack-allocated variable in
3511 Frequency : depends on hardware
3513 Description: lustre write performance loss in the SLES10 kernel
3514 Details : the performance loss is caused by using of write barriers in the
3515 ext3 code. The SLES10 kernel turns barrier support on by
3516 default. The fix is to undo that change for ldiskfs.
3518 ------------------------------------------------------------------------------
3520 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3522 * Support for kernels:
3523 2.6.9-42.0.3EL (RHEL 4)
3524 2.6.5-7.276 (SLES 9)
3525 2.4.21-47.0.1.EL (RHEL 3)
3526 2.6.12.6 vanilla (kernel.org)
3530 Frequency : quota enabled and large files being deleted
3532 Description: releasing more than 4GB of quota at once hangs OST
3533 Details : If a user deletes more than 4GB of files on a single OST it
3534 will cause the OST to spin in an infinite loop. Release
3535 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3540 Description: statfs data retrieved from /proc may be stale or zero
3541 Details : When reading per-device statfs data from /proc, in the
3542 {kbytes,files}_{total,free,avail} files, it may appear
3543 as zero or be out of date.
3546 Frequency : systems with MD RAID1 external journal devices
3548 Description: lconf's call to blkid is confused by RAID1 journal devices
3549 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3550 of returning all block devices that match the journal UUID.
3553 Frequency : always, for aggregate stripe size over 4GB
3555 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3556 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3557 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3560 Frequency : always if "lfs find" used on a local file/directory
3562 Description: "lfs find" segfaults if used on a local file/directory
3563 Details : The case where a directory component was not specified wasn't
3564 handled correctly. Handle this properly.
3567 Frequency : always on ppc64
3569 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3570 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3571 perform endianness conversion.
3574 Frequency : rarely (truncate to non-zero file size after write under load)
3575 Bugzilla : 10730, 10687
3576 Description: Files padded with zeros to next 4K multiple
3577 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3578 it is possible that files that are truncated to a non-zero size
3579 immediately after being written are filled with zero bytes beyond
3580 the truncated size. No file data is lost.
3582 Severity : enhancement
3584 Description: Allow recovery/failover for liblustre clients.
3585 Details : liblustre clients were unaware of failover configurations until
3588 Severity : enhancement
3590 Description: user file locks should fail when not mounting with flock option
3591 Details : Set up an error-returning stub in ll_file_operations.lock field
3592 to prevent incorrect behaviour when client is mounted without
3593 flock option. Also, set up properly f_op->flock field for
3597 Frequency : always on ia64
3599 Description: "lfs df" loops on printing out MDS statfs information
3600 Details : The obd_ioctl_data was not initialized and in some systems
3601 this caused a failure during the ioctl that did not return
3602 an error. Initialize the struct and return an error on failure.
3605 Frequency : SLES 9 only
3607 Description: Error of copying files with lustre special EAs as root
3608 Details : Client side always return success for setxattr call for lustre
3609 special xattr (currently only "trusted.lov").
3612 Frequency : rarely on clusters with both ia64+i386 clients
3614 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3615 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3616 clients are doing concurrent writes to the same file it is
3617 possible that the ia64 clients may LASSERT because the OST
3618 extent locks are not PAGE_SIZE aligned. Ensure that grown
3619 locks are always aligned on the request boundary.
3622 Frequency : specific use, occasional
3624 Description: Overwriting in use executable truncates on-disk binary image
3625 Details : If one node attempts to overwrite an executable in use by
3626 another node, we now correctly return ETXTBSY instead of
3627 truncating the file.
3629 Severity : enhancement
3631 Description: Async OSC create to avoid the blocking unnecessarily.
3632 Details : If a OST has no remain object, system will block on the creating
3633 when need to create a new object on this OST. Now, ways use
3634 pre-created objects when available, instead of blocking on an
3635 empty osc while others are not empty. If we must block, we block
3636 for the shortest possible period of time.
3641 Description: chmod on Lustre root is propagated to other clients
3642 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3643 invalid by the follow_mount time.
3646 Frequency : liblustre clients only
3648 Description: Race in 'instant cancel' lock handling could lead to such locks
3649 never to be granted in case of SMP MDS
3650 Details : Do not destroy not yet granted but cbpending locks in
3654 Frequency : replay/resend of open
3656 Description: non null lock assetion failure in mds_intent_policy
3657 Details : Trying to replay/resend lockless open requests resulted in
3658 mds_open() returning 0 with no lock. Now it sets a flag if
3659 a lock is going to be returned.
3661 Severity : enhancement
3663 Description: Checksum enhancements
3664 Details : New checksum enhancements allow for resending RPCs that failed
3667 Severity : enhancement
3669 Description: Tunables on number of dirty pages in cacche
3670 Details : Allow to set limit on number of dirty pages cached.
3675 Description: client crash on unmount - lock still has references
3676 Details : In some error handling cases it was possible to leak a lock
3677 reference on a client while accessing a file. This was not
3678 harmful to the client during operation, but would cause the
3679 client to crash when the filesystem is unmounted.
3682 Frequency : specific case, rare
3684 Description: ETXTBSY on mds though file not in use by client
3685 Details : ETXTBSY is no longer incorrectly returned when attempting to
3686 chmod or chown a directory that the user previously tried to
3687 execute or a currently-executing binary.
3690 Frequency : extremely rare except on liblustre-based clients
3692 Description: Lustre space not freed when files are deleted
3693 Details : Clean up open-unlinked files after client eviction. Previously
3694 the unlink was skipped and the files remained as orphans.
3699 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3700 Details : In some cases it was possible to send a blocking callback to
3701 a client doing a glimpse, even though that client didn't get
3702 a lock granted. When the glimpse lock is cancelled on the OST
3703 the freed lock is left on the waiting list and corrupted the list.
3706 Frequency : all core dumps
3708 Description: Broke core dumps to lustre
3709 Details : Negative dentry may be unhashed if parent does not have UPDATE
3710 lock, but some callers, e.g. do_coredump, expect dentry to be
3711 hashed after successful create, hash it in ll_create_it.
3713 ------------------------------------------------------------------------------
3715 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3717 * Support for kernels:
3718 2.6.9-42.0.2.EL (RHEL 4)
3719 2.6.5-7.276 (SLES 9)
3720 2.4.21-40.EL (RHEL 3)
3721 2.6.12.6 vanilla (kernel.org)
3725 Frequency : always on RHEL 3
3727 Description: Number of open files grows over time
3728 Details : The number of open files grows over time, whether or not
3729 Lustre is started. This was due to a filp leak introduced
3730 by one of our kernel patches.
3732 ------------------------------------------------------------------------------
3734 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3736 * Support for kernels:
3737 2.6.9-42.EL (RHEL 4)
3738 2.6.5-7.276 (SLES 9)
3739 2.4.21-40.EL (RHEL 3)
3740 2.6.12.6 vanilla (kernel.org)
3745 Bugzilla : 5719, 9635, 9792, 9684
3746 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3747 Details : If a server is under heavy load and cannot reply to new
3748 connection requests before the client resends the (re)connect,
3749 the connection handling code can behave badly if two service
3750 threads are concurrently handing separate (re)connections from
3751 the same client. Add better locking to the connection handling
3752 code, and ensure that only a single connection will be processed
3753 for a given client UUID, even if the lock is dropped.
3755 Severity : enhancement
3757 Description: add TCP zero-copy support to kernel
3758 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3759 sends if the hardware supports scatter-gather and checksumming.
3760 This allows socklnd to do client-write and server-read more
3761 efficiently and reduce CPU utilization from skbuf copying.
3764 Frequency : only if NFS exporting from client
3766 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3767 Details : If setting attributes on a file created under NFS that had
3768 never been opened it would be possible to oops the client
3769 if the file had no objects.
3772 Frequency : always for liblustre
3774 Description: liblustre client does MDS+OSTs setattr RPC for each write
3775 Details : When doing a write from a liblustre client, the client
3776 incorrectly issued an RPC to the MDS and each OST the file was
3777 striped over in order to update the timestamps. When writing
3778 with small chunks and many clients this could overwhelm the MDS
3779 with RPCs. In all cases it would slow down the write because
3780 these RPCs are unnecessary.
3782 Severity : enhancement
3784 Description: allow number of MDS service threads to be changed at module load
3785 Details : It is now possible to change the number of MDS service threads
3786 running. Adding "options mds mds_num_threads={N}" to the MDS's
3787 /etc/modprobe.conf will set the number of threads for the next
3788 time Lustre is restarted (assuming the "mds" module is also
3789 reloaded at that time). The default number of threads will
3790 stay the same, 32 for most systems.
3795 Description: OST crash if filesystem is unformatted or corrupt
3796 Details : If an OST is started on a device that has never been formatted
3797 or if the filesystem is corrupt and cannot even mount then the
3798 error handling cleanup routines would dereference a NULL pointer.
3803 Description: NULL pointer deref in llap_from_page.
3804 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3805 incorrect) mapping if the page was truncated/reclaimed while it was
3806 searched for. Check for this condition and skip such pages when
3807 doing readahead. Introduce extra check to llap_from_page() to
3808 verify page->mapping->host is non-NULL (so page is not anonymous).
3811 Frequency : Sometimes when using sys_sendfile
3813 Description: "page not covered by a lock" warnings from ll_readpage
3814 Details : sendfile called ll_readpage without right page locks present.
3815 Now we introduced ll_file_sendfile that does necessary locking
3816 around call to generic_file_sendfile() much like we do in
3820 Frequency : with certain MDS communication failures at client mount time
3822 Description: NULL pointer deref after failed client mount
3823 Details : a client connection request may delayed by the network layer
3824 and not be sent until after the PTLRPC layer has timed out the
3825 request. If the client fails the mount immediately it will try
3826 to clean up before the network times out the request. Add a
3827 reference from the request import to the obd device and delay
3828 the cleanup until the network drops the request.
3831 Frequency : occasionally during client (re)connect
3833 Description: assertion failure during client (re)connect
3834 Details : processing a client connection request may be delayed by the
3835 client or server longer than the client connect timeout. This
3836 causes the client to resend the connection request. If the
3837 original connection request is replied in this interval, the
3838 client may trip an assertion failure in ptlrpc_connect_interpret()
3839 which thought it would be the only running connect process.
3842 Frequency : only with obd_echo servers and clients that are rebooted
3844 Description: kernel BUG accessing uninitialized data structure
3845 Details : When running an obd_echo server it did not start the ping_evictor
3846 thread, and when a client was evicted an uninitialized data
3847 structure was accessed. Start the ping_evictor in the RPC
3848 service startup instead of the OBD startup.
3850 Severity : enhancement
3851 Bugzilla : 10193 (patchless)
3852 Description: Remove dependency on various unexported kernel interfaces.
3853 Details : No longer need reparent_to_init, exit_mm, exit_files,
3854 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3857 Frequency : rare (only users of deprecated and unsupported LDAP config)
3859 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3860 Details : LDAP apparently contains 'lustreName' attributes instead of
3861 'name'. A simple remapping of the name is sufficient.
3864 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3866 Description: waiting_locks_callback trips kernel BUG if client is evicted
3867 Details : Running with the dump_on_timeout debug flag turned on makes
3868 it possible that the waiting_locks_callback() can try to dump
3869 the Lustre kernel debug logs from an interrupt handler. Defer
3870 this log dumping to the expired_lock_main() thread.
3872 Severity : enhancement
3874 Description: Support NFS exporting on 2.6 kernels.
3875 Details : Implement non-rawops metadata methods for NFS server to use without
3876 changing NFS server code.
3879 Frequency : very rare (synthetic metadata workload only)
3881 Description: two racing renames might cause an MDS thread to deadlock
3882 Details : Running the "racer" program may cause one MDS thread to rename
3883 a file from being the source of a rename to being the target of
3884 a rename at exactly the same time that another thread is doing
3885 so, and the second thread has already enqueued these locks after
3886 doing a lookup of the target and is trying to relock them in
3887 order. Ensure that we don't try to re-lock the same resource.
