1 tbd Sun Microsystems, Inc.
3 * Server support for kernels:
4 2.6.18-164.6.1.el5 (RHEL 5)
5 2.6.18-164.6.1.0.1.el5 (OEL 5)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-164.6.1.el5 (RHEL 5),
9 2.6.18-164.6.1.0.1.el5 (OEL 5)
10 2.6.16.60-0.42.8 (SLES 10),
12 * Recommended e2fsprogs version: 1.41.6.sun1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 Description: Add cascading_rw to lustre mpi tests.
21 Description: Store SOM data in EA.
22 Details : Also several SOM related cleanups.
26 Description: Add refcount to md_open_data
30 Description: Use refcount on lov_request_set to prevent memory corruption.
34 Description: Statahead: use parent dir's i_mutex for synchronization.
38 Description: loadgen improvements.
42 Description: Enable peer health detection in ksocklnd and ko2iblnd.
46 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
50 Description: Fix race on obd_devcie:md_stats
54 Description: Fix SLES 11 client builds.
55 Details : Resolved uninitalized pointer gcc warnings.
59 Description: Allow non-root access for "lfs check".
60 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
64 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
66 Severity : enhancement
68 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
69 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
73 Description: Sanity test_27p fails.
74 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
75 Sanity tests 27 were cleaned up and improved.
79 Description: Unify req format on client/servers
80 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
81 buffers and swabbers approach..
86 Description: drop join file stuff.
88 Severity : enhancement
90 Description: Add 2.6.27 and SLES11 (patchless client) support
94 Description: Fix in ptlrpc_expire_one_request() to print the signed time
97 Severity : enhancement
99 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
102 Severity : enhancement
104 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
106 Severity : enhancement
108 Description: correctly handle big reply message.
109 Details : send LNet event if reply is bigger then buffer and adjust this buffer
114 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
116 Severity : enhancement
118 Description: Add support for OEL5.
122 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
126 Description: client could not reconnect to OST because of an active request.
127 Details : The client now retries to reconnect to the same server, if a
128 connect request failed with EBUSY or -EAGAIN. Test case 35b
129 conf-sanity.sh has been created.
133 Description: Don't allow make backward step on assignin osc next id.
134 Details : race between allocation next id and ll_sync thread can be cause
135 of set wrong osc next id and can be kill valid ost objects.
139 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
143 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
147 Description: Error handling in osc_statfs_interpret() has been improved.
148 Details : Check in osc_statfs_interpret() for EBADR.
152 Description: Do not send statfs() requests to OSTs disabled by administrator.
153 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
155 Severity : enhancement
157 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
159 Severity : enhancement
161 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
163 Severity : enhancement
165 Description: Enhancement for lfs(1) command to use numeric uid/gid.
169 Description: Improved cleanup procedure for conf-sanity test_46a.
173 Description: Do not start bulk transfer, if client tried to reconnect.
174 Details : Do not start bulk transfer, if client tried to reconnect,
175 but failed due to active rpcs. Check for exp_abort_active_req
176 has been implemented.
180 Description: Prevent inconsistences between linux and lustre mount structures.
181 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
182 Make the sleep interruptible.
184 Severity : enhancement
186 Description: provide server to client comms path
187 Details : server to client communications path via new LDLM_SET_INFO rpc
191 Description: truncate starts GFP_FS allocation under transaction causing deadlock
192 Details : ldiskfs_truncate calls grab_cache_page which may start page
193 allocation under an open transaction. This may lead to
194 calling prune_icache with consequent lustre reentrance.
198 Description: send correctly lsm on open replay
199 Details : MDS is trust to LSM size on replay open, but client can set wrong size
204 Description: enable client interface failover
205 Details : When a child reconnects from another NID, properly update export
206 nid hash position and ldlm reverse import.
208 Severity : enhancement
210 Description: Communicate OST degraded/readonly state via statfs to MDS
211 Details : Flags in the statfs returned from OSTs indicate whether the
212 OST is in a degraded RAID state, or if the filesystem has
213 turned read-only after a filesystem error is detected.
218 Description: don't panic if EPROTO was hit when reading symlink
219 Details : correctly handling request reference in error cases.
221 Severity : enhancement
223 Description: Add LustreNetLink, a kernel-userspace communcation path.
225 Severity : enhancement
227 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
232 Description: access to llog context before init.
233 Details : move handling CATALOGS file at osc layer and forbid access to llog
237 Frequency : rare, only when client is writing to OST on local node
239 Description: Avoid deadlock for local client writes
240 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
241 memory freeing context. This allows OST threads to set the
242 PF_MEMALLOC flag on task structures in order to allocate memory
243 from reserved pools and complete IO.
244 Use GFP_HIGHUSER for OST allocations for non-local client writes,
245 so that the OST threads generate memory pressure and allow
246 inactive pages to be reclaimed.
248 Severity : enhancement
250 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
255 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
256 Details : move ldlm namespace creation in setup phase to avoid grab
257 _lprocfs_lock with cli_sem held.
261 Description: Temporarily disable grant shrink.
262 Details : Disable the feature for debugging.
265 Frequency : only when formatting test filesystems
267 Description: Unable to run several mkfs.lustre on loop devices at the same time
268 Details : mkfs.lustre returns error 256 on the concurrent loop devices
269 formatting. The solution is to proper handle the error.
271 Severity : enhancement
273 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
275 Severity : enhancement
277 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
282 Description: resolve race between obd_disconnect and class_disconnect_exports
283 Details : if obd_disconnect will be called to already disconnected export he
284 forget release one reference and osc module can't unloaded.
286 Severity : enhancement
288 Description: Allow tuning service thread via /proc
289 Details : For each service a new
290 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
291 created that can be used to set min/max thread counts, and get the
292 current number of running threads.
297 Descriptoin: don't return error if have partially created objects for file.
298 Details : lov_update_create_set uses set->set_success as index for created
299 objects, so if some requests will be failed, they will have hole
300 at end of array and qos_shrink_lsm for allocate correct lsm.
302 Severity : enhancement
304 Description: Update OFED support to 1.4.1
309 Description: Reduce small size read RPC
310 Details : Set read-ahead limit for every file and only do read-ahead when
311 available read-ahead pages are bigger than 1M to avoid small size
314 Severity : enhancement
316 Description: move AT tunable parameters for more consistent usage
317 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
319 Severity : enhancement
321 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
322 Details : allow skip disconnected ost for send statfs request and hide error
328 Description: fix racy locking of mballoc block bitmaps causing BUG
329 Details : The locking of the mballoc buddy bitmap and the in-memory
330 block bitmap was using two different spin locks in some
331 cases. This made it possible to incorrectly access the
332 mballoc bitmap while another process was modifying it,
333 causing a sanity assertion to fail. While no on-disk corruption
334 was reported, there was some risk of this happening.
337 Frequency : rare, on llog test 6
339 Descriptoin: don't allow connect to already connected import
340 Details : allowing connect to already connected import is hide connecting problem.
345 Descriptoin: allow kill process which wait statahead result
346 Details : for some reasons 'ls' can stick in waiting result from statahead,
347 in this case need way for kill this process.
349 Severity : enhancement
351 Description: Add state history info file, enhance import info file
352 Details : Track import connection state changes in a new osc/mdc proc file;
353 add overview-type data to the osc/mdc import proc file.
355 Severity : enhancement
357 Description: MDS create should not wait for statfs RPC while holding DLM lock.
359 Severity : enhancement
361 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
366 Descriptoin: OST grant too much space to client even there are not enough space.
367 Details : Client will shrink its grant cache to OST if there are no write
368 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
369 this grant cache if there are already not enough avaible space
370 (left_space < total_clients * 32M).
373 Frequency : start MDS on uncleanly shutdowned MDS device
375 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
376 Details : stay in waiting mds<>ost recovery finished produce random bugs
377 due race between two ll_sync thread for one lov target. send
378 ACTIVATE event only if connect realy finished and import have
382 Frequency : rare, connect and disconnect target at same time
384 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
385 Details : don't call obd_disconnect under lov_lock. this long time
386 operation and can block ptlrpcd which answer to connect request.
391 Descriptoin: don't lose wakeup for imp_recovery_waitq
392 Details : recover_import_no_retry or invalidate_import and import_close can
393 both sleep on imp_recovery_waitq, but we was send only one wakeup
397 Frequency : always with long access acl
399 Descriptoin: mds can't pack reply with long acl.
400 Details : mds don't control size of acl but they limited by reint/getattr
403 Severity : enhancement
405 Description: Update to SLES10 kernel-2.6.16.60-0.33.
407 Severity : enhancement
409 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
412 Frequency : start MDS on uncleanly shutdowned MDS device
414 Descriptoin: aborting recovery hang on MDS
415 Details : don't throttle destroy RPCs for the MDT.
418 Frequency : on remount
420 Description: external journal device not working after the remount
421 Details : clear dev_rdonly flag for external journal devices in
427 Description: shutdown vs evict race
428 Details : client_disconnect_export vs connect request race.
429 if client will evicted at this time - we start invalidate
430 thread without referece to import and import can be freed
434 Frequency : rare, need acl's on inode.
436 Description: client can't handle ost additional correctly
437 Details : if ost was added after client connected to mds client can have
438 hit lnet_try_match_md ... to big messages to wide striped files.
439 in this case need teach client to handle config events about add
440 lov target and update client max ea size at that event.
442 Severity : enhancement
444 Description: Changelogs
445 Details : Changelogs are a lightweight mechanism to track filesystem
446 metadata and namespace changes. The changelog is recorded
447 permanently on the MDTs, and is periodically "consumed" / purged
448 when records are no longer needed.
450 Severity : enhancement
452 Description: compact fld format with extents
453 Details : Store range of seq rather than every seq in FLD. Seq
454 controller update FLD rather than clients. In Case of CMD, mdt0
455 has FLD, all other metadata server act as non persistent proxy
456 for FLD queries and cache fld entries in fld cache.
461 Description: don't skip ost target if they assigned to file
462 Details : Drop slow OSCs if we can, but not for requested start idx.
463 This means "if OSC is slow and it is not the requested
464 start OST, then it can be skipped, otherwise skip it only
465 if it is inactive/recovering/out-of-space.
469 Description: more cleanup in mds_lov
470 Details : not send LOV EA under replay, we can't know about they size at this
471 time. Don't allow client connect to mds before any ost connected,
472 for avoid problems with LOV EA size and returning EIO to client.
474 Severity : enhancement
476 Description: Interoperability at server side (Disk interoperability)
478 Severity : enhancement
480 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
482 Severity : enhancement
484 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
486 Severity : enhancement
488 Description: New client IO stack (CLIO).
490 Severity : enhancement
492 Description: Commit on sharing. Eliminate inter-client dependencies between
493 uncommitted transactions by doing transaction commits.
494 Thereby clients may recovery independently.
497 Frequency : Create a symlink file with a very long name
499 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
500 Details : If there is no extra space in the request for early cancels,
501 ldlm_req_handles_avail() returns 0 instead of a negative value.
503 Severity : enhancement
505 Description: Add /proc entry for import status
506 Details : The mdc, osc, and mgc import directories now have
507 an import directory that contains useful import data for debugging
510 Severity : enhancement
512 Description: Re-disable certain /proc logging
513 Details : Enable and disable client's offset_stats, extents_stats and
514 extents_stats_per_process stats logging on the fly.
517 Frequency : Only on FC kernels 2.6.22+
519 Description: oops in statahead
520 Details : Do not drop reference count for the dentry from VFS when lookup,
521 VFS will do that by itself.
523 Severity : enhancement
525 Description: Generic /proc file permissions
526 Details : Set /Proc file permissions in a more generic way to enable non-
527 root users operate on some /proc files.
531 Description: Hitting mdc_commit_close() ASSERTION
532 Details : Properly handle request reference release in
533 ll_release_openhandle().
537 Description: quota recovery deadlock during mds failover
538 Details : This patch includes att18982, att18236, att18237 in bz14840.
540 1. fix osts hang when mds does failover with quotaon
541 2. prevent watchdog storm when osts threads wait for the
546 Frequency : only patchless client
547 Description: add workaround for race between add/remove dentry from hash
549 Severity : enhancement
551 Description: Allow OST glimpses to return PW locks
555 Description: LBUG when llog conf file is full
556 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
560 Description: Prevent import from entering FULL state when server in recovery
564 Description: service mount cannot take device name with ":"
565 Details : Only when device name contains ":/" will mount treat it as
571 Description: replace ptlrpcd with the statahead thread to interpret the async
572 statahead RPC callback
576 Frequency : on recovery
577 Description: I/O failures after umount during fail back
578 Details : if client reconnected to restarted server we need join to recovery
579 instead of find server handler is changed and process self
580 eviction with cancel all locks.
582 Severity : enhancement
584 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
586 Severity : enhancement
588 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
590 Severity : enhancement
592 Description: Upcall on Lustre log has been dumped
593 Details : Allow for a user mode script to be called once a Lustre log has
594 been dumped. It passes the filename of the dumped log to the
595 script, the location of the script can be specified via
596 /proc/sys/lnet/debug_log_upcall.
601 Description: avoid idr_remove called for id which is not allocated.
602 Details : Move assigment s_dev for clustered nfs to end of initialization,
603 to avoid problem with error handling.
608 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
609 Details : When connection is reused this not moved from CONN_UNUSED_HASH
610 into CONN_USED_HASH and this prodice warning when put connection
611 again in unused hash.
614 Severity : enhancement
616 Description: File striping can now be set to use an arbitrary pool of OSTs.
618 Severity : enhancement
620 Description: Export bytes_read/bytes_write count on OSC/OST.
624 Description: Early reply size mismatch, MGC loses connection
625 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
626 the connect flags are properly negotiated.
631 Description: quotas are not honored with O_DIRECT
632 Details : all writes with the flag O_DIRECT will use grants which leads to
633 this problem. Now using OBD_BRW_SYNC to guard this.
637 Description: add quota statistics
638 Details : 1. sort out quota proc entries and proc code.
639 2. add quota statistics
641 Severity : enhancement
643 Description: enable quota support for HEAD.
647 Description: Properly propagate oinfo flags from lov to osc for statfs
648 Details : restore missing copy oi_flags to lov requests.
650 Severity : enhancement
652 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
653 : l_getgroups(8), lst(8), routerstat(8)
654 Details : included man pages for llobdstat(8), llstat(8),
655 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
657 Severity : enhancement
659 Description: configure's --enable-quota should check the
660 : kernel .config for CONFIG_QUOTA
661 Details : configure is terminated if --enable-quota is passed but
662 : no quota support is in kernel
666 Description: Remove portals compatibility
667 Details : Remove portals compatibility, not interoperable with releases
672 Description: Resolve device initialization race
673 Details : Prevent proc handler from accessing devices added to the
674 obd_devs array but yet be intialized.
676 Severity : enhancement
678 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
680 Severity : enhancement
682 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
687 Description: Using wrong pointer in osc_brw_prep_request
688 Details : Access to array[-1] can produce panic if kernel compiled with
689 CONFIG_PAGE_ALLOC enabled
693 Description: Client runs out of low memory
694 Details : Consider only lowmem when counting initial number of llap pages
698 Description: *optional* service tags registration
699 Details : if the "service tags" package is installed on a Lustre node
700 When the filesystem is mounted, a local-node service tag will
701 be created. See http://inventory.sun.com/ for more information
702 about the Service Tags asset management system.
706 Description: Kernel BUG tries to release flock
707 Details : Lustre does not destroy flock lock before last reference goes
708 away. So always drop flock locks when client is evicted and
709 perform unlock regardless of successfulness of speaking to MDS.
713 Description: add refcount protection for osc callbacks, avoid panic on shutdown
717 Description: sanity test 65a fails if stripecount of -1 is set
718 Details : handle -1 striping on filesystem in ll_dirstripe_verify
723 Description: ASSERTION(CheckWriteback(page,cmd)) failed
724 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
727 Severity : enhancement
729 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
733 Description: do not process already freed flock
734 Details : flock can possibly be freed by another thread before it reaches
735 to ldlm_flock_completion_ast.
739 Description: LBUG during stress test
740 Details : Need properly lock accesses the flock deadlock detection list.
744 Description: oops in page fault handler
745 Details : kernel page fault handler can return two special 'pages' in error
746 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
750 Description: timeout with invalidate import.
751 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
752 should be handled by ptlrpcd. This produce long age waiting and
753 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
755 Severity : enhancement
757 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
761 Description: enable MGS and MDT services start separately
762 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
763 a co-located MGS without starting the MGS, which is a complement
764 to 'nosvc' mount option.
768 Frequency : after recovery
769 Description: precreate to many object's after del orphan.
770 Details : del orphan st in oscc last_id == next_id and this triger growing
771 count of precreated objects. Set flag LOW to skip increase count
772 of precreated objects.
776 Frequency : rare, on clear nid stats
777 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
778 Details : when clean nid stats sometimes try destroy live entry,
779 and this produce panic in free.
783 Description: Stack overflow during MDS log replay
784 ease stack pressure by using a thread dealing llog_process.
788 Description: wait until IO finished before start new when do lock cancel.
789 Details : VM protocol want old IO finished before start new, in this case
790 need wait until PG_writeback is cleared until check dirty flag and
791 call writepages in lock cancel callback.
793 Severity : enhancement
795 Description: using special macro for print time and cleanup in includes.
799 Description: mds_mfd_close() ASSERTION(rc == 0)
800 Details : In mds_mfd_close(), we need protect inode's writecount change
801 within its orphan write semaphore to prevent possible races.
805 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
809 Frequency : rare, on shutdown ost
810 Description: don't hit live lock with umount ost.
811 Details : shrink_dcache_parent can be in long loop with destroy dentries,
812 use shrink_dcache_sb instead.
816 Description: don't panic with use echo client
817 Details : echo client pass NULL as client nid pointer and this produce null
822 Description: fix build on ppc32
823 Details : compile code with -m64 flag produce wrong object file for ppc32.
827 Description: add message levels for liblustreapi
831 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
832 Details : Move the 'good_osts' check before the 'total_bavail' check. This
833 will result in an -EAGAIN and in the exit call path we call
834 alloc_rr() which will with increasing aggressiveness attempt to
835 aquire precreated objects on the minimum number of required OSCs.
839 Description: Use old size assignment to avoid deadlock
840 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
841 the scheduling while holding a spinlock. We do not need locking
842 for size in ll_update_inode() because size is only updated from
843 the MDS for directories or files without objects, so there is no
844 other place to do the update, and concurrent access to such inodes
845 are protected by the inode lock.
849 Description: resolve "_IOWR redefined" build error on SLES10
853 Description: dump the memory debugging after all modules are unloaded to
854 suppress false negative in conf_sanity test 39
856 Severity : enhancement
858 Description: build kernel-ib packages for OFED 1.3 in our release cycle
863 Description: fix SLES kernel versioning
864 Details : the kernel version for our SLES 10 kernel did not include a "-"
865 before the "smp" at the end. while this was not a problem in
866 general, it did mean that software trying to use the kernel
867 version to try to detect a vendor specific kernel would fail.
868 this was most evident by the OFED build scripts.
872 Description: Don't update lov_desc members until making sure they are valid
873 Details : When updating lov_desc members via proc fs, need fix their
874 validities before doing the real update.
878 Description: don't put request into delay list while invalidate in flight.
879 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
880 import in flight. this produce timeout for invalidate and sometimes
881 can cause stale data.
885 Frequency : on ppc only
886 Description: not convert ost objects for directory because it's not exist.
887 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
891 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
893 Severity : enhancement
895 Description: Update to support 2.6.22.14 vanilla kernel.
899 Frequency : rare, on recovery
900 Description: read procfs can produce deadlock in some situation
901 Details : Holding lprocfs lock which send rpc can produce block for destroy
902 obd objects and this also block reconnect with -EALREADY. This
903 isn't fix all lprocfs bugs - but make it rare.
905 Severity : enhancement
907 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
910 Frequency : frequent on X2 node
912 Description: mdc_set_open_replay_data LBUG
913 Details : Set replay data for requests that are eligible for replay.
917 Description: lustre_mgs: operation 101 on unconnected MGS
918 Details : When MGC is disconnected from MGS long enough, MGS will evict the
919 MGC, and late on MGC cannot successfully connect to MGS and a lot
920 of the error messages complaining that MGS is not connected.
924 Frequency : on network error
925 Description: panic with double free request if network error
926 Details : mdc_finish_enqueue is finish request if any network error ocuring,
927 but it's true only for synchronus enqueue, for async enqueue
928 (via ptlrpcd) this incorrect and ptlrpcd want finish request
931 Severity : enhancement
933 Description: client-side metadata stat-ahead during readdir(directory readahead)
934 Details : perform client-side metadata stat-ahead when the client detects
935 readdir and sequential stat of dir entries therein
938 Frequency : on start mds
940 Description: Implement get_info(last_id) in obdfilter.
943 Frequency : occasional
945 Description: Correctly check stale fid, not start epoch if ost not support SOM
946 Details : open with flag O_CREATE need set old fid in op_fid3 because
947 op_fid2 was overwritten with new generated fid, but mds can answer
948 with one of these two fids and both is not stale. Setattr
949 incorrectly started an epoch and assume will be called
950 done_writing, but without SOM done_writing ever being called.
953 Frequency : rare, depends on device drivers and load
955 Description: MDS or OSS nodes crash due to stack overflow
956 Details : Code changes in 1.8.0 increased the stack usage of some functions.
957 In some cases, in conjunction with device drivers that use a lot
958 of stack the MDS (or possibly OSS) service threads could overflow
959 the stack. One change which was identified to consume additional
960 stack has been reworked to avoid the extra stack usage.
963 Frequency : occasional
965 Description: Do not fail import if osc_interpret_create gets -EAGAIN
966 Details : If osc_interpret_create got -EAGAIN it immediately exits and
967 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
968 oscc_has_objects and see OSC has no objests and call
969 oscc_internal_create to resend create request.
971 Severity : enhancement
973 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
975 Severity : enhancement
977 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
980 Frequency : very rare
982 Description: Fix replay if there is an un-replied request and open
983 Details : In some cases, older replay request will revert the
984 mcd->mcd_last_xid on MDS which is used to record the client's
987 Severity : enhancement
989 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
991 Severity : enhancement
993 Description: Add rhel5 support to HEAD.
995 Serverity : enhancement
997 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1002 Description: Don't allow skipping OSTs if index has been specified.
1003 Details : Don't allow skipping OSTs if index has been specified, make
1004 locking in internal create lots better.
1008 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1009 Details : access to bitfield in structure is always rounded to long
1010 and this produce problem with not atomic change any bit.
1014 Description: Lustre make rpms failed.
1015 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1016 builds Lustre rpms from tarball.
1019 Frequency : rare on shutdown ost
1021 Description: If llog cancel was not send before clean_exports phase, this can
1022 produce deadlock in llog code.
1023 Details : If llog thread has last reference to obd and call class_import_put
1024 this produce deadlock because llog_cleanup_commit_master wait when
1025 last llog_commit_thread exited, but this never success because was
1026 called from llog_commit_thread.
1030 Description: allow userland application know is lost one of stripes.
1031 Details : fill lvb_blocks with error code on ost and return it to
1032 application if error flag found.
1036 Description: NULL lov_tgts causing MDS oops
1037 Details : more safe checks for NULL lov_tgts for avoid oops.
1039 Severity : enhancement
1041 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1045 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1046 Details : If an OST is down the MDS will hang indefinitely in
1047 obd_statfs_rqset() waiting for the statfs data. While for
1048 MDS QOS usage of statfs, it should not stuck in waiting.
1050 Severity : enhancement
1052 Description: remote_acl support
1053 Details : Support ACL-based permission check for remote user.
1054 Support setfacl/getfacl for remote user with the utils
1055 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1056 the system "{s,g}etfacl" utils.