3890 Frequency : only very large systems with liblustre clients
3892 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3893 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3894 all OSTs in parallel. This allows the eviction of stale liblustre
3895 clients to proceed much faster than if they were done in series,
3896 and also offers similar improvements for other set_info RPCs.
3901 Description: excessive CPU usage during initial read phase on client
3902 Details : During the initial read phase on a client, it would agressively
3903 retry readahead on the file, consuming too much CPU and impacting
3904 performance (since 1.4.5.8). Improve the readahead algorithm
3905 to avoid this, and also improve some other common cases (read
3906 of small files in particular, where "small" is files smaller than
3907 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3912 Description: MDS crash when receiving packet with unknown intent.
3913 Details : Do not LBUG in unknown intent case, just return -EFAULT
3915 Severity : enhancement
3916 Bugzilla : 9293, 9385
3917 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3918 Details : Do not serialize getattr (non-intent version) and statfs.
3921 Frequency : occasional, when OST network is overloaded/intermittent
3923 Description: client evicted by OST after bulk IO timeout
3924 Details : If a client sends a bulk IO request (read or write) the OST
3925 may evict the client if it is unresposive to its data GET/PUT
3926 request. This is incorrect if the network is overloaded (takes
3927 too long to transfer the RPC data) or dropped the OST GET/PUT
3928 request. There is no need to evict the client at all, since
3929 the pinger and/or lock callbacks will handle this, and the
3930 client can restart the bulk request.
3933 Frequency : Always when mmapping file with no objects
3935 Description: client crashes when mmapping file with no objects
3936 Details : Check that we actually have objects in a file before doing any
3937 operations on objects in ll_vm_open, ll_vm_close and
3943 Description: Request leak when working with deleted CWD
3944 Details : Introduce advanced request refcount tracking for requests
3945 referenced from lustre intent.
3947 Severity : Enhancement
3949 Description: Cache open file handles on client.
3950 Details : MDS now will return special lock along with openhandle, if
3951 requested and client is allowed to hold openhandle, even if unused,
3952 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3953 closing files for every read/write openration.
3955 Severity : Enhancement
3957 Description: Cache open negative dentries on client when possible.
3958 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3959 negative dentries on lock revocation.
3964 Description: Remounting a client read-only wasn't possible with a zconf mount
3965 Details : It wasn't possible to remount a client read-only with llmount.
3967 Severity : enhancement
3968 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3969 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3970 MPICH is included to add Lustre-specific ADIO interfaces.
3971 This is based closely on the UFS ADIO layer and only differs
3972 in file creation, in order to allow the OST striping to be set.
3973 This is user-contributed code and not supported by CFS.
3978 Description: extended inode attributes (immutable, append-only) work improperly
3979 when 2.4 and 2.6 kernels are used on client/server or vice versa
3980 Details : Introduce kernel-independent values for these flags.
3982 Severity : enhancement
3985 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3986 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3987 granularity. Now, Fractional MB granularity is be supported,
3988 this is very useful for low memory system.
3990 Severity : enhancement
3992 Description: Getattr by fid
3993 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3994 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3995 have locks and avoids getattr rpc requests for it on every stat.
3998 Frequency : Always, for filesystems larger than 2TB
4000 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4001 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4002 filesystems larger than 2TB to be mounted with mballoc enabled.
4005 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4007 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4008 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4009 filesystems larger than 2TB is mounted then the kernel will
4010 silently corrupt the start of the filesystem. CONFIG_LBD is
4011 enabled for all CFS-supported kernels, but the possibility of
4012 this happening with a modified kernel config exists.
4014 Severity : enhancement
4016 Description: add client O_DIRECT support for 2.6 kernels
4017 Details : It is now possible to do O_DIRECT reads and writes to files
4018 in the Lustre client mountpoint on 2.6 kernel clients.
4020 Severity : enhancement
4022 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4023 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4024 parallel, not waiting for response from every OST before sending
4025 a rpc to the next OST.
4030 Description: setattr vs write race when updating file timestamps
4031 Details : Client processes that update a file timestamp into the past
4032 right after writing to the file (e.g. tar) it is possible that
4033 the updated file modification time can be reset to the current
4034 time due to a race between processing the setattr and write RPC.
4036 Severity : enhancement
4038 Description: Bring 'lfs find' closer in line with regular Linux find.
4039 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4040 -print0 options and obtains all the needed info through the lustre
4043 Severity : enhancement
4045 Description: support up to 1024 configured devices on one node
4046 Details : change obd_dev array from statically allocated to dynamically
4047 allocated structs as they are first used to reduce memory usage
4052 Description: Flush dirty partially truncated pages during truncate
4053 Details : Immediatelly flush partially truncated pages in filter_setattr,
4054 this way we completely avoid having any pages in page cache on OST
4055 and can retire ugly workarounds during writes to flush such pages.
4060 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4061 Details : protect inode from truncation within vfs_unlink() context
4062 just take a reference before calling vfs_unlink() and release it
4063 when parent's i_sem is free.
4068 Description: last_id value checked outside lock on OST caused LASSERT failure
4069 Details : If there were multiple MDS->OST object precreate requests in
4070 flight, it was possible that the OST's last object id was checked
4071 outside a lock and incorrectly tripped an assertion. Move checks
4072 inside locks, and discard old precreate requests.
4075 Frequency : always, if extents are used on OSTs
4077 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4078 Details : OSTs using the extents format would not zero the high 16 bits of
4079 the index physical block number. This is not a problem for any
4080 OST filesystems smaller than 16TB, and no kernels support ext3
4081 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4082 new/modified files) and can be fixed for existing filesystems
4083 with e2fsprogs-1.39-cfs1.
4088 Description: import connection selection may be incorrect if timer wraps
4089 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4090 import connections to be ignored if the 32-bit jiffies counter
4091 wraps. Use a 64-bit jiffies counter.
4094 Frequency : very large clusters immediately after boot
4096 Description: LNET request buffers exhausted under heavy short-term load
4097 Details : If a large number of client requests are generated on a service
4098 that has previously never seen so many requests it is possible
4099 that the request buffer growth cannot keep up with the spike in
4100 demand. Instead of dropping incoming requests, they are held in
4101 the LND until the RPC service can accept more requests.
4104 Frequency : Sometimes during replay
4106 Description: Assertion failure in ll_local_open after replay.
4107 Details : If replay happened on an open request reply before we were able
4108 to set replay handler, reply will become not swabbed tripping the
4109 assertion in ll_local_open. Now we set the handler right after
4110 recognising of open request
4113 Frequency : very rare
4115 Description: kernel reports "badness in vsnprintf"
4116 Details : Reading from the "recovery_status" /proc file in small chunks
4117 may cause a negative length in lprocfs_obd_rd_recovery_status()
4118 call to vsnprintf() (which is otherwise harmless). Exit early
4119 if there is no more space in the output buffer.
4121 Severity : enhancement
4123 Description: clear OBD RPC statistics by writing to them
4124 Details : It is now possible to clear the OBD RPC statistics by writing
4125 to the "stats" file.
4130 Description: Client mtime is not the same on different clients after utimes
4131 Details : In some cases, the client was using the utimes() syscall on
4132 a file cached on another node. The clients now validate the
4133 ctime from the MDS + OSTs to determine which one is right.
4138 Description: Inability to activate failout mode
4139 Details : lconf script incorrectly assumed that in pythong string's numeric
4140 value is used in comparisons.
4143 Frequency : always with multiple stripes per file
4145 Description: Inefficient object allocation for mutli-stripe files
4146 Details : When selecting which OSTs to stripe files over, for files with
4147 a stripe count that divides evenly into the number of OSTs,
4148 the MDS is always picking the same starting OST for each file.
4149 Return the OST selection heuristic to the original design.
4154 Description: mount failures may take full timeout to return an error
4155 Details : Under some heavy load conditions it is possible that a
4156 failed mount can wait for the full obd_timeout interval,
4157 possibly several minutes, before reporting an error.
4158 Instead return an error as soon as the status is known.
4160 Frequency : quota enabled and large files being deleted
4162 Description: releasing more than 4GB of quota at once hangs OST
4163 Details : If a user deletes more than 4GB of files on a single OST it
4164 will cause the OST to spin in an infinite loop. Release
4165 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4170 Description: statfs data retrieved from /proc may be stale or zero
4171 Details : When reading per-device statfs data from /proc, in the
4172 {kbytes,files}_{total,free,avail} files, it may appear
4173 as zero or be out of date.
4176 Frequency : systems with MD RAID1 external journal devices
4178 Description: lconf's call to blkid is confused by RAID1 journal devices
4179 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4180 of returning all block devices that match the journal UUID.
4183 Frequency : always, for aggregate stripe size over 4GB
4185 Description: assertion fails when trying to use 4GB stripe size
4186 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4187 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4190 Frequency : always on ppc64
4192 Description: the first write on an ext3 filesystem with mballoc got stuck
4193 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4194 perform endianness conversion.
4196 ------------------------------------------------------------------------------
4198 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4200 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4201 INTEROPERATE with older versions automatically. Please read the
4202 user documentation before upgrading any part of a live system.
4203 * WARNING: Lustre networking configuration changes are required with
4204 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4207 * Support for kernels:
4208 2.6.9-22.0.2.EL (RHEL 4)
4209 2.6.5-7.244 (SLES 9)
4210 2.6.12.6 vanilla (kernel.org)
4213 Severity : enhancement
4214 Bugzilla : 7981/8208
4215 Description: Introduced Lustre Networking (LNET)
4216 Details : LNET is new networking infrastructure for Lustre, it includes
4217 a reorganized network configuration mode (see the user
4218 documentation for full details) as well as support for routing
4219 between different network fabrics. Lustre Networking Devices
4220 (LNDs) for the supported network fabrics have also been
4221 created for this new infrastructure.
4223 Severity : enhancement
4224 Description: Introduced Access control lists
4225 Details : clients can set ACLs on files and directories in order to have
4226 more fine-grained permissions than the standard Unix UGO+RWX.
4227 The MDS must be started with the "-o acl" mount option.
4229 Severity : enhancement
4230 Description: Introduced filesystem quotas
4231 Details : Administrators may now establish per-user quotas on the
4234 Severity : enhancement
4236 Description: Configuration change for the XT3
4237 The PTLLND is now used to run Lustre over Portals on the XT3
4238 The configure option(s) --with-cray-portals are no longer used.
4239 Rather --with-portals=<path-to-portals-includes> is used to
4240 enable building on the XT3. In addition to enable XT3 specific
4241 features the option --enable-cray-xt3 must be used.
4246 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4247 Details : A race between allocating a new llcd and re-getting the llcd_lock
4248 allowed another thread to grab newly-allocated llcd.
4250 Severity : enhancement
4252 Description: 2.6 OST async journal commit and locking fix to improve performance
4253 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4254 2.6 kernels are now async as they already were in 2.4 kernels so
4255 that they can commit concurrently with the network bulk transfer.
4256 For block-allocated files the filter allocation semaphore is held
4257 to avoid filesystem fragmentation during allocation. BKL lock
4258 removed for 2.6 xattr operations where it is no longer needed.
4263 Description: lconf incorrectly determined whether two IP networks could talk
4264 Details : In some more complicated routing and multiple-network
4265 configurations, lconf will avoid trying to make a network
4266 connection to a disjoint part of the IP space. It was doing the
4267 math incorrectly for one set of cases.
4272 Description: Fix for potential infinite loop processing records in an llog.
4273 Details : If an llog record is corrupted/zeroed, it is possible to loop
4274 forever in llog_process(). Validate the llog record length
4275 and skip the remainder of the block on error.
4278 Frequency : occasional (liblustre only)
4280 Description: liblustre could not open files whose last component is a symlink
4281 Details : sysio_path_walk() would incorrectly pass the open intent to
4282 intermediate path components.