1058 Severity : enhancement
1060 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1062 Severity : enhancement
1064 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1068 Description: make mgs_setparam() handle fsname containing dash
1069 Details : fsname containing a dash does not work with lctl conf_param
1071 Severity : enhancement
1073 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1075 Severity : enhancement
1077 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1081 Frequency : rare, at shutdown
1082 Description: access already free / zero obd_namespace.
1083 Details : if client_disconnect_export was called without force flag set,
1084 and exist connect request in flight, this can produce access to
1085 NULL pointer (or already free pointer) when connect_interpret
1086 store ocd flags in obd_namespace.
1090 Frequency : only at startup
1091 Description: not alloc memory with spinlock held.
1092 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1093 if any spinlock held.
1095 Severity : enhancement
1097 Description: make lustre randomly fail allocating memory
1098 Details : Make lustre randomly failed allocating memory for testing purpose.
1100 Severity : enhancement
1102 Description: lost problems with lov objid file
1103 Details : Fixes some scability and access to not inited memory problems
1104 in work with lov objdid file.
1109 Description: lfs find does not continue on file error
1110 Details : Continue other files processing when a file/dir is absent.
1114 Description: Inconsistent usage of lustre_pack_reply()
1115 Details : Standardize the usage of lustre_pack_reply() such that it
1116 always generate a CERROR on failure.
1119 Frequency : occasional
1121 Description: MDS hang or stay in waiting lock
1122 Details : If client receive lock with CBPENDING flag ldlm need send lock
1123 cancel as separate rpc, to avoid situation when cancel request
1124 can't processed due all i/o threads stay in wait lock.
1128 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1132 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1136 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1140 Description: Update to RHEL4 latest kernel.
1142 Severity : enhancement
1144 Description: Build SLES10 patchless client fails
1145 Details : The configure was broken by run ./configure with
1146 --with-linux-obj=.... argument for patchless client. When the
1147 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1148 can't search header adequately. Use absolute path such as
1149 -I($LINUX)/include instead.
1153 Description: interrupt oig_wait produce painc on resend.
1154 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1155 private set, and this produce situation when rq_ptlrpcd_data not
1156 copyed to new allocated request and triggered LBUG on assert
1157 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1158 ptlrpcd set if request is changed and can be safety changed to use
1163 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1164 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1165 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1166 these macros and replace them with their difinitions instead.
1170 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1171 Details : Update which_patch & target file for SLES10 latest kernel.
1174 Description: Remove client patches from SLES 10 kernel.
1175 Details : This causes SLES 10 clients to behave as patchless clients
1176 even on a Lustre-patched (server) kernel.
1178 Severity : enhancement
1180 Description: self-adjustable client's lru lists
1181 Details : use adaptive algorithm for managing client cached locks lru
1182 lists according to current server load, other client's work
1183 pattern, memory activities, etc. Both, server and client
1184 side namespaces provide number of proc tunables for controlling
1187 Severity : enhancement
1189 Description: light-weight GSS support
1190 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1191 and reduce performance overhead.
1193 Severity : enhancement
1195 Description: Linux keyring support
1196 Details : Support using service of Linux keyring for Lustre GSS internal
1197 context refresh/cache mechanism.
1201 Description: Fix errors in lfs documentation
1202 Details : Fixes man pages
1206 Description: don't use GFP_* in generic Lustre code.
1207 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1211 Description: obdclass is limited by single OBD_ALLOC(idarray)
1212 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1216 Description: warnings with build patchless client with vanila 2.6.19 and up
1217 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1218 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1222 Frequency : Only for SLES
1223 Description: sanity_quota fail test_1
1224 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1225 file, which makes TSTID[2] inunique.
1229 Frequency : Only for SLES9
1230 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1231 Details : Update target/ChangeLog/which_patch .
1235 Frequency : Only for test_10 in sanity-quota.sh
1236 Description: a bug in quota test code
1237 Details : checking a test flag in a wrong place
1241 Frequency : Only for test_13 in sanity-quota.sh
1242 Description: a bug in quota test script
1243 Details : shouldn't check overfull allocated quota
1247 Description: jbd statistics
1248 Details : Port older jbd statistics patch for sles10
1250 Severity : enhancement
1252 Description: Add printing inode info into message about error in writepage.
1257 Description: reply_lock_interpret crash due to race with it and lock cancel.
1258 Details : Do not replay locks that are being cancelled. Do not reference
1259 locks by their address during replay, just by their handle.
1263 Frequency : When flocks are used.
1264 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1265 Details : Flock locks might destroy just granted lock if it could be merged
1266 with another existing flock, this is done in completion handler,
1267 so teach ldlm_cli_enquque_fini that this is a valid case for
1273 Description: Oops in read and write path when failing to allocate lock.
1274 Details : Check if lock allocation failed and return error back.
1278 Description: lstripe command fails for valid OST index
1279 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1280 instead of lov->desc.ld_active_tgt_count.
1284 Description: sanity.sh failed test 103
1285 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1290 Description: df doesn't work properly if diskfs blocksize != 4K
1291 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1295 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1296 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1298 Severity : enhancement
1300 Description: Adaptive timeouts
1301 Details : RPC timeouts adapt to changing server load and network
1302 conditions to reduce resend attempts and improve recovery time.
1306 Description: llapi_file_create() does not allow some changes
1307 Details : add llapi_file_open() that allows specifying the mode and
1308 open flags, and also returns an open file handle.
1312 Description: merge and cleanup kernel patches.
1313 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1317 Description: Add journal checksum support.(Kernel part)
1318 Details : The journal checksum feature adds two new flags i.e
1319 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1320 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1321 indicates that the commit block contains the checksum for
1322 the blocks described by the descriptor blocks. Now commit
1323 record can be sent to disk without waiting for descriptor
1324 blocks to be written to disk. This behavior is controlled
1325 using JBD2_FEATURE_ASYNC_COMMIT flag.
1329 Description: OSS needs mutliple precreate threads
1330 Details : Add ability to start more than one create thread per OSS.
1334 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1335 Details : execute lfs setstripe on client
1339 Description: mds_obd_create error creating tmp object
1340 Details : When the user sets quota on root, llog will be affected and can't
1341 create files and write files.
1344 Frequency : Always on ia64 patchless client, and possibly others.
1346 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1347 Details : This allows the patchless client to be loaded on architectures
1348 without this export.
1352 Description: RedHat Update kernel for RHEL5
1353 Details : Add the kernel config file for RHEL5.
1357 Description: RedHat Update kernel for RHEL5
1358 Details : Modify the kernel config file more closer RHEL5.
1362 Description: Build failure against Centos5 (RHEL5)
1363 Details : Define PAGE_SIZE when it isn't present.
1367 Description: test_7 on sanity-quota.sh can't be used on separate machines
1368 Details : add facet to handle it
1372 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1373 Details : with NFS, the anon dentry's parent was set to itself in
1374 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1375 corresponding dentry other than use rec->ur_name.
1377 Severity : enhancement
1379 Description: lfs setstripe enhancement
1380 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1384 Description: enable data checksumming by default
1385 Details : enable checksum by default, allow --disable-checksum
1386 configure option and "-o nochecksum" mount option. Checksums
1387 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1391 Description: lustre support for RHEL5
1392 Details : Add support for RHEL5.
1396 Description: Client eviction due to failover config
1397 Details : after a connection loss, the lustre client should attempt to
1398 reconnect to the last active server first before trying the
1399 other potential connections.
1403 Description: when mds and osts use different quota unit(32bit and 64bit),
1404 quota will be released repeatly.
1405 Details : void sending multiple quota reqs to mds, which will keep the
1406 status between the reqs.
1410 Description: osts not allocated evenly to files
1411 Details : change the condition to increase offset_idx
1414 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1416 Description: Data corruption for OSTs that are formatted larger than 2TB
1418 Details : When generating the bio request for lustre file writes the
1419 sector number would overflow a temporary variable before being
1420 used for the IO. The data reads correctly from Lustre (which
1421 will overflow in a similar manner) but other file data or
1422 filesystem metadata may be corrupted in some cases.
1426 Description: Tune the kernel for good SCSI performance.
1427 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1428 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1433 Description: rewrite ext2-derived code in obdclass/uuid.c
1434 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1435 from scratch preserving functionality.
1439 Description: "lfs find -obd UUID" prints directories
1440 Details : "lfs find -obd UUID" will return all directory names instead
1441 of just file names. It is incorrect because the directories
1442 do not reside on the OSTs.
1446 Description: use i_size_read and i_size_write in 2.6 port
1447 Details : replace inode->i_size access with i_size_read/write()
1450 Frequency : when removing large files
1452 Description: scheduling issue during removal of large Lustre files
1453 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1454 It causes scheduling issues when removing large files (17TB in the
1458 Frequency : only with liblustre clients on XT3
1460 Description: evictions taking too long
1461 Details : allow llrd to evict clients directly on OSTs
1464 Frequency : only on ppc
1466 Description: /proc/fs/lustre/devices broken on ppc
1467 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1468 We should make sure the type of 'index' is loff_t and then cast
1469 explicitly as needed below. Do not assign an explicitly cast
1473 Frequency : only for rhel5
1475 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1476 Details : Modify the target file & which_kernel.
1478 Severity : enhancement
1480 Description: omit set fsid for export NFS
1481 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1482 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1486 Frequency : Always, for kernels after 2.6.16
1487 Description: Fix warning idr_remove called for id=.. which is not allocated.
1488 Details : Last kernels save old s_dev before kill super and not allow
1489 to restore from callback - restore it before call kill_anon_super.
1493 Description: buffer overruns could theoretically occur
1494 Details : llapi_semantic_traverse() modifies the "path" argument by
1495 appending values to the end of the origin string, and a
1496 overrun may occur. Adding buffer overrun check in liblustreapi.
1500 Description: Fix error on 'ls .' at the top of the Lustre mount.
1501 Details : Don't revalidate dentry if it is a root dentry.
1505 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1506 Details : Modify vm-tunables-rhel4.patch.
1510 Description: Kernel config for 2.6.18-vanilla.
1511 Details : Modify targets/2.6-vanilla.target.in.
1512 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1513 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1514 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1515 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1519 Description: improve handling recoverable errors
1520 Details : if request processig with error which can be recoverable on server
1521 request should be resend, otherwise page released from cache and
1526 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1527 Details : Modify target file & which_patch.
1528 A flaw was found in the IA32 system call emulation provided
1529 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1530 value could be stored in the %RAX register, which could trigger an
1531 out-of-bounds system call table access. An untrusted local user
1532 could exploit this flaw to run code in the kernel
1533 (ie a root privilege escalation). (CVE-2007-4573).
1537 Description: O_DIRECT bypasses client statistics.
1538 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1539 read_ahead_stats were updated. copied the stats section from
1540 osc_send_oap_rpc() into async_internal().
1544 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1548 Description: change order of libsysio includes
1549 Details : '#include sysio.h' should always come before '#include xtio.h'
1553 Description: handle "serious error: objid * already exists" more gracefully
1554 Details : If LAST_ID value on disk is smaller than the objects existing in
1555 the O/0/d* directories, it indicates disk corruption and causes an
1556 LBUG(). If the object is 0-length, then we should use the existing
1557 object. This will help to avoid a full fsck in most cases.
1559 Severity : enhancement
1561 Description: adapt the lustre_config script to support the upgrade case
1562 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1563 server targets to 1.6 in parallel.
1567 Description: To avoid grant space > avaible space when the disk is almost
1568 full. Without this patch you might see the error "grant XXXX >
1569 available" or some LBUG about grant, when the disk is almost
1571 Details : In filter_check_grant, for non_grant cache write, we should
1572 check the left space by if (*left > ungranted + bytes), instead
1573 of (*left > ungranted), because only we are sure the left space
1574 is enough for another "bytes", then the ungrant space should be
1575 increase. In client, we should update cl_avail_grant only there
1576 is OBD_MD_FLGRANT in the reply.