4285 Frequency : rare (liblustre only with non-standard tuning)
4286 Bugzilla : 7201 (7350)
4287 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4288 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4289 after it has been released, only do this for referenced locks
4291 Severity : enhancement
4293 Description: specifying an (invalid) directory default stripe_size of -1
4294 would reset the directory default striping
4295 Details : stripe_size -1 was used internally to signal directory stripe
4296 removal, now use "all default" to signal dir stripe removal
4297 as a directory striping of "all default" is not useful
4300 Frequency : common for large clusters running liblustre clients
4302 Description: doing an ls when liblustre clients are running is slow
4303 Details : sending a glimpse AST to a liblustre client waits for every AST
4304 to time out, as liblustre clients will not respond. Since they
4305 cannot cache data we refresh the OST lock LVB from disk instead.
4307 Severity : enhancement
4309 Description: doing an ls at the same time as file IO can be slow
4310 Details : enqueue and other "small" requests can be blocked behind many
4311 large IO requests. Create a new OST IO portal for non-IO
4312 requests so they can be processed faster.
4315 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4317 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4318 created file with mode 000
4319 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4320 allow an ftruncate() as a non-root user to a file with mode 000.
4321 Lustre now respects this flag to disable mode checks when
4322 truncating a file owned by the user
4325 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4328 Description: Revoking locks from clients that went dead or catatonic might take
4330 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4331 cancellation of such locks instant on servers without waiting for
4332 any reply from clients. Clients drops these locks when cancel
4333 notification from server is received without replying.
4336 Frequency : liblustre-only, when liblustre client dies or becomes busy
4338 Description: Doing ls on Linux clients can take a long time with active
4340 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4341 granting such locks to it in the first place if possible. Locks
4342 are taken by proxy on the OST during the read or write and
4343 dropped immediately afterward. Add connect flags handling, do
4344 not grant locks to liblustre clients for glimpse ASTs.
4346 Severity : enhancement
4348 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4349 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4350 sequential reads and behaves badly with applications doing random
4351 reads. Improve it by reading ahead at least read region, and
4352 avoiding excessive large RPC for small reads.
4354 Severity : enhancement
4356 Description: Creating more than 1000 files for a single job may cause a load
4357 imbalance on the OSTs if there are also a large number of OSTs.
4358 Details : qos_prep_create() uses an OST index reseed value that is an
4359 even multiple of the number of available OSTs so that if the
4360 reseed happens in the middle of the object allocation it will
4361 still utilize the OSTs as uniformly as possible.
4366 Description: OST or MDS may oops in ping_evictor_main()
4367 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4368 but doesn't restart at beginning of obd_exports_timed list
4371 Severity : enhancement
4373 Description: improve by-nid export eviction on the MDS and OST
4374 Details : allow multiple exports with the same NID to be evicted at one
4375 time without re-searching the exports list.
4378 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4380 Description: MDS may oops in groups_free()
4381 Details : in rare race conditions a newly allocated group_info struct is
4382 freed again, and this can be NULL. The 2.4 compatibility code
4383 for groups_free() checked for a NULL pointer, but 2.6 did not.
4386 Frequency : common for liblustre clients doing little filesystem IO
4387 Bugzilla : 9352, 7313
4388 Description: server may evict liblustre clients accessing contended locks
4389 Details : if a client is granted a lock or receives a completion AST
4390 with a blocking AST already set it would not reply to the AST
4391 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4394 Frequency : lfs setstripe, only systems with more than 160 OSTs
4396 Description: unable to set striping with a starting offset beyond OST 160
4397 Details : llapi_create_file() incorrectly limited the starting stripe
4398 index to the maximum single-file stripe count.
4401 Frequency : LDAP users only
4403 Description: lconf did not handle in-kernel recovery with LDAP properly
4404 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4406 Severity : enhancement
4408 Description: bind OST threads to NUMA nodes to improve performance
4409 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4410 node and do their allocations there to localize memory access
4412 Severity : enhancement
4414 Description: llmount can determine client NID directly from Myrinet (GM)
4415 Details : the client NID code from gmnalnid was moved directly into
4416 llmount, removing the need to use this or specifying the
4417 client NID explicitly when mounting GM clients with zeroconf
4420 Frequency : if client is started with down MDS
4422 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4423 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4424 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4425 ptlrpc_import_delayed_req() because we didn't check if the
4426 request was interrupted, and we also didn't break out of the
4427 event loop if there was no timeout
4432 Description: data loss during non-page-aligned writes to a single file from
4433 both multiple nodes and multiple threads on one node at same time
4434 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4435 inconsistency led to false short-reads, that were cached and later
4436 used by ->prepare_write() to fill in partially written page,
4437 leading to data loss.
4440 Frequency : always, if lconf --abort_recovery used
4442 Description: lconf --abort_recovery fails with 'Operation not supported'
4443 Details : lconf was attempting to abort recovery on the MDT device and not
4446 Severity : enhancement
4448 Description: remove cleanup logs
4449 Details : replace lconf-generated cleanup logs with lustre internal
4450 cleanup routines. Eliminates the need for client-cleanup and
4453 Severity : enhancement
4455 Description: add support for EAs (user and system) on lustre filesystems
4456 Details : it is now possible to store extended attributes in the Lustre
4457 client filesystem, and with the user_xattr mount option it
4458 is possible to allow users to store EAs on their files also
4460 Severity : enhancement
4462 Description: Add possibility (config option) to show minimal available OST free
4464 Details : When compiled with --enable-mindf configure option, statfs(2)
4465 (and so, df) will return least minimal free space available from
4466 all OSTs as amount of free space on FS, instead of summary of
4467 free spaces of all OSTs.
4469 Severity : enhancement
4471 Description: do not expand extent locks acquired on OST-side
4472 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4473 by server: they are not cached anyway.
4476 Frequency : when mmap is used/binaries executed from Lustre
4478 Description: Unmmap pages before throwing them away from read cache.
4479 Details : llap_shrink cache now attempts to unmap pages before discarding
4480 them (if unmapping failed - do not discard). SLES9 kernel has
4481 extra checks that trigger if this unmapping is not done first.
4486 Description: lconf didn't resolve symlinks before checking to see whether a
4487 given mountpoint was already in use
4490 Frequency : when migrating failover services
4491 Bugzilla : 6395, 9514
4492 Description: When migrating a subset of services from a node (e.g. failback
4493 from a failover service node) the remaining services would
4494 time out and evict clients.
4495 Details : lconf --force (implied by --failover) sets the global obd_timeout
4496 to 5 seconds in order to quickly disconnect, but this caused
4497 other RPCs to time out too quickly. Do not change the global
4498 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4500 Severity : enhancement
4501 Frequency : if MDS is started with down OST
4502 Bugzilla : 9439,5706
4503 Description: Allow startup/shutdown of an MDS without depending on the
4504 availability of the OSTs.
4505 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4506 Add appropriate locking and lov-osc refcounts for safe
4507 cleaning. Add osc abort_inflight calls in case the
4508 synchronize never started.
4511 Frequency : occasional (Cray XT3 only)
4513 Description: root not authorized to access files in CRAY_PORTALS environment
4514 Details : The client process capabilities were not honoured on the MDS in
4515 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4516 been accessed by an authorized user then root was able to access
4517 the file on the local client also. The root user capabilities
4518 are now allowed on the MDS, as this environment has secure UID.
4521 Frequency : occasional
4523 Description: ldiskfs "too long searching" message happens too often
4524 Details : A debugging message (otherwise harmless) prints too often on
4525 the OST console. This has been reduced to only happen when
4526 there are fragmentation problems on the filesystem.
4531 Description: Division by zero in statfs when all OSCs are inactive
4532 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4533 lov_statfs divides by value returned by lov_get_stripecnt().
4537 Bugzilla : 9489, 3273
4538 Description: First write from each client to each OST was only 4kB in size,
4539 to initialize client writeback cache, which caused sub-optimal
4540 RPCs and poor layout on disk for the first writen file.
4541 Details : Clients now request an initial cache grant at (re)connect time
4542 and so that they can start streaming writes to the cache right
4543 away and always do full-sized RPCs if there is enough data.
4544 If the OST is rebooted the client also re-establishes its grant
4545 so that client cached writes will be honoured under the grant.
4550 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4551 Details : Now I/O RPCs go to different portal number and (presumably) fast
4552 lock requests (and glimses) and other RPCs get their own service
4553 threads pool that should be able to service those RPCs
4556 Severity : enhancement
4558 Description: Ability to exchange lustre version between client and servers and
4559 issue warnings at client side if client is too old. Also for
4560 liblustre clients there is ability to refuse connection of too old
4562 Details : New 'version' field is added to connect data structure that is
4563 filled with version info. That info is later checked by server and
4567 Frequency : rare, liblustre only.
4568 Bugzilla : 9296, 9581
4569 Description: Two simultaneous writes from liblustre at offset within same page
4570 might proceed at the same time overwriting eachother with stale
4572 Details : I/O lock withing llu_file_prwv was released too early, before data
4573 actually was hitting the wire. Extended lock-holding time until
4574 server acknowledges receiving data.
4577 Frequency : extremely rare. Never observed in practice.
4579 Description: avoid generating lustre_handle cookie of 0.
4580 Details : class_handle_hash() generates handle cookies by incrementing
4581 global counter, and can hit 0 occasionaly (this is unlikely, but
4582 not impossible, because initial value of cookie counter is
4583 selected randonly). Value of 0 is used as a sentinel meaning
4584 "unassigned handle" --- avoid it. Also coalesce two critical
4585 sections in this function into one.
4587 Severity : enhancement
4589 Description: allow liblustre clients to delegate truncate locking to OST
4590 Details : To avoid overhead of locking, liblustre client instructs OST to
4591 take extent lock in ost_punch() on client's behalf. New connection
4592 flag is added to handle backward compatibility.
4594 Severity : enhancement
4595 Bugzilla : 4928, 7341, 9758
4596 Description: allow number of OST service threads to be specified
4597 Details : a module parameter allows the number of OST service threads
4598 to be specified via "options ost ost_num_threads={N}" in the
4599 OSS's /etc/modules.conf or /etc/modprobe.conf.
4603 Bugzilla : 6146, 9635, 9895
4604 Description: servers crash with bad pointer in target_handle_connect()
4605 Details : In rare cases when a client is reconnecting it was possible that
4606 the connection request was the last reference for that export.
4607 We would temporarily drop the export reference and get a new
4608 one, but this may have been the last reference and the export
4609 was just destroyed. Get new reference before dropping old one.
4611 Severity : enhancement
4612 Frequency : if client is started with failover MDS
4614 Description: Allow multiple MDS hostnames in the mount command
4615 Details : Try to read the configuration from all specified MDS
4616 hostnames during a client mount in case the "primary"
4619 Severity : enhancement
4621 Description: Stop sending data to evicted clients as soon as possible.
4622 Details : Check if the client we are about to send or are sending data to
4623 was evicted already. (Check is done every second of waiting,
4624 for which l_wait_event interface was extended to allow checking
4625 of exit condition at specified intervals).
4628 Frequency : rare, normally only when NFS exporting is done from client
4630 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4632 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4635 Frequency : always, if srand() is called before liblustre initialization
4637 Description: Liblustre uses system PRNG disturbing its usage by user application
4638 Details : Introduce internal to lustre fast and high-quality PRNG for
4639 lustre usage and make liblustre and some other places in generic
4640 lustre code to use it.
4642 Severity : enhancement
4643 Bugzilla : 9477, 9557, 9870
4644 Description: Verify that the MDS configuration logs are updated when xml is
4645 Details : Check if the .xml configuration logs are newer than the config
4646 logs stored on the MDS and report an error if this is the case.
4647 Request --write-conf, or allow starting with --old_conf.
4649 Severity : enhancement
4651 Description: Handle symlinks in the path when checking if Lustre is mounted.
4652 Details : Resolve intermediate symlinks when checking if a client has
4653 mounted a filesystem to avoid duplicate client mounts.