1580 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1581 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1582 described in CVE-2007-4573. This problem could allow untrusted
1583 local users to gain root access.
1586 Frequency : when using O_DIRECT and quotas
1588 Description: Incorrect file ownership on O_DIRECT output files
1589 Details : block usage reported by 'lfs quota' does not take into account
1590 files that have been written with O_DIRECT.
1595 Description: touch file failed when fs is not full
1596 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1597 otherwise we can get ENOSP while fs is not full.
1601 Description: parallel lock callbacks
1602 Details : Instead of sending blocking and completion callbacks as separated
1603 requests, adding them to a set and sending in parallel.
1606 Frequency : only for Cray XT3
1607 Bugzilla : 12829/13455
1608 Description: Changing primary group doesn't change the group lustre assigns to
1610 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1611 overridden with the primary group provided by the group upcall,
1612 whereas the supplied fsgid can be trusted if it is in the list of
1613 supplementary groups returned by the group upcall.
1615 Severity : enhancement
1617 Description: Allow masking D_WARNING, D_ERROR messages from console
1618 Details : Console messages can now be disabled via lnet.printk.
1622 Description: User code with malformed file open parameter crashes client node
1623 Details : Before packing join_file req, all the related reference should be
1624 checked carefully in case some malformed flags cause fake
1625 join_file req on client.
1629 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1630 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1631 mode is equal to the granted mode. However, it should also check
1632 that LDLM_FL_CP_REQD is not set.
1636 Description: Heavy nfs access might result in deadlocks
1637 Details : After ELC code landed, it is now improper to enqueue any mds
1638 locks under och_sem, because enqueue might want to decide to
1639 cancel open locks for same inode we are holding och_sem for.
1643 Description: Client eviction while running blogbench
1644 Details : A lot of unlink operations with concurrent I/O can lead to a
1645 deadlock causing evictions. To address the problem, the number of
1646 oustanding OST_DESTROY requests is now throttled to
1647 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1648 callbacks are processed in priority.
1652 Description: enable ACLs on MDS by default
1653 Details : ACLs must be enabled on MDS by default.
1656 Frequency : PPC/PPC64 only
1658 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1659 Details : fix duplicated definitions between asm-ppc64/types.h and
1660 lustre_types.h on PPC.
1663 Frequency : PPC/PPC64 only
1665 Description: asm-ppc/segment.h does not exist
1666 Details : fix compile issue on PPC.
1670 Description: better handle error messages in extents code
1673 Frequency : RHEL4 only
1675 Description: mkfs is very slow on IA64/RHEL4
1676 Details : A performance regression has been discovered in the MPT Fusion
1677 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1678 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1679 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1681 Severity : enhancement
1683 Description: SNMP support enhancement
1684 Details : Adding total number of sampled request for an MDS node in snmp
1687 Severity : enhancement
1689 Description: Optimize ldlm waiting list processing for PR extent locks
1690 Details : When processing waiting list for read extent lock and meeting read
1691 lock that is same or wider to it that is not contended, skip
1692 processing rest of the list and immediatelly return current
1693 status of conflictness, since we are guaranteed there are no
1694 conflicting locks in the rest of the list.
1698 Description: Time out and refuse to reconnect
1699 Details : When the failover node is the primary node, it is possible
1700 to have two identical connections in imp_conn_list. We must
1701 compare not conn's pointers but NIDs, otherwise we can defeat
1702 connection throttling.
1706 Description: port llog fixes from b1_6 into HEAD
1707 Details : Port llog reference couting and some llog cleanups from b1_6
1708 (bug 10800) into HEAD, for protect from panic and access to already
1709 free llog structures.
1713 Description: Detect stride IO mode in read-ahead
1714 Details : When a client does stride read, read-ahead should detect that and
1715 read-ahead pages according to the detected stride pattern.
1719 Description: data checksumming impacts single node performance
1720 Details : add support for several checksum algorithm. Currently, only CRC32
1721 and Adler-32 are supported. The checksum type can be changed on
1722 the fly via /proc/fs/lustre/osc/*/checksum_type.
1726 Description: use adler32 for page checksums
1727 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1732 Description: build for x2 fails
1733 Details : fix compile issue on Cray systems.
1737 Description: Properly match for duplicate locks
1738 Details : Due to different lock order from skiplists code, we need to
1739 traverse entire list for now
1742 Frequency : only on PPC/SLES10
1744 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1745 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1746 be defined. Add a hack in mkfs_lustre.c to work around this compile
1751 Description: LASSERT on MDS when client holding flock lock dies
1752 Details : ldlm pool logic depends on number of granted locks equal to
1753 number of released locks which is not true for flock locks, so
1754 just exclude such locks from consideration.
1758 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1759 Details : Use fsfilt_sync() for both the whole filesystem sync and
1760 individual file sync to eliminate dangerous inode locking
1761 with I_LOCK that can lead to a deadlock.
1765 Description: performance in 1.6.3
1766 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1767 sure that 1MB requests are not fragmented by the block layer.
1769 Severity : enhancement
1771 Description: organize the server-side client stats on per-nid basis
1772 Details : Change the structure of stats under obdfilter and mds to
1780 The "uuid"s file would list the uuids of _active_ exports.
1781 And the clear entry is to clear all stats and stale nids.
1783 Severity : enhancement
1785 Description: eliminate client locks in face of contention
1786 Details : file contention detection and lockless i/o implementation
1787 for contended files.
1791 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1794 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1796 Description: MDS may lose file striping (and hence file data) in some cases
1797 Details : If there are additional extended attributes stored on the MDS,
1798 in particular ACLs, SELinux, or user attributes (if user_xattr
1799 is specified for the client mount options) then there is a risk
1800 of attribute loss. Additionally, the Lustre file striping
1801 needs to be larger than default (e.g. striped over all OSTs),
1802 and an additional attribute must be stored initially in the
1803 inode and then increase in size enough to be moved to the
1804 external attribute block (e.g. ACL growing in size) for file
1809 Description: skiplist implementation simplification
1810 Details : skiplists are used to group compatible locks on granted list
1811 that was implemented as tracking first and last lock of each
1812 lock group the patch changes that to using doubly linked lists
1816 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1817 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1818 because LASSERT on both the data supplied by a client, and the
1819 data on disk is dangerous and incorrect.
1821 Severity : enhancement
1823 Description: Slow truncate/writes to huge files at high offsets.
1824 Details : Directly associate cached pages to lock that protect those pages,
1825 this allows us to quickly find what pages to write and remove
1826 once lock callback is received.
1830 Description: more ldlm soft lockups
1831 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1832 starve other threads from the resource lock for a long time in
1833 case of long waiting queue, so change the debug level from
1834 D_OTHER to the less frequently used D_INFO.
1836 Severity : enhancement
1838 Description: add -gid, -group, -uid, -user options to lfs find
1842 Description: Hung threads in invalidate_inode_pages2_range
1843 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1844 once one is completed. As a result, some RPCs are stuck in the
1845 queue and are never sent.
1849 Description: filter threads hungs on waiting journal commit
1850 Details : Cleanup filter group llog code, then only filter group llog will
1851 be only created in the MDS/OST syncing process.
1855 Description: Procfs and llog threads access destoryed import sometimes.
1856 Details : Sync the import destoryed process with procfs and llog threads by
1857 the import refcount and semaphore.
1859 Severity : enhancement
1861 Description: openlock cache of b1_6 port to HEAD
1866 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1867 Details : If insertion of an extent fails, then discard the inode
1868 preallocation and free data blocks else it can lead to duplicate
1873 Description: don't always update ctime in ext3_xattr_set_handle()
1874 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1875 In some cases the ctime should not be updated, for example for
1876 2.0->1.8 compatibility it is necessary to delete an xattr and it
1877 should not update the ctime.
1881 Bugzilla : 15713/16362
1882 Description: Assertion in iopen_connect_dentry in 1.6.3
1883 Details : looking up an inode via iopen with the wrong generation number can
1884 populate the dcache with a disconneced dentry while the inode
1885 number is in the process of being reallocated. This causes an
1886 assertion failure in iopen since the inode's dentry list contains
1887 both a connected and disconnected dentry.
1891 Description: assertion failure in ldlm_handle2lock()
1892 Details : fix a race between class_handle_unhash() and class_handle2object()
1893 introduced in lustre 1.6.5 by bug 13622.
1898 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1899 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1904 Description: oops in sd_iostats_seq_show()
1905 Details : unloading/reloading the scsi low level driver triggers a kernel
1906 bug when trying to access the sd iostat file.
1911 Description: Kernel panics during QLogic driver reload
1912 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1913 patch, causing memory corruption.
1918 Description: journal_dev option does not work in b1_6
1919 Details : pass mount option during pre-mount.
1921 Severity : enhancement
1923 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1924 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1925 extent information of a file. It can be used to map logical blocks
1926 in a file to physical blocks in the block device.
1930 Description: LDLM soft lockups - improvement
1931 Details : It is be possible to send the lock handle along with each read
1932 or write request because the client is already doing a lock match
1933 itself so there isn't any reason the OST should have to re-do that
1937 Frequency : only on Cray X2
1939 Description: X2 build failures
1940 Details : fix build failures on Cray X2.
1944 Description: xid & resent requests
1945 Details : Initialize RPC XID from clock at startup (randomly if clock is
1948 Severity : enhancement
1950 Description: Add lustre_start utility to start or stop multiple Lustre servers
1955 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1956 Details : In case of memory pressure, list_del() can be called twice on
1957 req->rq_history_list, causing a kernel oops.
1961 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1962 Details : fix stack overflow in the distributed lock manager by defering
1963 export eviction after a failed AST to the elt thread instead of
1964 handling it in the dlm interpret routine.
1968 Description: Convert some comments to new format.
1969 Details : Update documenting comments to match doxygen conventions.
1973 Description: Grammar fixes.
1974 Details : A couple of trivial spelling fixes.
1978 Description: OSD_COUNTERS-mandatory
1979 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1980 to simplify the code.
1984 Description: simplify cmm_device freeing logic.
1985 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1989 Description: Add lockdep support to dt_object_operations locking interface.
1990 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1991 parameter indicating lock ordering. Update mdd code to use new
1996 Description: Introduce failloc constants for lockless IO tests.
1997 Details : Add two new failloc constants to test lockless IO. Only one of
1998 them in implemented---another is checked in yet to be landed
2003 Description: Add lockdep support for inode mutex.
2004 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2005 in the situations where Lustre has to lock more than one inode
2010 Description: Add optional invariants checking support.
2011 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2012 configure switch. This macro is to be used for consistency and
2013 sanity checks that are too expensive to be left in `production'
2018 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2019 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2024 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2025 Details : Use spin_lock_nested() in (the only) situation where more than
2026 one ldlm_lock is locked simultaneously. Also, fix possible
2027 dead-lock in ldlm_lock_change_resource() by enforcing particular
2032 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2033 Details : Instead of specifying each ldlm_lock call-back through separate
2034 parameter, wrap them into struct ldlm_callback_suite.
2038 Description: Kill join_lru obd method and its callers.
2039 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2040 in memory---a requirement implemented through ->o_join_lru() obd
2041 method in HEAD. Remove this method and its users.
2045 Description: Add asynchronous ldlm ENQUEUE completion handler.
2046 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2047 case that stock ldlm_completion_ast() cannot handle as it waits
2048 until lock is granted. Introduce new ldlm_completion_ast_async()
2049 for this. Also comment ldlm_completion_ast().
2053 Description: ldlm_error <-> errno conversion.
2054 Details : Add functions to map (rather arbitrary) between LDLM error codes
2055 and standard errno values. CLIO needs this to prevent LDLM specific
2056 constants from escaping ldlm and osc.
2060 Description: Kill unused ldlm_handle2lock_ns() function.
2061 Details : Kill unused ldlm_handle2lock_ns() function.
2065 Description: Add lu_ref support to ldlm_lock
2066 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2067 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2068 and ldlm_resource. LDLM interface has to be changed, because code that
2069 releases a reference on a lock, has to "know" what reference this is.
2070 In the most frequent case
2072 lock = ldlm_handle2lock(handle);
2074 LDLM_LOCK_PUT(lock);
2076 no changes are required. When any other reference (received _not_ from
2077 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2078 instead of LDLM_LOCK_PUT().
2080 Arguably, changes are pervasive, and interface requires some discipline
2081 for proper use. On the other hand, it was very instrumental in finding
2082 a few leaked lock references.
2086 Description: Add ldlm_lock_addref_try().
2087 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2088 attempts to addref a lock that might be being canceled
2093 Description: Add ldlm_weigh_callback().
2094 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2095 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2096 hard-coded `number of pages' logic.
2100 Description: Add lockdep annotations to llog code.
2101 Details : Use appropriately tagged _nested() locking calls in the places
2102 where llog takes more than one ->lgh_lock lock.
2106 Description: Add loi_kms_set().
2107 Details : Wrap kms updates into a helper function.
2111 Description: Constify instances of struct lsm_operations.
2112 Details : Constify instances of struct lsm_operations.
2116 Description: lu_conf support.
2117 Details : On a server, a file system object is uniquely identified
2118 by a fid, which is sufficient to locate and load all object
2119 state (inode). On a client, on the other hand, more data are
2120 necessary instantiate an object. Change lu_object_find() and
2121 friends to take additional `lu_conf' argument describing object.
2122 Typically this includes layout information.
2126 Description: lu_context fixes.
2127 Details : Introduce new lu_context functions that are needed on the client
2128 side, where some system threads (ptlrpcd) are shared by multiple
2129 modules, and so cannot be stopped during module shutdown.
2133 Description: Add start and stop methods to lu_device_type_operations.
2134 Details : Introduce two new methods in lu_device_type_operations, that are
2135 invoked when first instance of a given type is created and last one
2136 is destroyed respectively. This is need by CLIO.
2140 Description: Add lu_ref support to struct lu_device.
2141 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2142 track leaked references.
2146 Description: Introduce lu_kmem_descr.
2147 Details : lu_kmem_descr and its companion interface allow to create
2148 and destroy a number of kmem caches at once.
2152 Description: Fix lu_object finalization race.
2153 Details : Fix a race between lu_object_find() finding an object and its
2154 concurrent finalization. This race is (most likely) not possible
2155 on the server, but might happen on the client.
2159 Description: Introduce lu_ref interface.
2160 Details : lu_ref is a debugging module allowing to track references to
2161 a given object. It is quite cpu expensive, and has to be
2162 explicitly enabled with --enable-lu_ref. See usage description
2167 Description: Factor lu_site procfs stats into a separate function.
2168 Details : Separate lu_site stats printing code into a separate function
2169 that can be reused on a client.
2173 Description: Constify instances of struct {lu,dt,md}_device_operations.
2174 Details : Constify instances of struct {lu,dt,md}_device_operations.
2178 Description: Introduce struct md_site and move meta-data specific parts of
2179 struct lu_site here.
2180 Details : Move md-specific fields out of struct lu_site into special struct
2181 md_site, so that lu_site can be used on a client.
2185 Description: Kill mdd_lov_destroy().
2186 Details : Remove unused mdd code.
2190 Description: Add st_block checking to multistat.c.
2191 Details : Add st_block checking to multistat.c.
2195 Description: Add lu_ref support to struct obd_device.
2196 Details : Add lu_ref tracking to obd_device.
2200 Description: Kill obd_set_fail_loc().
2201 Details : Remove unused code.
2205 Description: Add special type for ptlrpc_request interpret functions.
2206 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2207 there. Actual usage will be in CLIO.
2211 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2212 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2213 doesn't work with lockdep.
2217 Description: Add rwv.c test program.
2218 Details : New testing program exercising readv(2) and writev(2) (Qian).
2222 Description: Add sendfile.c test program.
2223 Details : New testing program exercising sendfile(2) (Jay).
2227 Description: Ratelimit a message that can be very frequent.
2228 Details : Ratelimit a memory allocation failure message that can
2233 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2234 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2238 Description: Liblustre build fixes.
2239 Details : Liblustre build fixes.
2243 Description: libcfs: add cfs_{need,cond}_resched() interface.
2244 Details : libcfs: add cfs_{need,cond}_resched() definition and
2245 implementations for Linux, NT, and liblustre.
2247 Severity : enhancement
2249 Description: More exported tunables for mballoc
2250 Details : Add support for tunable preallocation window and new tunables for
2251 large/small requests
2255 Description: Detect corruption of block bitmap and checking for preallocations
2256 Details : Checks validity of on-disk block bitmap. Also it does better
2257 checking of number of applied preallocations. When corruption is
2258 found, it turns filesystem readonly to prevent further corruptions.
2262 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2263 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2264 window to 0. If it does hit cache miss, and read-pattern is still
2265 stride-io mode, does not reset the stride window, but also does not
2266 increase the stride window length in this case.
2270 Frequency : only for big-endian servers
2271 Description: Check if system is big-endian while mounting fs with extents feature
2272 Details : Mounting a filesystem with extents feature will fail on big-endian
2273 systems since ext3-based ldiskfs is not supported on big-endian
2274 systems. This can be over-riden with "bigendian_extents" mount option.
2276 Severity : enhancement
2278 Description: The root squash functionality
2279 Details : A security feature, which is to prevent users from being able
2280 to mount lustre on their desktop, run as root, and delete
2281 all of the files in the filesystem. The goal is accomplished by
2282 remapping user id (UID) and group id (GID) of the root user to
2283 a UID and GID specified by the system administartor via Lustre
2284 configuration management server (MGS). The functionality also
2285 allows to specify sets of clients for which the remapping does
2290 Description: Excessive recovery window
2291 Details : With AT enabled, the recovery window can be excessively long (6000+
2292 seconds). To address this problem, we no longer use
2293 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2294 timeout no longer depends on the service time, it is set to
2295 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2296 time via pb_service_time.
2300 Description: Watchdog triggered on MDS failover
2301 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2302 the OSTs know that the MDS "UUID" can be reused for the same export
2303 from a different NID, so we do not need to wait for the export to be
2307 Frequency : rare, only if using MMP with Linux RAID
2309 Description: MMP doesn't work with Linux RAID
2310 Details : While using HA for Lustre servers with Linux RAID, it is possible
2311 that MMP will not detect multiple mounts. To make this work we
2312 need to unplug the device queue in RAID when the MMP block is being
2313 written. Also while reading the MMP block, we should read it from
2314 disk and not the cached one.
2316 Severity : enhancement
2318 Description: open file using fid
2319 Details : A file can be opened using just its fid, like
2320 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2323 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2326 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2327 Details : add proper locking for creation of jbd_4k slab cache
2331 Description: MMP check in ext3_remount() fails without displaying any error
2332 Details : When multiple mount protection fails during remount, proper error
2335 Severity : enhancement
2337 Description: Allow stripe size to be up to 4G-64k
2338 Details : Fix math logic to allow large stripe sizes.
2342 Description: add check for >8TB ldiskfs filesystems
2343 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2344 Don't allow >8TB ldiskfs filesystems to be mounted without
2345 force_over_8tb mount option
2350 Descriptoin: OSS complains "too many blocks" and crashes
2351 Details : In filter_commitrw_write():
2352 1. lquota_chkquota();
2353 2. obdo_from_inode();
2354 3. lquota_pending_commit();
2355 "2." has changed uid and gid in this case, then caused
2356 lquota_chkquota() and lquota_pending_commit() to use
2357 different uid and gid. That is the root of the bug.
2361 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2362 Details : Detect and print out the 'stripe count = -1' case properly.
2366 Description: Out or order replies might be lost on replay
2367 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2368 request with tid smaller then one currently being added, add it
2369 to the start, not end of the list.
2371 --------------------------------------------------------------------------------
2373 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2375 * Support for kernels:
2376 2.6.5-7.283 (SLES 9),
2377 2.6.9-55.EL (RHEL 4),
2378 2.6.16.46-0.14 (SLES 10),
2379 2.6.18.8 vanilla (kernel.org)
2380 * Client support for unpatched kernels:
2381 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2382 2.6.16 - 2.6.22 vanilla (kernel.org)
2383 * Due to recently discovered recovery problems, we do not recommend
2384 using patchless RHEL 4 clients with this or any earlier release.
2385 * Recommended e2fsprogs version: 1.39.cfs8
2386 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2387 * Starting with this release, the ldiskfs backing filesystem required
2388 by Lustre is now in its own package, lustre-ldiskfs. This package
2389 should be installed. It is versioned separately from Lustre and
2390 may be released separately in future.
2394 Description: block reactivating mgc import until all deactivates complete
2395 Details : Fix race when failing back MDT/MGS to itself (testing)
2398 Frequency : at statup only
2400 Description: mds_lov_synchronize race leads to various problems
2401 Details : simultaneous MDT->OST connections at startup can cause the
2402 sync to abort, leaving the OSC in a bad state.
2404 Severity : enhancement
2406 Description: add optional extra BUILD_VERSION info
2407 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2408 to override the lustre version.
2411 Frequency : 2.6.18 servers only
2413 Description: ll_kern_mount() doesn't release the module reference
2414 Details : The ldiskfs module reference count never drops down to 0
2415 because ll_kern_mount() doesn't release the module reference.
2420 Description: server LBUG when using old ost_num_threads parameter
2421 Details : Accept the old ost_num_threads parameter but warn that it
2422 is deprecated, and fix an off-by-one error that caused an LBUG.
2427 Description: Transient SCSI error results in persistent IO issue
2428 Details : iobuf->dr_error is not reinitialized to 0 between two
2432 Frequency : sometimes when underlying device returns I/O errors
2434 Description: OSTs not going read-only during write failures
2435 Details : OSTs are not remounted read-only when the journal commit threads
2436 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2437 instead of the ext3 wrappers.
2441 Description: poor connect scaling with increasing client count
2442 Details : Don't run filter_grant_sanity_check for more than 100 exports
2443 to improve scaling for large numbers of clients.
2446 Frequency : SLES10 only
2448 Description: sanity-quota.sh quotacheck failed: rc = -22
2449 Details : Quotas cannot be enabled on SLES10.
2452 Frequency : liblustre clients only
2454 Description: getdirentries does not give error when run on compute nodes
2455 Details : getdirentries does not fail when the size specified as an argument
2456 is too small to contain at least one entry
2458 Severity : enhancement
2460 Description: Add LNET router traceability for debug purposes
2461 Details : If a checksum failure occurs with a router as part of the
2462 IO path, the NID of the last router that forwarded the bulk data
2463 is printed so it can be identified.
2468 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2469 Details : Due to a race condition, liblustre clients were occasionally
2470 evicted incorrectly.
2472 Severity : enhancement
2474 Description: lfs setstripe use optional parameters instead of postional
2477 Severity : enhancement
2479 Description: Nanosecond timestamp support for ldiskfs
2480 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2481 resolution timestamps. There is not yet support for this at
2482 the Lustre filesystem level.
2485 Frequency : during server recovery
2487 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2488 Details : request with rq_no_resend flag not awake l_wait_event if they get
2492 Frequency : nfs export on patchless client
2494 Description: connectathon hang when test nfs export over patchless client
2495 Details : Disconnected dentry cannot be found with lookup, so we do not need
2496 to unhash it or make it invalid
2499 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2500 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2501 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2506 Description: open req refcounting wrong on reconnect
2507 Details : If reconnect happened between getting open reply from server and
2508 call to mdc_set_replay_data in ll_file_open, we will schedule
2509 replay for unreferenced request that we are about to free.
2510 Subsequent close will crash in variety of ways.
2511 Check that request is still eligible for replay in
2512 mdc_set_replay_data().
2517 Description: disable writes to filesystem when reading health_check file
2518 Details : the default for reading the health_check proc file has changed
2519 to NOT do a journal transaction and write to disk, because this
2520 can cause reads of the /proc file to hang and block HA state
2521 checking on a healthy but otherwise heavily loaded system. It
2522 is possible to return to the previous behaviour during configure
2523 with --enable-health-write.