4658 Description: lconf can hit an error exception but still return success.
4659 Details : The lconf command catches the Command error exception at the top
4660 level script context and will exit with the associated exit
4661 status, but doesn't ensure that this exit status is non-zero.
4666 Description: failure of ptlrpc thread startup can cause oops
4667 Details : Starting a ptlrpc service thread can fail if there are a large
4668 number of threads or the server memory is very fragmented.
4669 Handle this without oopsing.
4672 Frequency : always, only if liblustre and non-default acceptor port was used
4674 Description: liblustre cannot connect to servers with non-default acceptor port
4675 Details : tcpnal_set_default_params() was not called and was therefore
4676 ignoring the environment varaible TCPNAL_PORT, as well as other
4677 TCPNAL_ environment variables
4682 Description: two objects could be created on the same OST for a single file
4683 Details : If an OST is down, in some cases it was possible to create two
4684 objects on a single OST for a single file. No problems other
4685 than potential performance impact and spurious error messages.
4689 Bugzilla : 5681, 9562
4690 Description: Client may oops in ll_unhash_aliases
4691 Details : Client dcache may become inconsistent in race condition.
4692 In some cases "getcwd" can fail if the current directory is
4698 Description: Inode refcounting problems in NFS export code
4699 Details : link_raw functions used to call d_instantiate without obtaining
4700 extra inode reference first.
4704 Bugzilla : 9942, 9903
4705 Description: Referencing freed requests leading to crash, memleaks with NFS.
4706 Details : We used to require that call to ll_revalidate_it was always
4707 followed by ll_lookup_it. Also with revalidate_special() it is
4708 possible to call ll_revalidate_it() twice for the same dentry
4709 even if first occurence returned success. This fix changes semantic
4710 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4711 reference on a request referred from the intent.
4712 ll_intent_release() then releases such a request.
4715 Frequency : rare, normally benchmark loads only
4717 Description: unlinked inodes were kept in memory on the client
4718 Details : If a client is repeatedly creating and unlinking files it
4719 can accumulate a lot of stale inodes in the inode slab cache.
4720 If there is no other client load running this can cause the
4721 client node to run out of memory. Instead flush old inodes
4722 from client cache that have the same inode number as a new inode.
4725 Frequency : SLES9 2.6.5 kernel and long filenames only
4726 Bugzilla : 9969, 10379
4727 Description: utime reports stale NFS file handle
4728 Details : SLES9 uses out-of-dentry names in some cases, which confused
4729 the lustre dentry revalidation. Change it to always use the
4733 Frequency : rare, unless heavy write-truncate concurrency is continuous
4734 Bugzilla : 4180, 6984, 7171, 9963, 9331
4735 Description: OST becomes very slow and/or deadlocked during object unlink
4736 Details : filter_destroy() was holding onto the parent directory lock
4737 while truncating+unlinking objects. For very large objects this
4738 may block other threads for a long time and slow overall OST
4739 responsiveness. It may also be possible to get a lock ordering
4740 deadlock in this case, or run out of journal credits because of
4741 the combined truncate+unlink. Solution is to do object truncate
4742 first in one transaction without parent lock, and then do the
4743 final unlink in a new transaction with the parent lock. This
4744 reduces the lock hold time dramatically.
4747 Frequency : rare, 2.4 kernels only
4749 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4750 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4751 before unmounting their filesystem in order to avoid deadlock.
4752 The kernel_locked() function in 2.4 kernels only checks whether
4753 the kernel lock is held, not whether it is this process that is
4754 holding it as 2.6 kernels do.
4759 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4760 Details : The client ptlrpc code may be trying to reconnect to a down
4761 server before a previous connection attempt has timed out.
4762 Increase the reconnect interval to be longer than the connection
4763 timeout interval to avoid sending duplicate connections to
4767 Frequency : echo_client brw_test command
4769 Description: fix echo_client to work with OST preallocated code
4770 Details : OST preallocation code (5137) didn't take echo_client IO path
4771 into account: echo_client calls filter methods outside of any
4772 OST thread and, hence, there is no per-thread preallocated
4773 pages and buffers to use. Solution: hijack pga pages for IO. As
4774 a byproduct, this avoids unnecessary data copying.
4778 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4779 Description: Client can oops in mdc_commit_close() after open replay
4780 Details : It was possible for the MDS to return an open request with no
4781 transaction number in mds_finish_transno() if the client was
4782 evicted, but without actually returning an error. Clients
4783 would later try to replay that open and may trip an assertion
4784 Simplify the client close codepath, and always return an error
4785 from the MDS in case the open is not successful.
4788 Frequency : rare, 2.6 OSTs only
4790 Description: OST may deadlock under high load on fragmented files
4791 Details : If there was a heavy load and highly-fragmented OST filesystems
4792 it was possible to have all the OST threads deadlock waiting on
4793 allocation of biovecs, because the biovecs were not released
4794 until the entire RPC IO was completed. Instead, release biovecs
4795 as soon as they are complete to ensure forward IO progress.
4797 Severity : enhancement
4799 Description: Support for specifying external journal device at mount
4800 Details : If an OST or MDS device is formatted with an external journal
4801 device, this device major/minor is stored in the ext3 superblock
4802 and may not be valid for failover. Allow detecting and
4803 specifying the external journal at mount time.
4808 Description: Mounting an MDS with pending unlinked files may cause oops
4809 Details : target_finish_recovery() calls mds_postrecov() which returned
4810 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4811 considers this an error and immediately begins cleaning up the
4812 lov, just after starting the mds_lov process
4814 Severity : enhancement
4816 Description: Implement 'lfs df' to report actual free space on per-OST basis
4817 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4818 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4819 usage of objects; '-h' to report in human readable format.
4821 ------------------------------------------------------------------------------
4823 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4830 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4831 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4832 at startup during orphan unlinking. Instead, skip buddy update
4833 before setup, as it will be regenerated after recovery is complete.
4838 Description: If an OST is inactive, its locks might reference stale inodes.
4839 Details : lov_change_cbdata() must iterate over all namespaces, even if
4840 they are inactive to clear inode references from the lock.
4842 Severity : enhancement
4843 Frequency : occasional, if non-standard max_dirty_mb used
4845 Description: Client will block write RPCs if not enough grant
4846 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4847 then the client will block writes while waiting for another RPC
4848 to complete instead of consuming its dirty limit. With change
4849 we get improved performance when max_dirty_mb is small.
4851 Severity : enhancement
4852 Bugzilla : 3389, 6253
4853 Description: Add support for supplementary groups on the MDS.
4854 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4855 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4856 lookups for user supplementary groups into a cache.
4860 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4861 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4862 for liblustre O_EXCL is enforced by MDS.
4866 Description: readdir never returns NULL in liblustre.
4867 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4868 dentry in struct dirent.
4872 Frequency : liblustre only, depends on application IO pattern
4873 Description: liblustre clients evicted if not contacting servers
4874 Details : Don't put liblustre clients into the ping_evictor list, so
4875 they will not be evicted by the pinger ever.
4877 Severity : enhancement
4879 Description: Add ability to evict clients by NID from MDS.
4880 Details : By echoing "nid:$NID" string into
4881 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4882 $NID would be instantly evicted from this MDS and from all active
4883 OSTs connected to it.
4887 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4888 Details : lookup_it_finish() used to query file size from OSTs that was not
4893 Description: service threads change working directory to that of init
4894 Details : Starting lustre service threads may pin the working directory
4895 of the parent thread, making that filesystem busy. Threads
4896 now change to the working directory of init to avoid this.
4900 Frequency : during shutdown only
4901 Description: shutdown with a failed MDS or OST can cause unmount to hang
4902 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4907 Frequency : 2.6 only
4908 Description: chmod/chown may include an extra supplementary group
4909 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4910 supplementary group and if none is specified this is used.
4913 Bugzilla : 5479 (6816)
4915 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4916 Details : If lookup is in progress on a file that is unlinked we might try
4917 to revalidate the inode and fail in revalidate after lookup is
4918 complete and ll_file_open() enqueues the open again but
4919 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4922 Frequency : always, if lconf --abort_recovery used
4924 Description: lconf --abort_recovery fails with 'Operation not supported'
4925 Details : lconf was attempting to abort recovery on the MDT device and not
4928 ------------------------------------------------------------------------------
4930 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4935 Frequency : rare (only unsupported configurations with a node running as an
4937 Bugzilla : 6514, 5137
4938 Description: Mounting a Lustre file system on a node running as an OST could
4940 Details : OSTs now preallocates memory needed to write out data at
4941 startup, instead of when needed, to avoid having to
4942 allocate memory in possibly low memory situations.
4943 Specifically, if the file system is mounted on on OST,
4944 memory pressure could force it to try to write out data,
4945 which it needed to allocate memory to do. Due to the low
4946 memory, it would be unable to do so and the node would
4947 become unresponsive.
4949 Severity : enhancement
4951 Description: Addition of lconf --service command line option
4952 Details : lconf now accepts a '--service <arg>' option, which is
4953 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4955 Severity : enhancement
4957 Description: Failover mode is now the default for OSTs.
4958 Details : By default, OSTs will now run in failover mode. To return to
4959 the old behaviour, add '--failout' to the lmc line for OSTs.
4961 Severity : enhancement
4963 Description: Health checks are now provided for MDS and OSTs
4964 Details : Additional detailed health check information on MSD and OSTs
4965 is now provided through the procfs health_check value.
4968 Frequency : occasional, depends on IO load
4970 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4971 after numerous create/delete cycles
4972 Details : The ext3 inode allocation policy would not allocate new inodes
4973 very well on the OSTs because there are no new directories
4974 being created. Instead we look for groups with free space if
4975 the parent directories are nearly full.
4980 Description: Network or server problems during mount may cause partially
4981 mounted clients instead of returning an error.
4982 Details : The config llog parsing code may overwrite the error return
4983 code during mount error handling, returning success instead
4989 Description: MDS can fail to allocate large reply buffers
4990 Details : After long uptimes the MDS can fail to allocate large reply
4991 buffers (e.g. zconf client mount config records) due to memory
4992 fragmentation or consumption by the buffer cache. Preallocate
4993 some large reply buffers so that these replies can be sent even
4994 under memory pressure.
4998 Frequency : rare (liblustre)
4999 Description: fsx running with liblustre complained that using truncate() to
5000 extend the file doesn't work. This patch corrects that issue.
5001 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5002 ATTR_SIZE and lsm use in llu_setattr_raw.
5006 Frequency : rare, only 2.6 kernels
5007 Description: Unusual file access patterns on the MDS may result in inode
5008 data being lost in very rare circumstances.
5009 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5010 code in the 2.6 kernel caused the inode and/or EA data not to
5011 be read from disk, causing single-file corruption.
5015 Frequency : rare, only 2.6 filesystems using extents
5016 Description: Heavy concurrent write and delete load may cause data corruption.
5017 Details : It was possible under high-load situations to have an extent
5018 metadata block in the block device cache from a just-unlinked
5019 file overwrite a newly-allocated data block. We now unmap any
5020 metadata buffers that alias just-allocated data blocks.
5024 Frequency : filesystems with default stripe_count larger than 77
5025 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5026 Details : lconf specifies an inode size of 4096 bytes when the default
5027 stripe_count is larger than 77. This conflicts with the default
5028 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5029 this case to avoid pinning too much memory for large EAs.
5031 ------------------------------------------------------------------------------
5033 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5038 Frequency : rare (extremely heavy IO load with hundreds of clients)
5040 Description: Client is evicted, gets IO error writing to file
5041 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5042 caused clients to be evicted for AST timeouts. The fixes in
5043 bug 5192 mean we no longer need to have such short AST timeouts
5044 so ldlm_timeout has been increased.