2525 Severity : enhancement
2527 Description: 64-bit inode version
2528 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2529 to the inode. This will be required for version-based recovery.
2534 Description: MDS fails to start if a duplicate client export is detected
2535 Details : in some rare cases it was possible for a client to connect to
2536 an MDS multiple times. Upon recovery the MDS would detect this
2537 and fail during startup. Handle this more gracefully.
2539 Severity : enhancement
2541 Description: Add -o localflock option to simulate old noflock
2543 Details : This will achieve local-only flock/fcntl locks
2549 Description: log_commit_thread vs filter_destroy race leads to crash
2550 Details : Take import reference before releasing llog record semaphore
2555 Description: Wrong request locking in request set processing
2556 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2557 imp_delayed_list, in this place should be used imp_lock.
2560 Frequency : when reconnection
2562 Description: Grant Leak when osc reconnect to OST
2563 Details : When osc reconnect ost, OST(filter) should check whether it
2564 should grant more space to client by comparing fed_grant and
2565 cl_avail_grant, and return the granted space to client instead
2566 of "new granted" space, because client will call osc_init_grant
2567 to update the client grant space info.
2570 Frequency : when client reconnect to OST
2572 Description: Grant Leak when osc do resend and replay bulk write
2573 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2574 bulk write request, because the grant info while be sync between
2575 OSC and OST when reconnect, and we should ignore the grant info
2576 these of resend/replay write req.
2581 Description: Grant space more than avaiable left space sometimes.
2582 Details : When then OST is about to be full, if two bulk writing from
2583 different clients came to OST. Accord the avaliable space of the
2584 OST, the first req should be permitted, and the second one
2585 should be denied by ENOSPC. But if the seconde arrived before
2586 the first one is commited. The OST might wrongly permit second
2587 writing, which will cause grant space > avaiable space.
2590 Frequency : when client is evicted
2592 Description: Grant might be wrongly erased when osc is evicted by OST
2593 Details : when the import is evicted by server, it will fork another
2594 thread ptlrpc_invalidate_import_thread to invalidate the
2595 import, where the grant will be set to 0. While the original
2596 thread will update the grant it got when connecting. So if
2597 the former happened latter, the grant will be wrongly errased
2598 because of this race.
2603 Description: Checking Stale with correct fid
2604 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2605 to check whether it is stale, because sometimes, we want the
2606 enqueue happened anyway, and op_data.fid2 will not be initialized.
2608 Severity : enhancement
2610 Description: update patchless client
2611 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2614 Frequency : only with 2.4 kernel
2616 Description: random memory corruption
2617 Details : size of struct ll_inode_info is to big for union inode.u and this
2618 can be cause of random memory corruption.
2623 Description: Memory leak in recovery
2624 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2625 It should also check obd_fail before fsfilt_start, otherwise if
2626 fsfilt_start return -EROFS,(failover mds during mds recovery).
2627 then the req will return with repmsg->transno = 0 and rc = EROFS.
2628 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2629 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2630 be freed no matter whether fsfilt_commit success or not.
2633 Frequency : only with huge count clients
2635 Description: Prevents from taking the superblock lock in llap_from_page for
2637 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2638 for llap_from_page call in ll_removepage prevents from taking the
2639 superblock lock for a soon died page.
2644 Description: Not check open intent error before release open handle
2645 Details : in some rare cases, the open intent error is not checked before
2646 release open handle, which may cause
2647 ASSERTION(open_req->rq_transno != 0), because it tries to release
2648 the failed open handle.
2653 Description: Set cat log bitmap only after create log success.
2654 Details : in some rare cases, the cat log bitmap is set too early. and it
2655 should be set only after create log success.
2659 Description: Accessing a block bevice can re-enable I/O when Lustre is
2660 tearing down a device.
2661 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2666 Description: service threads may hog cpus when there are a lot of requests
2667 Details : Insert cond_resched to give other threads a chance to use some CPU
2672 Description: the cat log was not initialized in recovery
2673 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2674 unlink log on mds will not be initialized until mds post
2675 recovery. And also in mds post recovery, the unlink log will
2676 initialization will be done asynchronausly, so there will be race
2677 between add unlink log and unlink log initialization.
2681 Description: brw_stats were being printed incorrectly
2682 Details : brw_stats were being printed as log2 but all of them were not
2683 recorded as log2. Also remove some code duplication arising from
2684 filter_tally_{read,write}.
2688 Frequency : rare, only in recovery.
2689 Description: ASSERTION(req->rq_type != LI_POISON) failed
2690 Details : imp_lock should be held while iterating over imp_sending_list for
2691 prevent destroy request after get timeout in ptlrpc_queue_wait.
2695 Description: replay-single.sh test 52 fails
2696 Details : A lock's skiplist need to be cleanup when it being unlinked
2697 from its resource list.
2701 Description: Short directio read returns full requested size rather than
2703 Details : Direct I/O operations should return actual amount of bytes
2704 transferred rather than requested size.
2706 Severity : enhancement
2708 Description: metadata RPC reduction (e.g. for rm performance)
2709 Details : decrease the amount of synchronous RPC between clients and servers
2710 by canceling conflicing lock before the operation on the client
2711 and packing thier handles into the main operation RPC to server.
2713 Severity : enhancement
2715 Description: Async OSC create to avoid the blocking unnecessarily.
2716 Details : If a OST has no remain object, system will block on the creating
2717 when need to create a new object on this OST. Now, ways use
2718 pre-created objects when available, instead of blocking on an
2719 empty osc while others are not empty. If we must block, we block
2720 for the shortest possible period of time.
2724 Description: improve handling recoverable errors
2725 Details : if request processig with error which can be recoverable on server
2726 request should be resend, otherwise page released from cache and
2729 Severity : enhancement
2731 Description: refine locking for avoid write wrong info into lov_objid file
2732 Details : fix possible races with add new target and write/update data in
2735 --------------------------------------------------------------------------------
2737 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2742 Frequency : on some architectures
2744 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2745 Details : Uninitialized flags sometimes cause configuration commands to
2749 Frequency : patchless clients only
2751 Description: missing __iget() symbol export
2752 Details : The __iget() symbol export is missing. To avoid the need for
2753 this on patchless clients the deathrow inode reaper is turned
2754 off, and we depend on the VM to clean up old inodes. This
2755 dependency was during via the fix for bug 12181.
2757 --------------------------------------------------------------------------------
2759 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2761 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2762 INTEROPERATE with older versions automatically. In many cases a
2763 special upgrade step is needed. Please read the
2764 user documentation before upgrading any part of a live system.
2765 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2766 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2767 * WARNING: Lustre configuration and startup changes are required with
2768 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2772 Severity : enhancement
2774 Description: MountConf
2775 Details : Lustre configuration is now managed via mkfs and mount
2776 commands instead of lmc and lconf. New obd types (MGS, MGC)
2777 are added for dynamic configuration management. See
2778 https://mail.clusterfs.com/wikis/lustre/MountConf for
2781 Severity : enhancement
2783 Description: dynamic OST addition
2784 Details : OSTs can now be added to a live filesystem
2786 Severity : enhancement
2788 Description: startup order invariance
2789 Details : MDTs and OSTs can be started in any order. Clients only
2790 require the MDT to complete startup.
2792 Severity : enhancement
2794 Description: parallel, asynchronous orphan cleanup
2795 Details : orphan cleanup is now performed in separate threads for each
2796 OST, allowing parallel non-blocking operation.
2798 Severity : enhancement
2800 Description: optimized stripe assignment
2801 Details : stripe assignments are now made based on ost space available,
2802 ost previous usage, and OSS previous usage, in order to try
2803 to optimize storage space and networking resources.
2805 Severity : enhancement
2807 Description: Permanently set tunables
2808 Details : All writable /proc/fs/lustre tunables can now be permanently
2809 set on a per-server basis, at mkfs time or on a live system.
2811 Severity : enhancement
2813 Description: Lustre message v2
2814 Details : Add lustre message format v2.
2816 Severity : enhancement
2818 Description: client OST exclusion list
2819 Details : Clients can be started with a list of OSTs that should be
2820 declared "inactive" for known non-responsive OSTs.
2824 Description: SPEC SFS validation failure on NFS v2 over lustre.
2825 Details : Changes the blocksize for regular files to be 2x RPC size,
2826 and not depend on stripe size.
2828 Severity : enhancement
2830 Description: Multiple MD RPCs in flight.
2831 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2832 To avoid overly-overloading MDS, introduce a limit on number of
2833 MDS RPCs in flight for a single client and add /proc controls
2834 to adjust this limit.
2836 Severity : enhancement
2838 Description: client read/write statistics
2839 Details : Add client read/write call usage stats for performance
2840 analysis of user processes.
2841 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2842 file access. extents_stats shows chunk size distribution.
2843 extents_stats_per_process show chunk size distribution per
2846 Severity : enhancement
2848 Description: mds statistics
2849 Details : Add detailed mds operations statistics in
2850 /proc/fs/lustre/mds/*/stats.
2854 Description: Failure of copying files with lustre special EAs.
2855 Details : Client side always return success for setxattr call for lustre
2856 special xattr (currently only "trusted.lov").
2861 Description: Refcount LNET uuids
2862 Details : The global LNET uuid list grew linearly with every startup;
2863 refcount repeated list entries instead of always adding to
2866 Severity : enhancement
2868 Description: Dynamic service threads
2869 Details : Within a small range, start extra service threads
2870 automatically when the request queue builds up.
2873 Frequency : mixed-endian client/server environments
2875 Description: mixed-endian crashes
2876 Details : The new msg_v2 system had some failures in mixed-endian
2879 Severity : enhancement
2881 Description: Easy OST removal
2882 Details : OSTs can be permanently deactivated with e.g. 'lctl
2883 conf_param lustre-OST0001.osc.active=0'
2885 Severity : enhancement
2887 Description: MGS proc entries
2888 Details : Added basic proc entries for the MGS showing what filesystems
2891 Severity : enhancement
2893 Description: provide MGS failover
2894 Details : Added config lock reacquisition after MGS server failover.
2896 Severity : enhancement
2898 Description: add Linux 2.4 support
2899 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2903 Description: a large application tries to do I/O to the same resource and dies
2904 in the middle of it.
2905 Details : Check the req->rq_arrival time after the call to
2906 ost_brw_lock_get(), but before we do anything about
2907 processing it & sending the BULK transfer request. This
2908 should help move old stale pending locks off the queue as
2909 quickly as obd_timeout.
2912 Frequency : when an incorrect nid is specified during startup
2914 Description: ptlrpc connect to non-existant node causes kernel crash
2915 Details : LNET can't be re-entered from an event callback, which
2916 happened when we expire a message after the export has been
2917 cleaned up. Instead, hand the zombie cleanup off to another
2920 Severity : enhancement
2922 Description: plain/inodebits lock performance improvement
2923 Details : Grouping plain/inodebits in granted list by their request modes
2924 and bits policy, thus improving the performance of search through
2928 Frequency : only if OST filesystem is corrupted
2930 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2931 Details : for a short time RPCs with bulk IO are in the replay list,
2932 but replay of bulk IOs is unimplemented. If the OST filesystem
2933 is corrupted due to disk cache incoherency and then replay is
2934 started it is possible to trip an assertion. Avoid putting
2935 committed RPCs into the replay list at all to avoid this issue.
2938 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2940 Description: System hang on startup
2941 Details : This bug allowed the liblustre (e.g. catamount) client to
2942 return to the app before handling all startup RPCs. This
2943 could leave the node unresponsive to lustre network traffic
2944 and manifested as a server ptllnd timeout.
2946 Severity : enhancement
2948 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2949 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2950 boolean to control whether to print peer debug info when a
2951 client's RPC times out.
2954 Frequency : only for kernels with patches from Lustre below 1.4.3
2956 Description: Remove old rdonly API
2957 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2960 Frequency : only for devices with external journals
2962 Description: Set external device read-only also
2963 Details : During a commanded failover stop, we set the disk device
2964 read-only while the server shuts down. We now also set any
2965 external journal device read-only at the same time.
2968 Frequency : when upgrading from 1.4 while trying to change parameters
2970 Description: The wrong (new) MDC name was used when setting parameters for
2971 upgraded MDT's. Also allows changing of OSC (and MDC)
2972 parameters if --writeconf is specified at tunefs upgrade time.
2975 Frequency : when setting specific ost indicies
2977 Description: QOS code breaks on skipped indicies
2978 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2979 created with --index need not be sequential.
2981 Severity : enhancement
2983 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2984 Details : The uninit_groups feature works in conjunction with the kernel
2985 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2986 up the pass1 processing of e2fsck. This is a read-only feature
2987 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2988 mount filesystems that have had this feature enabled.
2990 Severity : enhancement
2992 Description: Improve multi-block allocation algorithm to avoid fragmentation
2993 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2994 allocation locality and avoid filesystem fragmentation.
2996 ------------------------------------------------------------------------------
2998 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3000 * Support for kernels:
3001 2.4.21-47.0.1.EL (RHEL 3)
3002 2.6.5-7.283 (SLES 9)
3003 2.6.9-42.0.10.EL (RHEL 4)
3004 2.6.12.6 vanilla (kernel.org)
3005 2.6.16.27-0.9 (SLES 10)
3006 * Recommended e2fsprogs version: 1.39.cfs5
3008 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3012 Frequency : occasional, depends on client load and configuration
3013 Bugzilla : 12181, 12203
3014 Description: data loss for recently-modified files
3016 Details : In some cases it is possible that recently written or created
3017 files may not be written to disk in a timely manner (this should
3018 normally be within 30s unless client IO load is very high).
3019 The problem appears as zero-length files or files that are a
3020 multiple of 1MB in size after a client crash or client eviction
3021 that are missing data at the end of the file.
3023 This problem is more likely to be hit on clients where files are
3024 repeatedly created and unlinked in the same directory, clients
3025 have a large amount of RAM, have many CPUs, the filesystem has
3026 many OSTs, the clients are rebooted frequently, and/or the files
3027 are not accessed by other nodes after being written.
3029 The presence of the problem can be detected by looking at
3030 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3031 smaller than the second (nr_unused) then dirty files will not
3032 be flushed automatically to disk. "sync; sleep 10" should be
3033 run several times on the node before unmounting it to update
3034 Lustre (this is also safe to run on nodes without this problem).
3036 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3037 that can cause this same problem, so customers using that kernel
3038 also need to update the kernel in addition to Lustre. In order
3039 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3041 It is normal that files written just before a client crash (less
3042 than 30s) may not yet have been flushed to disk, even for local
3046 Frequency : frequent on thin XT3 nodes
3048 Description: UUID collision on thin XT3 Linux nodes
3049 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3050 randomly, since we relied on an insufficiently-seeded PRNG.
3055 Description: OSS hangs after "All ost request buffers busy"
3056 Details : A deadlock between quota and journal operations caused OSS
3057 hangs after printing "All ost request buffers busy."
3060 Frequency : always on liblustre builds
3062 Description: Cleanup compiler warnings on liblustre
3065 Frequency : always on liblustre builds on XT3
3067 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3071 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3072 > 32000 subdirectories
3073 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3074 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3075 subdirectory count crosses 32000. This will aid e2fsck to
3076 correctly handle more than 32000 subdirectories.
3079 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3081 Description: System hang on startup
3082 Details : This bug allowed the liblustre (e.g. catamount) client to
3083 return to the app before handling all startup RPCs. This
3084 could leave the node unresponsive to lustre network traffic
3085 and manifested as a server ptllnd timeout.
3087 Severity : enhancement
3089 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3090 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3091 boolean to control whether to print peer debug info when a
3092 client's RPC times out.
3097 Description: make O_SYNC working on 2.6 kernels
3098 Details : 2.6 kernels use different method for mark pages for write,
3099 so need add a code to lustre for O_SYNC work.
3104 Description: Failure to close file and release space on NFS
3105 Details : Put inode details into lock acquired in ll_intent_file_open.
3106 Use mdc_intent_lock in ll_intent_open to properly
3107 detect all kind of errors unhandled by mdc_enqueue.
3112 Description: proc file read during shutdown sometimes raced obd removal,
3114 Details : Add lock to prevent obd access after proc file removal.
3117 Frequency : Only for files larger than 4GB on 32-bit clients.
3119 Description: improperly doing page alignment of locks
3120 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3121 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3126 Description: RPCs being resent when they shouldn't be
3127 Details : Some RPCs that should not be resent are being resent. This
3128 can cause inconsistencies in the RPC state machine. Do not
3129 resend such requests.
3132 Frequency : rare, only with NFS export
3134 Description: Crash on NFS re-export node
3135 Details : under very unusual load conditions an assertion is hit in
3136 ll_intent_file_open()
3139 Frequency : only if OST filesystem is corrupted
3141 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3142 Details : for a short time RPCs with bulk IO are in the replay list,
3143 but replay of bulk IOs is unimplemented. If the OST filesystem
3144 is corrupted due to disk cache incoherency and then replay is
3145 started it is possible to trip an assertion. Avoid putting
3146 committed RPCs into the replay list at all to avoid this issue.
3151 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3152 Details : Large single O_DIRECT read and write calls can fail to allocate
3153 a sufficiently large buffer to process the request. In case of
3154 allocation failure the allocation is retried with a smaller
3155 buffer and broken into smaller requests.
3157 Severity : enhancement
3159 Description: Add -o localflock option to simulate old noflock behaviour.
3160 Details : This will achieve local-only flock/fcntl locks coherentness.
3165 Description: versioning check is incomplete
3166 Details : Checking the version difference of client vs. server, report
3167 error if the gap is too big.
3172 Description: add support PG_writeback bit
3173 Details : add support for PG_writeback bit for Lustre, for more carefull
3174 work with page cache in 2.6 kernel. This also fix some deadlocks
3175 and remove hack for work O_SYNC with 2.6 kernel.
3177 Severity : enhancement
3179 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3180 Details : The uninit_groups feature works in conjunction with the kernel
3181 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3182 up the pass1 processing of e2fsck. This is a read-only feature
3183 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3184 mount filesystems that have had this feature enabled.
3186 Severity : enhancement
3188 Description: Improve multi-block allocation algorithm to avoid fragmentation
3189 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3190 allocation locality and avoid filesystem fragmentation.
3192 ------------------------------------------------------------------------------
3194 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3196 * Support for kernels:
3197 2.6.9-42.0.3.EL (RHEL 4)
3198 2.6.5-7.276 (SLES 9)
3199 2.4.21-47.0.1.EL (RHEL 3)
3200 2.6.12.6 vanilla (kernel.org)
3201 2.6.16.21-0.8 (SLES10)
3202 * Recommended e2fsprogs version: 1.39.cfs2-0
3204 * The backwards-compatible /proc/sys/portals symlink has been removed
3205 in this release. Before upgrading, please ensure that you change
3206 any configuration scripts or /etc/sysctl.conf files that access
3207 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3208 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3209 in advance of the upgrade on any system running Lustre 1.4.6 or
3210 newer, since /proc/sys/lnet was added in that version.
3211 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3215 Frequency : only when quota is used
3217 Description: avoid scanning export list for quota master
3218 Details : Change the algorithms to avoid scanning export list in order
3219 to improve the efficiency.
3222 Frequency : MDS failover only, very rarely
3224 Description: "went back in time" messages on mds failover
3225 Details : The greatest transno may be lost when the current operation
3226 finishes with an error (transno==0) and the client's last_rcvd
3227 record is over-written. Save the greatest transno in the
3228 mds_last_transno for this case.
3231 Frequency : always for specific kernels and striping counts
3233 Description: client may get "Matching packet too big" without ACL support
3234 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3235 when trying to access files in certain configurations. The
3236 clients should in fact be denied when mounting because they do
3237 not understand ACLs.
3240 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3242 Description: many clients connecting with IO in progress causes connect timeouts
3243 Details : Avoid synchronous journal commits to avoid delays caused by many
3244 clients connecting/disconnecting when bulk IO is in progress.
3245 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3246 OST_IO_PORTAL to avoid delays behind potentially many pending
3250 Frequency : occasionally with multiple writers to a single file
3252 Description: shared writes to file may result in wrong size reported by stat()
3253 Details : Allow growing of kms when extent lock is cancelled
3256 Frequency : always with random mmap IO to multi-striped file
3258 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3259 Details : Only if the hole is at the end of OST object so that kms is too
3260 small. Fix is to increase kms accordingly in ll_nopage.
3263 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3265 Description: writes to a missing object would leak memory on the OST
3266 Details : If there is an inconsistency between the MDS and OST filesystems,
3267 such that the MDS references an object that doesn't exist, writes
3268 to that object will leak memory due to incorrect cleanup in the
3269 error handling path, eventually running out of memory on the OST.
3274 Description: Creating too long symlink causes lustre errors
3275 Details : Check symlink and name lengths before sending requests to MDS.
3278 Frequency : only if flock is enabled (not on by default)
3280 Description: posix locks not released on fd closure on 2.6.9+
3281 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3282 kernels, this caused such locks not to be released on fd close and
3283 then assertions on fs unmount about still used locks.
3286 Frequency : MDS failover only, very rarely
3288 Description: clients may get ASSERTION(granted_lock != NULL)
3289 Details : When request was taking a long time, and a client was resending
3290 a getattr by name lock request. The were multiple lock requests
3291 with the same client lock handle and
3292 mds_getattr_name->fixup_handle_for_resent_request found one of the
3293 lock handles but later failed with ASSERTION(granted_lock != NULL).
3298 Description: handle->h_buffer_credits > 0, assertion failure
3299 Details : h_buffer_credits is zero after truncate, causing assertion
3300 failure. This patch extends the transaction or creates a new
3304 Frequency : NFS re-export or patchless client
3305 Bugzilla : 11179, 10796
3306 Description: Crash on NFS re-export node (__d_move)
3307 Details : We do not want to hash the dentry if we don't have a lock.
3308 But if this dentry is later used in d_move, we'd hit uninitialised
3309 list head d_hash, so we just do this to init d_hash field but
3310 leave dentry unhashed.
3313 Frequency : NFS re-export or patchless client
3315 Description: NFS exports has problem with symbolic link
3316 Details : lustre client didn't properly install dentry when re-exported
3317 to NFS or running patchless client.
3320 Frequency : NFS re-export or patchless client
3322 Description: Various nfs/patchless fixes.
3323 Details : fixes reuse disconected alias for lookup process - this fixes
3324 warning "find_exported_dentry: npd != pd",
3325 fix permission error with open files at nfs.
3326 fix apply umask when do revalidate.
3329 Frequency : occasional
3331 Description: Crash on NFS re-export node
3332 Details : calling clear_page() on the wrong pointer triggered oops in
3333 generic_mapping_read().
3336 Frequency : rarely, using O_DIRECT IO
3338 Description: unaligned directio crashes client with LASSERT
3339 Details : check for unaligned buffers before trying any requests.
3342 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3344 Description: stale data returned from RAID cache
3345 Details : If only a small amount of IO is done to the RAID device before
3346 reading it again it is possible to get stale data from the RAID
3347 cache instead of reading it from disk.
3350 Frequency : always for sles10 kernel
3352 Description: sles10 support
3353 Details : ll_follow_link: compile fixes and using of nd_set_link
3354 under newer kernels.
3357 Frequency : depends on arch, kernel and compiler version, always on sles10
3360 Description: recursive or deep enough symlinks cause stack overflow
3361 Details : getting rid of large stack-allocated variable in
3365 Frequency : depends on hardware
3367 Description: lustre write performance loss in the SLES10 kernel
3368 Details : the performance loss is caused by using of write barriers in the
3369 ext3 code. The SLES10 kernel turns barrier support on by
3370 default. The fix is to undo that change for ldiskfs.