5047 Frequency : occasional during --force or --failover shutdown under load
5048 Bugzilla : 5949, 4834
5049 Description: Server oops/LBUG if stopped with --force or --failover under load
5050 Details : a collection of import/export refcount and cleanup ordering
5051 issues fixed for safer force cleanup
5054 Frequency : only filesystems larger than 120 OSTs
5055 Bugzilla : 5990, 6223
5056 Description: lfs getstripe would oops on a very large filesystem
5057 Details : lov_getconfig used kfree on vmalloc'd memory
5060 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5061 Bugzilla : 6242, 6243
5062 Description: reading from files that had been truncated to a non-zero size
5063 but never opened returned no data
5064 Details : ll_file_read() reads zeros from no-object files to EOF
5069 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5070 Details : There's little harm in aborting MDS/OSS recovery and letting it
5071 try again, so I removed the LASSERT and return an error instead.
5073 Severity : enhancement
5075 Description: New debugging infrastructure for tracking down data corruption
5076 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5077 (b) cover more of the client-side code path, and (c) try to narrow
5078 down where problems occurred
5082 Bugzilla : 3819, 4364, 4397, 6313
5083 Description: Racing close and eviction MDS could cause assertion in mds_close
5084 Details : It was possible to get multiple mfd references during close and
5085 client eviction, leading to one thread referencing a freed mfd.
5087 Severity: : enhancement
5088 Bugzilla : 3262, 6359
5089 Description: Attempts to reconnect to servers are now more aggressive.
5090 Details : This builds on the enhanced upcall-less recovery that was added
5091 in 1.4.2. When trying to reconnect to servers, clients will
5092 now try each server in the failover group every 10 seconds. By
5093 default, clients would previously try one server every 25 seconds.
5098 Description: After recovery, certain operations trigger a failed
5099 assertion on a client.
5100 Details : Failing over an mds, using lconf -d --failover, while a
5101 client was doing a readdir() call would cause the client to
5102 LBUG after recovery completed and the readdir() was resent.
5104 Severity : enhancement
5106 Description: Default groups are now added by lconf
5107 Details : You can now run lconf --group <servicename> without having to
5108 manually add groups with lmc.
5111 Frequency : occasional
5113 Description: Nodes with an elan id of 0 trigger a failed assertion
5116 Frequency : always when accessing e.g. tty/console device nodes
5118 Description: tty and some other devices nodes cannot be used on lustre
5119 Details : file's private_data field is used by device data and lustre
5120 values in there got lost. New field was added to struct file to
5121 store fs-specific private data.
5124 Frequency : when exporting Lustre via NFS
5126 Description: NFSD failed occasionally when looking up a path component
5127 Details : NFSD is looking up ".." which was broken in ext3 directories
5128 that had grown large enough to become hashed.
5131 Frequency : Clusters with multiple interfaces not on the same subnet
5133 Description: Nodes will repeatedly try to reconnect to an interface which it
5134 cannot reach and report an error to the log.
5135 Details : Extra peer list entries will be created by lconf with some peers
5136 unreachable. lconf now validates the peer before adding it.
5139 Frequency : Only if a default stripe is set on the filesystem root.
5141 Description: Setting a default stripe on the filesystem root prevented the
5142 filesystem from being remounted.
5143 Details : The client was sending extra request flags in the root getattr
5144 request and did not allocate a reply buffer for the dir EA.
5147 Frequency : occasional, higher if lots of files are accessed by one client
5148 Bugzilla : 6159, 6097
5149 Description: Client trips assertion regarding lsm mismatch/magic
5150 Details : While revalidating inodes the VFS looks up inodes with ifind()
5151 and in rare cases can find an inode that is being freed.
5152 The ll_test_inode() code will free the lsm during ifind()
5153 when it finds an existing inode and then the VFS later attaches
5154 this free lsm to a new inode.
5158 Bugzilla : 6422, 7030
5159 Description: MDS deadlock between mkdir and client eviction
5160 Details : Creating a new file via mkdir or mknod (starting a transaction
5161 and getting the ns lock) can deadlock with client eviction
5162 (gets ns lock and trying to finish a synchronous transaction).
5165 Frequency : occasional
5166 Description: While starting a server, the fsfilt_ext3 module could not be
5168 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5169 kernels. Previously, lconf would still use the ext3 name
5170 when trying to load modules. Now, it will correctly use
5171 ext3 on 2.4 and ldiskfs on 2.6.
5173 Severity : enhancement
5174 Description: The default stripe count has been changed to 1
5175 Details : The interpretation of the default stripe count (0, to lfs
5176 or lmc) has been changed to mean striping across a single
5177 OST, rather than all available. For general usage we have
5178 found a stripe count of 1 or 2 works best.
5180 Severity : enhancement
5181 Description: Add support for compiling against Cray portals.
5182 Details : Conditional compiling for some areas that are different
5186 Frequency : occasional
5187 Bugzilla : 6409, 6834
5188 Description: Creating files with an explicit stripe count may lead to
5189 a failed assertion on the MDS
5190 Details : If some OSTs are full or unavailable, creating files may
5191 trigger a failed assertion on the MDS. Now, Lustre will
5192 try to use other servers or return an error to the
5196 Frequency : occasional
5198 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5199 caused lock timeout during readahead (since 1.4.2).
5200 Details : Processes doing readahead might match a lock that hasn't been
5201 granted yet if there are overlapping and conflicting lock
5202 requests. The readahead process waits on ungranted lock
5203 (original lock is CBPENDING), while OST waits for that process
5204 to cancel CBPENDING read lock and eventually evicts client.
5206 Severity : enhancement
5208 Description: Initial enabling of flock support for clients
5209 Details : Implements fcntl advisory locking and file status functions.
5210 This feature is provided as an optional mount flag (default
5211 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5212 locking are implemented yet, and those that are are not guaranteed
5213 to be completely correct in production environments.
5214 mount -t lustre -o [flock|noflock] ...
5217 Frequency : occasional
5219 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5220 assertion in the ext3 JBD (journaling) layer.
5221 Details : The b_committed_data struct is protected by the big kernel lock
5222 in 2.4 kernels, serializing journal_commit_transaction() and
5223 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5224 access to this struct. In 2.6 kernels there is finer grained
5225 locking to improve SMP performance of the JBD layer.
5229 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5234 Description: merge and cleanup kernel patches.
5237 -----------------------------------------------------------------------------
5239 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5241 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5242 versions of Lustre. Please update all servers and clients to
5243 version 1.4.2 or later at the same time. You must also run
5244 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5245 to update the configuration logs.
5247 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5248 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5249 - swab llog records of type '0' so we get proper header size/idx (5861)
5250 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5251 - fix rename of one directory over another leaking an inode (5953)
5252 - avoid SetPageDirty on 2.6 (5981)
5253 - don't re-add just-being-destroyed locks to the waiting list (5653)
5254 - when creating new directories, inherit the parent's custom
5255 striping settings if present parent (3048)
5256 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5257 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5258 - don't hold client locks on temporary worklist from l_lru (5666)
5259 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5260 - automatically evict dead clients (5921)
5261 - Update file size properly in create+truncate+fstat case (6196)
5262 - Do not unhash mountpoint dentries, do not allow removal of
5264 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5265 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5266 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5267 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5268 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5269 - Correctly handle reads of files with no objects (6243)
5270 - lctl recover will also mark a device active if deactivate used (5933)
5272 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5273 - allow --write-conf on an MDS with different nettype than client (5619)
5274 - don't write config llogs to MDS for mounts not from that MDS (5617)
5275 - lconf should create multiple TCP connections from a client (5201)
5276 - init scripts are now turned off by default; run chkconfig --on
5277 lustre and chkconfig --on lustrefs to use them
5278 - upcalls are no longer needed for clients to recover to failover
5280 - add --abort-recovery option to lconf to abort recovery on device
5282 - add support for an arbitrary number of OSTs (3026)
5283 - Quota support protocol changes.
5284 - forward compatibility changes to wire structs (6007)
5285 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5286 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5287 - improved "lustre status" script
5288 - initialize blocksize for non-regular files (6062)
5289 - added --disable-server and --disable-client configure options (5782)
5290 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5291 - Vanilla 2.4.29 support
5292 - increase maximum number of obd devices to 520 (6242)
5293 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5294 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5295 SLES 9 2.6.5 kernels
5296 - limit stripes per file to 160 (the maximum EA size) (6093)
5298 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5301 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5302 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5303 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5304 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5305 - small ext3 extents cleanups and fixes (5733)
5306 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5307 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5308 - avoid needless client->OST connect, fix handle mismatch (5317)
5309 - fix DLM error path that led to out-of-sync client, long delays (5779)
5310 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5311 - fix several locking issues related to i_size (5492,5624,5654,5672)
5312 - don't move pending lock onto export if it is already evicted (5683)
5313 - fix kernel oops when creating .foo in unlinked directory (5548)
5314 - fix deadlock in obdfilter statistics vs. object create (5811)
5315 - use time_{before,after} to avoid timer jiffies wrap (5882)
5316 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5317 - Do not leak request if server was not able to process it (5154)
5318 - If mds_open unable to find parent dir, make that negative lookup(5154)
5319 - don't create new directories with extent-mapping (5909, 5936)
5321 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5322 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5323 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5324 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5325 - support for RHEL4 kernel on i686 (5773)
5326 - provide error messages when incompatible logs are encountered (5898)
5328 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5329 * version 1.4.0.10 (1.4.1 release candidate 1)
5331 - don't keep a lock reference when lock is not granted (4238)
5332 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5333 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5334 - drop import inflight refcount on signal_completed_replay error (5255)
5335 - unlock page after async write error during send (3677)
5336 - handle missing objects in filter_preprw_read properly (5265)
5337 - no transno return for symlink open, don't save no-trasno open (3440)
5338 - don't try to complete elan receive that already failed (4012)
5339 - free RPC server reply state on error (5406)
5340 - clean up thread from ptlrpc_start_thread() on error (5160)
5341 - readahead could read extra page into cache that wasn't ejected (5388)
5342 - prevent races in class_attach/setup/cleanup/detach (5260)
5343 - don't dereference de->d_inode after l_dput of de (5458)
5344 - use "int" for stripe value returned from lock_to_stripe (5544)
5345 - mballoc allocation and error-checking fixes in 2.6 (5504)
5346 - block device patches to fix I/O request sizes in 2.6 (5482)
5347 - look up hostnames for IB nals (5602)
5348 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5349 - don't start multiple acceptors for the same port (5277)
5350 - fix incorrect LASSERT in mds_getattr_name (5635)
5351 - export a proc file for general "ping" checking (5628)
5352 - fix "lfs check" to not block when the MDS is down (5628)
5354 - service request history (4965)
5355 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5356 - create an "evict_client" /proc entry on OSTs, like the MDS has
5357 - fix mount usage message, return errors per mount(8) (5168)
5358 - change grep [] to grep "[]" in tests so they work in more UMLs
5359 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5360 - remove ancient LOV_MAGIC_V0 EA support (5047)
5361 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5362 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5363 - allow readahead allocations to fail when low on memory (5383)
5364 - mmap locking landed again, after considerable improvement (2828)
5365 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5367 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5370 - send OST transaction number in read/write reply to free req (4966)
5371 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5372 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5373 - the watchdog thread now runs as interruptible (5246)
5374 - flock/lockf fixes (but it's still disabled, pending 5135)
5375 - don't use EXT3 constants in llite code (5094)
5376 - memory shortage at startup could cause assertion (5176)
5378 - reorganization of lov code
5379 - single portals codebase
5381 - add extents/mballoc support (5025)
5382 - direct I/O reads in the obdfilter (4048)
5383 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5385 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5388 - send OST transaction number in read/write reply to free req (4966)
5389 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5390 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5391 - the watchdog thread now runs as interruptible (5246)
5392 - handle missing objects in filter_preprw_read properly (5265)
5393 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5394 - drop import inflight refcount on signal_completed_replay error (5255)
5395 - unlock page after async write error during send (3677)
5396 - return original error code on reconstructed replies (3761)
5397 - no transno return for symlink open, don't save no-trasno open (3440)
5399 - add pid to ldlm debugging output (4922)
5400 - bump the watchdog timeouts -- we can't handle 30sec yet
5401 - extra debugging for orphan dentry/inode bug (5259)
5403 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5406 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5407 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5408 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5409 - fix NULL dereference in /proc/sys/portals/routes (4827)
5410 - allow failed mdc_close() operations to be interrupted (4561)
5411 - stop precreate on OST before MDS would time out on it (4778)
5412 - don't send partial-page writes before EOF from client (4410)
5413 - discard client grant for sub-page writes on large-page clients (4520)
5414 - don't free dentries not owned by NFS code, check generation (4806)
5415 - fix lsm leak if mds_create_objects() fails (4801)
5416 - limit debug_daemon file size, always print CERROR messages (4789)
5417 - use transno after validating reply (3892)
5418 - process timed out requests if import state changes (3754)
5419 - update mtime on OST during writes, return in glimpse (4829)
5420 - add mkfsoptions to LDAP (4679)
5421 - use ->max_readahead method instead of zapping global ra (5039)
5422 - don't interrupt __l_wait_event() during strace
5424 - add software watchdogs to catch hung threads quickly (4941)
5425 - make lustrefs init script start after nfs is mounted
5426 - fix CWARN/ERROR duplication (4930)
5427 - return async write errors to application if possible (2248)
5428 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5429 - print NAL number in %x format (4645)
5430 - update barely-supported suse-2.4.21-171 series (4842)
5431 - support for sles 9 %post scripts
5432 - support for building 2.6 kernel-source packages
5433 - support for sles km_* packages
5435 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5438 - ignore -ENOENT errors in osc_destroy (3639)
5439 - notify osc create thread that OSC is being cleaned up (4600)
5440 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5441 - reconstruct ost_handle() like mds_handle() (4657)
5442 - create a new thread to do import eviction to avoid deadlock (3969)
5443 - let lconf resolve symlinked-to devices (4629)
5444 - don't unlink "objects" from directory with default EA (4554)
5445 - hold socknal file ref over connect in case target is down (4394)
5446 - allow more than 32000 subdirectories in a single directory (3244)
5447 - fix blocks count for O_DIRECT writes (3751)
5448 - OST returns ENOSPC from object create when no space left (4539)
5449 - don't send truncate RPC if file size isn't changing (4410)
5450 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5451 - bind to privileged port in socknal and tcpnal (3689)
5453 - rate limit CERROR/CWARN console message to avoid overload (4519)
5454 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5455 - basic mmap support (3918)
5456 - kernel patch series update from b1_4 (4711)
5458 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5461 - avoid crash during MDS cleanup with OST shut down (2775)
5462 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5463 - don't use bad inodes on the MDS (3744)
5464 - dynamic object preallocation to improve recovery speed (4236)
5465 - don't hold spinlock over lock dumping or change debug flags (4401)
5466 - don't zero obd_dev when it is force cleaned (3651)
5467 - print grants to console if they go negative (4431)
5468 - "lctl deactivate" will stop automatic recovery attempts (3406)
5469 - look for existing locks in ldlm_handle_enqueue() (3764)
5470 - don't resolve lock handle twice in recovery avoiding race (4401)
5471 - revalidate should check working dir is a directory (4134)
5473 - don't always mark "slow" obdfilter messages as errors (4418)
5475 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5478 - don't close LustreDB during write_conf until it is done (3860)
5479 - fix typo in lconf for_each_profile (3821)
5480 - allow dumping logs from multiple threads at one time (3820)
5481 - don't allow multiple threads in OSC recovery (3812)
5482 - fix debug_size parameters (3864)
5483 - fix mds_postrecov to initialize import for llog ctxt (3121)
5484 - replace config semaphore with spinlock (3306)
5485 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5486 - don't allow enqueue to complete on a destroyed export (3822)
5487 - down write_lock before checking llog header bitmap (3825)
5488 - recover from lock replay timeout (3764)
5489 - up llog sem before sending rpc (3652)
5490 - reduce ns lock hold times when setting kms (3267)
5491 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5492 - fix NULL deref and obd_dev leak on setup error (3312)
5493 - replace some LBUG about llog ops with error handling (3841)
5494 - don't match INVALID dentries from d_lookup and spin (3784)
5495 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5496 - fix invalid assertion in ptlrpc_set_wait (3880)
5498 - add libwrap support for the TCP acceptor (3996)
5499 - add /proc/sys/portals/routes for non-root route listing (3994)
5500 - allow setting MDS UUID in .xml (2580)
5501 - print the stack of a process that LBUGs (4228)
5503 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5506 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5507 - make sure to unset replay flag from failed open requests (3440)
5508 - if default stripe count is 0, use OST count for inode size (3636)
5509 - update parent mtime/ctime on client for create/unlink (2611)
5510 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5511 - free recovery state on server during a forced cleanup (3571)
5512 - unregister_reply for resent reqs (3063)
5513 - loop back devices mounting and status check on 2.6 (3563)
5514 - fix resource-creation race that can provoke i_size == 0 (3513)
5515 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5516 - more debugging for page-accounting assertion (3746)
5517 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5518 - avoid deadlock after precreation failure (3758)
5519 - fix race and lock order deadlock in orphan handling (3450, 3750)
5520 - add validity checks when grabbing inodes from l_ast_data (3599)
5522 - add /proc/.../recovery_status to obdfilter (3428)
5523 - lightweight CDEBUG infrastructure, debug daemon (3668)
5524 - change default OSC RPC parameters to be better on small clusters
5525 - turn off OST read cache for files smaller than 32MB
5526 - install man pages and include them in rpms (3100)
5527 - add new init script for (un)mounting lustre filesystems (2593)
5528 - run chkconfig in %post for init scripts (3701)
5529 - drop scimac NAL (unmaintained)
5531 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5534 - clean kiobufs before and after use (3485)
5535 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5536 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5537 - change init script to fail more clearly if not run as root (1528)
5538 - allow clients to reconnect during replay (1742)
5539 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5540 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5541 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5542 - fix cond_resched (3554)
5543 - only evict unfinished clients after recovery (3515)
5544 - allow bulk resend, prevent data loss (3570)
5545 - dynamic ptlrpc request buffer allocation (2102)
5546 - don't allow unlinking open directory if it isn't empty (2904)
5547 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5548 - remove extraneous obd dereference causing LASSERT failure (3334)
5549 - don't use get_cycles() when creating temp. files on the mds (3156)
5550 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5551 - handle EEXIST for set-stripe, set proper directory name (3336)
5553 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5554 - fix ksocknal_fmb_callback() error messages (2918)
5556 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5559 - don't copy lvb into (possibly NULL) reply on error (2983)
5560 - don't deref dentry after dput, don't free lvb on error (2922)
5561 - use the kms to determine writeback rpc length (2947)
5562 - increment oti_logcookies when osc is inactive (2948)
5563 - update client's i_blocks count via lvb messages (2543)
5564 - handle intent open/close of special files properly (1557)
5565 - mount MDS with errors=remount-ro, like obdfilter (2009)
5566 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5567 - don't use cancelling-locks' kms values (2947)
5568 - use highest lock extent for kms, not last one (2925)
5569 - don't dereference ERR_PTR() dentry in error handling path (3107)
5570 - fix thread race in portals_debug_dumplog() (3122)
5571 - create lprocfs device entries at setup instead of at attach (1519)
5572 - common AST error handler, don't evict client on completion race (3145)
5573 - zero nameidata in detach_mnt in 2.6 (3118)
5574 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5575 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5576 - initialize RPC timeout timer earlier for 2.6 (3219)
5577 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5578 - print nal/nid for unknown nid (3258)
5579 - additional checks for oscc recovery before doing precreate (3284)
5580 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5581 - don't crash in mdc_close for bad permissions on open (3285)
5582 - zero i_rdev for non-device files (3147)
5583 - clear page->private before handing to FS, better assertion (3119)
5584 - tune the read pipeline (3236)
5585 - fix incorrect decref of invalidated dentry (2350)
5586 - provide read-ahead stats and refine rpc in flight stats (3328)
5587 - don't hold journal transaction open across create RPC (3313)
5588 - update atime on MDS at close time (3265)
5589 - close LDAP connection when recovering to avoid server load (3315)
5590 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5591 - don't leak open file on MDS after open resend (3325)
5592 - serialize filter_precreate and filter_destroy_precreated (3329)
5593 - loop device shouldn't call sync_dev() for nul device (3092)
5594 - clear page cache after eviction (2766)
5595 - resynchronize MDS->OST in background (2824)
5596 - refuse to mount the same filesystem twice on same mountpoint (3394)
5597 - allow llmount to create routes for mounting behind routers (3320)
5598 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5599 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5600 - fix rare problem with rename on htree directories (3417)
5602 - allow default OST striping configuration per directory (1414)
5603 - fix compilation for qswnal for 2.6 kernels (3125)
5604 - increase maximum number of MDS request buffers for large systems
5605 - change liblustreapi to be useful for external progs like lfsck (3098)
5606 - increase local configuration timeout for slow disks (3353)
5607 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5609 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5612 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5613 - don't overwrite extent policy data in reply if lock was blocked (2901)
5614 - drop filter export grants atomically with removal from device (2663)
5615 - del obd_self_export from work_list in class_disconnect_exports (2908)
5616 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5617 - swab reply message in mdc_close, other PPC fixes (2464)
5618 - fix destroying of named logs (2325)
5619 - overwrite old logs when running lconf --write_conf (2264)
5620 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5621 - fix race in target_handle_connect (2898)
5622 - mds_reint_create() should take same inode create lock (2926)
5623 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5624 - don't close files for self_export to avoid uninitialized obd (2936)
5625 - allow MDS with the same name as client node (2939)
5626 - hold dentry reference for closed log files for unlink (2325)
5627 - reserve space for all logs during transactions (2059)
5628 - don't evict page beyond end of stripe extent (2925)
5629 - don't oops on a deleted current working directory (2399)
5630 - handle hard links to targets without a parent properly (2517)
5631 - don't dereference NULL lock when racing during eviction (2867)
5632 - don't grow lock extents when lots of conflicting locks (2919)
5634 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5637 - account for cache space usage on clients to avoid data loss (974)
5638 - lfsck support in lustre kernel code (2349)
5639 - reduce journal credits needed for BRW writes (2370)
5640 - orphan handling to avoid losing space on client/server crashes
5641 - ptlrpcd can be blocked, stopping ALL progress (2477)
5642 - use lock value blocks to assist in proper KMS, faster stat (1021)
5643 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5644 - recovery for initial connections (2355)
5645 - fixes for mds_cleanup_orphans (1934)
5646 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5647 - block all file creations until orphan recovery completes (1901)
5648 - client remove rq_connection from request struct (2423)
5649 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5650 - recovery timer race (2670)
5651 - mdc_close recovey bug (2532)
5652 - ptlrpc cleanup bug (2710)
5653 - mds timeout on local locks (2588)
5654 - namespace lock held during RPCs (2431)
5655 - handle interrupted sync write properly (2503)
5656 - don't try to handle a message that hasn't been replied to (2699)
5657 - client assert failure during cleanup after abort recovery (2701)
5658 - leak mdc device after failed mount (2712)
5659 - ptlrpc_check_set allows timedout requests to complete (2714)
5660 - wait for inflight reqs when ptlrpcd finishes (2710)
5661 - make sure unregistered services are removed from the srv_list
5662 - reset bulk XID's when resending them (caught by 1138 test)
5663 - unregister_bulk after timeout
5664 - fix lconf error (2694)
5665 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5666 - readahead locks pages, leaves pending causing memory pressure (2673)
5667 - increase OST request buffers to 4096 on large machines (2729)
5668 - fix up permission of existing directories in simple_mkdir (2661)
5669 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5670 - don't assign transno to errored transactions (2742)
5671 - don't delete objects on OST if given a bogus objid from MDS (2751)
5672 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5673 - if rq_no_resend, then timeout request after recovery (2432)
5674 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5675 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5676 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5677 - MDS recovery shouldn't delete objects that it has given out (2730)
5678 - if enqueue arrives after completion, don't clobber LVB (2819)
5679 - don't unlock pages twice when trigger_group_io returns error (2814)
5680 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5681 - don't write pages to disk if there was an error (1450)
5682 - don't ping imports that have recovery disabled (2676)
5683 - take buffered bytes into account when balancing socknal conn (2817)
5684 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5685 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5686 - remove little-endian swabbing of llog records (1987)
5687 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5688 - retry reposting request buffers if they fail (1191)
5689 - grow extent at grant time to avoid granting a revoked lock (2809)
5690 - lock revoke doesn't evict page if covered by a second lock (2765)
5691 - disable VM readahead to avoid reading outside lock extents (2805)
5693 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5694 - updated kernel patches for hp-2.4.20 kernel (2681)
5696 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5699 - fix truncated write corruption (2366)
5700 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5702 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5703 - ensure len(uuid) < 37 in lmc (1171)
5704 - fix ia64 OOPS in llog_test (2255)
5705 - zero end of page at obdfilter for partial page writes (2648)
5706 - don't leave stale dentries around after renames (bug 2428)
5707 - fix timeouts when evicting a client with a single lock held (2642)
5708 - set deadline for the initial HELLO message to drain (2634)
5709 - print out dotted-quad IP addresses in the socknal (2302)
5711 - additional debugging for MDS client eviction problem (2443)
5712 - fix mkfsoptions support for osts (2603, 2604)
5714 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5717 - add series for the vanilla 2.6.0 kernel
5718 - add series for the vanilla 2.4.24 kernel
5719 - add series for a cray x86/64 UL kernel drop
5720 - fix xattr patches for the vanilla 2.4.19 series
5722 - generate true UUIDs in lmc (1171)
5723 - have portals stack dumping break in UML (2466)
5724 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5725 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5726 - wake callback waiting threads on client eviction (2460)
5727 - Add --ptldebug and --subsystem to lmc (1719)
5728 - update assertion to allow safe interrupt allocation
5729 - set rq_no_resend for cancel requests (2432)
5730 - recalculate ptlrpcd timeout after resend (2494)
5731 - call vfs_rmdir when removing pending directories (2368)
5732 - fix renaming a file to itself (2429)
5733 - lmc creates a default one-stripe lov (2454)
5734 - expand procfs space to handle large clusters (2326)
5735 - increase UML stack to avoid overflow
5736 - update lconf's list of debug and subsystem masks
5737 - fix lfs find --obd (2510)
5738 - /proc tunable for disabling filter read caching (2591)
5739 - stop rpm packages from altering slapd.conf (2301)
5740 - disable nagle in the socknal under 0conf (2578)
5741 - choose mds inode size based on stripe count (2572)
5742 - fix kernel-source rpm problems (2516)
5744 - add --disable-doc to avoid pdf generation (2421)
5745 - update documentation, tests, type-os, comments
5746 - avoid format warnings on ia64
5747 - remove the TOE NAL
5748 - tiny code cleanups by removing unused fields
5750 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5753 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5754 - remove the most blatant lies from BUILDING file (2371)
5755 - change default debug level to reasonable production setting
5756 - reduce client side cache size to reduce cache flush time
5757 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5758 - make TCP zerocopy and pinger support enabled by default (2476)
5759 - sync writes completed after process exits caused crashes (2319)
5760 - maintain correct mount count on the MDS (2356)
5761 - backout 1557, because 2316 wasn't really fixed
5762 - better file I/O statistics gathering in /proc
5763 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5764 - another kernel patch to fix zero-copy TCP function export
5765 - don't take duplicate lock when processing re-sent getattr (2420)
5766 - lctl uses obd_self_export instead of creating new conn (2353)
5767 - MDS/OST recovery case which requires object creation asserted (2425)
5768 - move lfs from /usr/sbin to /usr/bin in packages
5769 - fix race between mds_client_add and mds_client_free (2417)
5770 - use kmalloc instead of slabs in portals (2430)
5771 - don't create duplicate records when a failover MDS is present (2442)
5772 - remove unnecessary mount age check (2332)
5773 - don't remove directory inodes from locks prematurely (2451)
5774 - don't break if MDS service name is the same as hostname (2103)
5775 - fix races in client write RPC generation when cache full (2482)
5777 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5780 - remove now-unused request->rq_obd (278)
5781 - if an allocation fails, print out how much memory we've used (1933)
5782 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5783 - add the "configurable stack size" patch to most series files (1256)
5784 - ability to write large log records, for 100+ OST configs (2306)
5785 - fix NULL deref when filter_prep fails (2314)
5786 - fix operator precedence error in filter_sync
5787 - dynamic allocation of socknal TX descriptors (2315)
5788 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5789 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5790 - ensure that CWARN messages in Portals always get to the syslog
5791 - __init/__exit are not for prototype decls (ldlm_init/exit)
5792 - x86-64 compile warning fixes
5793 - fix gateway LMC keyword conflict (2318)
5794 - fix MDS lock inversions in getattr/reint paths (1844)
5795 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5796 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5797 - DLM race condition prevented some lock evictions (2328)
5798 - ENOMEM detection and retry on socknal sends (2230)
5799 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5800 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5801 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5802 - better error messages when a client is rejected during recovery (1505)
5803 - avoid cancelling locks which were never granted, after failure (2330)
5804 - fix i_sem/journal inversion in mds_client_add (2333)
5805 - fix truncate/getattr lock cycle deadlock (2334)
5806 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5807 - fix two rare exit paths which could leak an l_lock() ref (2321)
5808 - fencepost error in MDS/OST orphan recovery (2226)
5809 - make log record alignment 8 bytes (1988)
5810 - lstripe now fails when requested offset > ost_count (2237)
5811 - ensure that all kernel series have a complete list.h (1607)
5812 - fix crashes in special-file operations (2316)
5813 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5815 - allow configurable automake binary, for testing new versions
5816 - small update to the lfs documentation
5818 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5820 * fix negative export reference count in fsfilt_sync (2312)
5822 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5823 * release candidate 0.9.1
5825 - orphans are moved into the PENDING directory for possible recovery
5826 - replayed opens now open by fid for orphan/rename safety (1042)
5827 - last close of an orphan inode generates a transno (683)
5828 - chdir() and mount() now pin the directory entry (1020)
5829 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5830 - discard very old requests without processing them (1502)
5831 - remove some common, well-understood CERRORs (1505)
5832 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5833 - clear "grant" flags in OST replies until OST grant code lands (1644)
5834 - fix read performance by not clobbering i_blksize on client (1598)
5835 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5836 - make LRU size a /proc tunable, clears locks when reduced (707)
5837 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5838 - server locks take references on exports now (1558)
5839 - build fixes for 2.4.20-rh trees (1663)
5840 - return an error from lov_create if all OSCs are inactive (1751)
5841 - fix import levels when a reconnect happens without a timeout (1597)
5842 - exit early from mds_open if we get a lookup error (1749)
5843 - partial page read at EOF wouldn't wait for disk before sending (1642)
5844 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5845 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5846 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5847 - write/truncate lock inversion (1639)
5848 - Don't auto-load obdclass, portals modules during cleanup (1495)
5849 - fix timestamps from jumping to "now" (1763)
5850 - extra journal assertions (1648)
5851 - add an extra multiunlink test (1771)
5852 - fix read_record/write_record API (1776)
5853 - fix leak of offset_extent, possible incorrect i_size later (1772)
5854 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5855 - Debugging for the kqswnal_get_idle_tx problems (1820)
5856 - Allow recovery to be attempted multiple times (1536)
5857 - Write out MDS last_rcvd file after it is first created (1600)
5858 - Fix tx_descriptor leak in failed transmit situations (1827)
5859 - ext3 journaling fixes for assertion failure after IO error (1871)
5860 - class_export_put() on freed export after completion AST error (1896)
5861 - Fix revalidate looping in VFS (1322)
5862 - Don't access a freed export during MDS_REINT timeout (1521)
5863 - Add open-unlink recovery support on the MDS (1673,1764)
5864 - Return an error if no MDS data was read from last_rcvd (1946)
5865 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5866 - Don't setup a disk device that doesn't match exported UUID (317)
5867 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5868 - avoid committing NULL handle in force close
5869 - local.sh is now a one-stripe LOV configuration
5870 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5871 - fix ext3 htree duplicate directory entry corruption (1516)
5872 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5873 - update to new LOV EA format (2097)
5874 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5876 2003-06-15 Phil Schwan <phil@clusterfs.com>
5879 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5880 - per-import/export recovery handling (958, 931, 959)
5881 - multiple last-rcvd slots, for serving multiple FSes (949)
5882 - connections are again shared between multiple imp/exports (963, 964)
5883 - "umount -f" would hang if any requests needed to be sent (393, 978)
5884 - avoid pinning large req buffer by copying for queued messages (989)
5885 - add "uuid" to "lctl device" command to help upcalls (991)
5886 - "open" RPCs with transnos would confuse recovery counters (1037)
5887 - do proper endian conversion of all wire messages (288, 340, 891)
5888 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5889 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5890 - fix ptlrpc_connection leak in target_handle_connect (1174)
5891 - fix import refcounting bug in OST and MDS cleanup (1134)
5892 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5893 - fix very unlikely obd_types race condition (501)
5894 - remove osc_open hack for echo_client (1187)
5895 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5896 - a failure in read_inode2 leads to deadlock (1139)
5897 - cancel ack-locks as soon as transaction is committed (1072)
5898 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5899 - make sure to commitrw after any preprw to avoid deadlock (1162)
5900 - failing to execute a file in a lustre FS would lock inode (1203)
5901 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5902 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5903 - obd_uuid2tgt would walk off the end of the list (1255)
5904 - on IA64 the osc would give portals incorrect bulk size (1258)
5905 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5906 - fix lock inversion caused by new llite matching code (1282)
5907 - limit the number of dirty pages on a client to 10MB (1286)
5908 - timed out locks were not being corrected cancelled (1289)
5909 - fix O_DIRECT above 4GB on IA-32 (1292)
5910 * major user-visible changes
5911 - fail out/fail over policy now controlled by the upcall (993)
5913 - add OBD_PING to check server availability and failure (954)
5914 - lustre messages are now sent in sending host order (288, 340, 891)
5915 - add eadatalen to MDS getattr reply (340)
5916 - OST read replies may contain second buffer, with per-page status (593)
5918 2003-03-11 Phil Schwan <phil@clusterfs.com>
5921 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5922 - failed open()s could cause deadlock; fixed (867, 869)
5923 - stop cancelling OST locks when files are closed (481)
5924 - overlapping XID spaces caused network corruption (851, 853)
5925 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5926 - setattr_raw added, to do single-RPC, server-side setattrs
5927 - lmc/lconf syntax change for OST UUIDs
5928 - fix crashy race condition between ptlrpc_free_req and osc_close
5929 - don't use request in mdc_enqueue if we hit a timeout (889)
5930 - don't set the inode i_size for regular files from the MDS (896)
5931 - handle out of order completion AST (842)
5932 - don't LBUG if a lock request times out after receiving AST (913)
5933 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5934 - if a bad lock AST arrives, send an error instead of dropping entirely
5935 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5936 - fix leak in bulk IO when only partially completed (899, 900, 926)
5937 - fix O_DIRECT for ia64 (55)
5938 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5939 - C-z after timeout could hang a process forever; fixed (977)
5941 - client-side I/O cache (678, 924, 929, 941, 970)
5943 - READPAGE and SETATTRs which don't take server-side locks get
5946 2003-02-11 Phil Schwan <phil@clusterfs.com>
5949 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5950 - Fix fsfilt_extN_readpage() to read a full page of directory
5951 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5952 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5953 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5954 - Most intent operations were removed, in favour of a new RPC mode
5955 that does a single RPC to the server and bypasses most of the VFS
5956 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5957 - Aggressively cancel local locks on DLM servers
5958 - mds_reint_unlink sends EA to the client if it's the last nlink.
5959 client uses that EA to unlink OST objects.
5960 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5961 - recursive symlinks were fixed (439)
5962 - fixed NULL deref in DEBUG_REQ
5963 - filter_update_lastobjid no longer calls sync, which annoyed extN
5964 - fixed multi-client small-writes to a single file problem (445)
5965 - fixed mtime updates during file writes (607)
5966 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5967 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5968 - fixed Linux OST size reporting problem (444, 656)
5969 - OST now updates object mtime with writes or setattr (607, 619)
5970 - client verifies file size before zeroing page past EOF (445)
5971 - OST now writes last allocated objid to disk with allocation (108)
5972 - LOV on echo now works (409)
5974 - mds_reint_unlink sends a new buffer, with the EA included. this
5975 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5976 set if a regular file was being unlinked, and it was the last link
5977 - use PtlGet from the target for bulk writes (315)
5978 - OST now updates object mtime with writes or setattr (607, 619)
5979 - LDLM now has a grant-time callback to revalidate locked items, if
5981 - Many MDS operations were reorganized to combat race conditions
5983 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5984 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5986 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5989 - Fully reactivate OST imports after reconnection (512, others)
5990 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5991 - More graceful error handling for truncating on dead OST (515)
5992 - Don't error out unless we're actually accessing dead stripes (474)
5993 - Fix garbage sizes when stripes are missing (410)
5994 - LRU counters were broken, causing constant lock purge (433, 432)
5995 - garbage on read from stripes with failed OSTs (441)
5996 - mark OSCs as active before reconnecting during recovery (438)
5997 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5998 - lfind did not preserve OST order in output (443)
5999 - symlinks cause hung clients, incorrect data (439)
6000 - stop dereferencing request after dropping refcount (457)
6001 - don't LASSERT(spin_is_locked) on non-SMP (455)
6002 - fixes for many rename() bugs
6003 - fstat didn't correctly synchronize attributes (399)
6004 - server must handle lock cancellation during blocking AST prep (487)
6005 - bulk descriptors were free()d too soon (511)
6006 - fix paths in lconf, which would load incorrect modules (451, 507)
6007 - fix confusing lconf 'host not found' error message (386)
6008 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6009 - fix race condition in mdc_blocking_ast() for inode access (526)
6010 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6011 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6012 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6013 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6014 - fix lstripe in several ways (kernel side) (527)
6015 - fix request leak in ldlm_cli_enqueue (262)
6016 - incorrect OSC was marked inactive after OST failure
6017 - call mds_fs_cleanup before unmounting filesystem (524)
6018 - fix races between taking ns_lock and ldlm_lock_change_resource
6019 - fix races updating LOV export open file list
6020 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6021 - fix recovery NULL deref in ldlm_cli_cancel_unused
6022 - fix some DLM races by using new hash table for lock handles (419)
6023 - permit the client to specify desired inodes, at replay
6024 - duplicate requests when we queue them for replay reintegration
6025 - fix last_rcvd offset calculation
6026 - sync after each recovered transaction, so we always make progress
6027 - never, not always, ERESTART requests without transnos
6028 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6030 - skip replay if the MDS says that the client is already connected
6031 - don't check for a recovery-enabled export to match lctl's UUID
6032 - don't INC_USE_COUNT for phantom exports
6033 - don't crash when cleaning up phantom exports (567)
6034 - don't double-finish or set replay data for errored mdc_open requests
6035 - abort requests when they time out, so we don't get old replies
6036 - send/receive replies for AST messages again
6037 - if the client says that it doesn't have the lock, cancel it on the
6039 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6040 mark it as destroyed, it will all work out when decref is called
6041 - fix module use counts (22, 581)
6043 - ASTs now expect a reply (server cancels lock on error reply)
6045 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6048 - fix many simultaneous client startup (392)
6049 - fix dentry->d_it clobbering
6050 - credentials weren't being shipped for readdir/getattr operations
6051 - remove invalid assertions triggered during some concurrent MD
6053 - proper Lustre versions added (336, 389)
6054 - fix memory leak for create error case (398)
6055 - fix LOV locking bug that would get cli/srv out of sync
6056 - fix echo client over LOV (409)
6057 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6058 - fix double-O_EXCL intent crash (424)
6059 - avoid sending multiple lock CANCELs (352)
6061 - MDS can do multi-client recovery (modulo bugs in new code)
6063 - many updates, edits, cleanups
6065 2002-11-18 Phil Schwan <phil@clusterfs.com>
6068 - fix null d_it dereference (346)
6069 - fix full OST/dbench hang (333)
6070 - fix permission problem with file removal (286)
6071 - fix removal of OSCs from LOV when they fail
6072 - fix NULL deref during bulk timeout (214)
6073 - fix problems related to multiple filesystems on one MDS (241)
6074 - fixed serious subtle metadata locking bugs
6075 - free locks on clients when inodes are removed due to memory
6077 - fix inode pointer in lock data (285)
6078 - partial support for multiple MDS on a single host (241)
6079 - data locks weren't cancelled at clear_inode time (290, 311)
6080 - intent locks could lead to unbounded lock growth (205)
6081 - added a maximum lock count, an LRU list, and a flusher
6082 - fix multiple rename (365)
6083 - properly abstracted the echo client
6084 - OSC locked 1 byte too many; fixed
6085 - rewrote brw callback code:
6086 - fixed recovery bugs related to LOVs (306)
6087 - fixed too-many-pages-in-one-write crash (191)
6088 - fixed (again) crash in sync_io_timeout (214)
6089 - probably fixed callback-related race (385)
6091 - Add capability to MDS protocol
6092 - LDLM cancellations and callbacks on different portals
6094 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6097 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6098 - defer open object destruction to close time (601981, 138)
6099 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6100 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6101 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6102 - don't call obd_setattr in ll_file_release for destroyed objects
6103 * protocol change to lustre_msg: move |version| and add |flags|
6104 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6105 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6106 * added replay of create, unlink, link and rename operations during
6107 MDS failover; recovery should be much more robust now
6108 * remove failed OSCs from LOVs (only lov_create uses this so far)
6109 * the lustre-HOWTO was brought (more) up to date (582544)
6111 2002-10-23 Phil Schwan <phil@clusterfs.com>
6114 - in-use dentries weren't being reused properly (617851)
6115 - prevent multiple LDLM setup (599178)
6116 - fix LOV size calculations for truncate (617853)
6117 - fix client handling of MDS intent errors (POSIX)
6118 - fix permission bug in lovstripe.c test (624321)
6119 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6120 - truncate past end of file could corrupt data
6121 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6122 - a race in recovery could return ETIMEDOUT to apps (623947)
6123 - building outside the source directory was fixed
6124 * the lustre-HOWTO was brought (more) up to date (582544)
6125 * major progress was made on recovery functionality
6127 2002-10-10 Phil Schwan <phil@clusterfs.com>
6130 - recovery deadlock fix
6131 - rm -rf causes LBUG fix (617817)
6132 - file open by multiple tasks fix (618962)
6133 - directory permissions bugs (602707 and 620007)
6134 - journal_stop fixed with locking (611313)
6135 - O_APPEND failures resolved (618273, perhaps 614459)
6136 - lconf PATH fix (619770)
6137 - IA64 build fix (621450)
6138 - RPC buffer sizes scale with amount of memory
6140 2002-10-01 Phil Schwan <phil@clusterfs.com>
6143 - locks would be cancelled without throwing away data pages,
6144 resulting in inconsistent data (605627)
6145 - inode attributes were not always being refreshed (605627, 612449)
6146 - lconf now continues to cleanup after lctl reports an error
6147 - MDS now enforces user permissions (602707)
6148 - lprocfs cleanup fixed, but not yet enabled (614157)
6149 - fixed infinite server hang, should a client not respond to an AST
6150 - avoid going into recovery if user calls readlink() with a buffer
6151 that's too small (613941)
6152 - AST RPCs no longer require replies (614867) -- this may be changed
6153 - don't crash server if client sends an IOV that's too big (611336)
6154 - fixed lock conversion deadlock (611892)
6155 - fixed the following of symlinks (614622)
6156 * recovery: the server can remove locks from a client that dies, other
6157 clients can make progress
6158 * more extN patch fixes
6159 * compile-time configurable ptlrpc buffer allocations
6161 - collaborative read cache document
6162 - Lustre Lite Performance CDR document-in-progress
6164 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6167 - fix typo in patch-2.4.18
6169 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6172 - clear ptlrpc request each time in handle_incoming_request()
6173 - unlink of files now destroys the object on the OST
6175 2002-09-19 Peter Braam <braam@clusterfs.com>
6177 * add hard link support
6178 * change obdfile creation method
6179 * kernel patch changed
6181 2002-09-19 Peter Braam <braam@clusterfs.com>
6184 - stack overflow bug in extN fixed
6186 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6188 * documentation updates
6189 - add man pages for config tools
6190 - update tests/README to describe testing with new config tools
6191 - finish metadata API descriptions
6192 * bug fixes and cleanups
6193 - statfs workaround for 16TB limit
6194 - LOV stripe allocation improved, can stripe on subset of OSTs
6195 - LOV file size/IO offset was wrong for files > 4GB in size
6196 - object EA data was being dropped, caused files to be unreadable
6197 - memory overflow with non-LOV OST caused memory corruption
6198 - fixed regression tests to work with new config tools, obdfilter
6199 - fixed bug when directory size became larger than 1 block
6200 - fixed bug (for single client case) when PWD was deleted
6201 - invalidate local directory pages when doing intent-based ops
6202 - avoid LDLM oops when lock callback contained bad data
6204 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6206 * documentation updates
6207 * bug fixes and cleanups
6208 - configuration tools
6211 - 64-bit compile warnings
6212 - 64-bit internal statfs data
6214 * test_brw on persistent OST devices
6216 * lprocfs (disabled)
6218 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6220 * documentation updates
6221 * bug fixes and cleanups
6222 * configuration tools
6224 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6227 * many small fixes to 0_5_4
6228 * io/network handling
6229 * thinkos in MDS operations
6231 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6234 * crucial basic fixes to 0.5.3
6235 * IOR, Iozone work over Elan
6238 2002-08-07 Phil Schwan <phil@clusterfs.com>
6239 * version 0_5_3, our first alpha
6240 * we use the new Portals iovs
6241 * documentation updates
6242 * bug fixes and cleanups
6243 * small changes in the DLM wire protocol
6245 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6246 * version 0_5_1 with some initial stability,
6247 * locking on MD and file I/O.
6248 * documentation updates
6249 * several bug fixes since 0.5.0
6250 * small changes in wire protocol
6252 2002-07-18 Phil Schwan <phil@clusterfs.com>
6254 * delivered as Lustre Light Alpha
6255 * fixed a crash after handling invalid MDS requests
6256 * fixed directory pages for architectures with non-4k pages sizes
6258 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6259 * release version v0_4_4
6260 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6262 * Fixes a number of interruption problems with OST operations.
6263 * Update documentation for portals header changes
6264 * Move all wire protocol structs/defines to lustre_idl.h
6265 * Fixes symlink length bug.
6266 * Add tcpdump to repository.
6268 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6269 * release version v0_4_3
6270 * Fixes statfs for inodes on extN.
6271 * Fixes bug in runtests which would delete /etc/hosts.
6272 * Use 64-bit object IDs wherever possible (not into VFS though)
6273 Remove ost_get_info, which is unused by lustre, and out of date.
6275 2002-07-03 Peter Braam <braam@clusterfs.com>
6276 * release version v0_4_2 Fixes a lookup error (type not passed)
6277 * move forward to head of Portals
6278 * move forward to latest Lustre kernel
6280 2002-06-25 Peter Braam <braam@clusterfs.com>
6281 * release version v0_4_1. Hopefully stable on single node use.