3372 ------------------------------------------------------------------------------
3374 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3376 * Support for kernels:
3377 2.6.9-42.0.3EL (RHEL 4)
3378 2.6.5-7.276 (SLES 9)
3379 2.4.21-47.0.1.EL (RHEL 3)
3380 2.6.12.6 vanilla (kernel.org)
3384 Frequency : quota enabled and large files being deleted
3386 Description: releasing more than 4GB of quota at once hangs OST
3387 Details : If a user deletes more than 4GB of files on a single OST it
3388 will cause the OST to spin in an infinite loop. Release
3389 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3394 Description: statfs data retrieved from /proc may be stale or zero
3395 Details : When reading per-device statfs data from /proc, in the
3396 {kbytes,files}_{total,free,avail} files, it may appear
3397 as zero or be out of date.
3400 Frequency : systems with MD RAID1 external journal devices
3402 Description: lconf's call to blkid is confused by RAID1 journal devices
3403 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3404 of returning all block devices that match the journal UUID.
3407 Frequency : always, for aggregate stripe size over 4GB
3409 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3410 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3411 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3414 Frequency : always if "lfs find" used on a local file/directory
3416 Description: "lfs find" segfaults if used on a local file/directory
3417 Details : The case where a directory component was not specified wasn't
3418 handled correctly. Handle this properly.
3421 Frequency : always on ppc64
3423 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3424 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3425 perform endianness conversion.
3428 Frequency : rarely (truncate to non-zero file size after write under load)
3429 Bugzilla : 10730, 10687
3430 Description: Files padded with zeros to next 4K multiple
3431 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3432 it is possible that files that are truncated to a non-zero size
3433 immediately after being written are filled with zero bytes beyond
3434 the truncated size. No file data is lost.
3436 Severity : enhancement
3438 Description: Allow recovery/failover for liblustre clients.
3439 Details : liblustre clients were unaware of failover configurations until
3442 Severity : enhancement
3444 Description: user file locks should fail when not mounting with flock option
3445 Details : Set up an error-returning stub in ll_file_operations.lock field
3446 to prevent incorrect behaviour when client is mounted without
3447 flock option. Also, set up properly f_op->flock field for
3451 Frequency : always on ia64
3453 Description: "lfs df" loops on printing out MDS statfs information
3454 Details : The obd_ioctl_data was not initialized and in some systems
3455 this caused a failure during the ioctl that did not return
3456 an error. Initialize the struct and return an error on failure.
3459 Frequency : SLES 9 only
3461 Description: Error of copying files with lustre special EAs as root
3462 Details : Client side always return success for setxattr call for lustre
3463 special xattr (currently only "trusted.lov").
3466 Frequency : rarely on clusters with both ia64+i386 clients
3468 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3469 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3470 clients are doing concurrent writes to the same file it is
3471 possible that the ia64 clients may LASSERT because the OST
3472 extent locks are not PAGE_SIZE aligned. Ensure that grown
3473 locks are always aligned on the request boundary.
3476 Frequency : specific use, occasional
3478 Description: Overwriting in use executable truncates on-disk binary image
3479 Details : If one node attempts to overwrite an executable in use by
3480 another node, we now correctly return ETXTBSY instead of
3481 truncating the file.
3483 Severity : enhancement
3485 Description: Async OSC create to avoid the blocking unnecessarily.
3486 Details : If a OST has no remain object, system will block on the creating
3487 when need to create a new object on this OST. Now, ways use
3488 pre-created objects when available, instead of blocking on an
3489 empty osc while others are not empty. If we must block, we block
3490 for the shortest possible period of time.
3495 Description: chmod on Lustre root is propagated to other clients
3496 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3497 invalid by the follow_mount time.
3500 Frequency : liblustre clients only
3502 Description: Race in 'instant cancel' lock handling could lead to such locks
3503 never to be granted in case of SMP MDS
3504 Details : Do not destroy not yet granted but cbpending locks in
3508 Frequency : replay/resend of open
3510 Description: non null lock assetion failure in mds_intent_policy
3511 Details : Trying to replay/resend lockless open requests resulted in
3512 mds_open() returning 0 with no lock. Now it sets a flag if
3513 a lock is going to be returned.
3515 Severity : enhancement
3517 Description: Checksum enhancements
3518 Details : New checksum enhancements allow for resending RPCs that failed
3521 Severity : enhancement
3523 Description: Tunables on number of dirty pages in cacche
3524 Details : Allow to set limit on number of dirty pages cached.
3529 Description: client crash on unmount - lock still has references
3530 Details : In some error handling cases it was possible to leak a lock
3531 reference on a client while accessing a file. This was not
3532 harmful to the client during operation, but would cause the
3533 client to crash when the filesystem is unmounted.
3536 Frequency : specific case, rare
3538 Description: ETXTBSY on mds though file not in use by client
3539 Details : ETXTBSY is no longer incorrectly returned when attempting to
3540 chmod or chown a directory that the user previously tried to
3541 execute or a currently-executing binary.
3544 Frequency : extremely rare except on liblustre-based clients
3546 Description: Lustre space not freed when files are deleted
3547 Details : Clean up open-unlinked files after client eviction. Previously
3548 the unlink was skipped and the files remained as orphans.
3553 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3554 Details : In some cases it was possible to send a blocking callback to
3555 a client doing a glimpse, even though that client didn't get
3556 a lock granted. When the glimpse lock is cancelled on the OST
3557 the freed lock is left on the waiting list and corrupted the list.
3560 Frequency : all core dumps
3562 Description: Broke core dumps to lustre
3563 Details : Negative dentry may be unhashed if parent does not have UPDATE
3564 lock, but some callers, e.g. do_coredump, expect dentry to be
3565 hashed after successful create, hash it in ll_create_it.
3567 ------------------------------------------------------------------------------
3569 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3571 * Support for kernels:
3572 2.6.9-42.0.2.EL (RHEL 4)
3573 2.6.5-7.276 (SLES 9)
3574 2.4.21-40.EL (RHEL 3)
3575 2.6.12.6 vanilla (kernel.org)
3579 Frequency : always on RHEL 3
3581 Description: Number of open files grows over time
3582 Details : The number of open files grows over time, whether or not
3583 Lustre is started. This was due to a filp leak introduced
3584 by one of our kernel patches.
3586 ------------------------------------------------------------------------------
3588 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3590 * Support for kernels:
3591 2.6.9-42.EL (RHEL 4)
3592 2.6.5-7.276 (SLES 9)
3593 2.4.21-40.EL (RHEL 3)
3594 2.6.12.6 vanilla (kernel.org)
3599 Bugzilla : 5719, 9635, 9792, 9684
3600 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3601 Details : If a server is under heavy load and cannot reply to new
3602 connection requests before the client resends the (re)connect,
3603 the connection handling code can behave badly if two service
3604 threads are concurrently handing separate (re)connections from
3605 the same client. Add better locking to the connection handling
3606 code, and ensure that only a single connection will be processed
3607 for a given client UUID, even if the lock is dropped.
3609 Severity : enhancement
3611 Description: add TCP zero-copy support to kernel
3612 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3613 sends if the hardware supports scatter-gather and checksumming.
3614 This allows socklnd to do client-write and server-read more
3615 efficiently and reduce CPU utilization from skbuf copying.
3618 Frequency : only if NFS exporting from client
3620 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3621 Details : If setting attributes on a file created under NFS that had
3622 never been opened it would be possible to oops the client
3623 if the file had no objects.
3626 Frequency : always for liblustre
3628 Description: liblustre client does MDS+OSTs setattr RPC for each write
3629 Details : When doing a write from a liblustre client, the client
3630 incorrectly issued an RPC to the MDS and each OST the file was
3631 striped over in order to update the timestamps. When writing
3632 with small chunks and many clients this could overwhelm the MDS
3633 with RPCs. In all cases it would slow down the write because
3634 these RPCs are unnecessary.
3636 Severity : enhancement
3638 Description: allow number of MDS service threads to be changed at module load
3639 Details : It is now possible to change the number of MDS service threads
3640 running. Adding "options mds mds_num_threads={N}" to the MDS's
3641 /etc/modprobe.conf will set the number of threads for the next
3642 time Lustre is restarted (assuming the "mds" module is also
3643 reloaded at that time). The default number of threads will
3644 stay the same, 32 for most systems.
3649 Description: OST crash if filesystem is unformatted or corrupt
3650 Details : If an OST is started on a device that has never been formatted
3651 or if the filesystem is corrupt and cannot even mount then the
3652 error handling cleanup routines would dereference a NULL pointer.
3657 Description: NULL pointer deref in llap_from_page.
3658 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3659 incorrect) mapping if the page was truncated/reclaimed while it was
3660 searched for. Check for this condition and skip such pages when
3661 doing readahead. Introduce extra check to llap_from_page() to
3662 verify page->mapping->host is non-NULL (so page is not anonymous).
3665 Frequency : Sometimes when using sys_sendfile
3667 Description: "page not covered by a lock" warnings from ll_readpage
3668 Details : sendfile called ll_readpage without right page locks present.
3669 Now we introduced ll_file_sendfile that does necessary locking
3670 around call to generic_file_sendfile() much like we do in
3674 Frequency : with certain MDS communication failures at client mount time
3676 Description: NULL pointer deref after failed client mount
3677 Details : a client connection request may delayed by the network layer
3678 and not be sent until after the PTLRPC layer has timed out the
3679 request. If the client fails the mount immediately it will try
3680 to clean up before the network times out the request. Add a
3681 reference from the request import to the obd device and delay
3682 the cleanup until the network drops the request.
3685 Frequency : occasionally during client (re)connect
3687 Description: assertion failure during client (re)connect
3688 Details : processing a client connection request may be delayed by the
3689 client or server longer than the client connect timeout. This
3690 causes the client to resend the connection request. If the
3691 original connection request is replied in this interval, the
3692 client may trip an assertion failure in ptlrpc_connect_interpret()
3693 which thought it would be the only running connect process.
3696 Frequency : only with obd_echo servers and clients that are rebooted
3698 Description: kernel BUG accessing uninitialized data structure
3699 Details : When running an obd_echo server it did not start the ping_evictor
3700 thread, and when a client was evicted an uninitialized data
3701 structure was accessed. Start the ping_evictor in the RPC
3702 service startup instead of the OBD startup.
3704 Severity : enhancement
3705 Bugzilla : 10193 (patchless)
3706 Description: Remove dependency on various unexported kernel interfaces.
3707 Details : No longer need reparent_to_init, exit_mm, exit_files,
3708 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3711 Frequency : rare (only users of deprecated and unsupported LDAP config)
3713 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3714 Details : LDAP apparently contains 'lustreName' attributes instead of
3715 'name'. A simple remapping of the name is sufficient.
3718 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3720 Description: waiting_locks_callback trips kernel BUG if client is evicted
3721 Details : Running with the dump_on_timeout debug flag turned on makes
3722 it possible that the waiting_locks_callback() can try to dump
3723 the Lustre kernel debug logs from an interrupt handler. Defer
3724 this log dumping to the expired_lock_main() thread.
3726 Severity : enhancement
3728 Description: Support NFS exporting on 2.6 kernels.
3729 Details : Implement non-rawops metadata methods for NFS server to use without
3730 changing NFS server code.
3733 Frequency : very rare (synthetic metadata workload only)
3735 Description: two racing renames might cause an MDS thread to deadlock
3736 Details : Running the "racer" program may cause one MDS thread to rename
3737 a file from being the source of a rename to being the target of
3738 a rename at exactly the same time that another thread is doing
3739 so, and the second thread has already enqueued these locks after
3740 doing a lookup of the target and is trying to relock them in
3741 order. Ensure that we don't try to re-lock the same resource.
3744 Frequency : only very large systems with liblustre clients
3746 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3747 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3748 all OSTs in parallel. This allows the eviction of stale liblustre
3749 clients to proceed much faster than if they were done in series,
3750 and also offers similar improvements for other set_info RPCs.
3755 Description: excessive CPU usage during initial read phase on client
3756 Details : During the initial read phase on a client, it would agressively
3757 retry readahead on the file, consuming too much CPU and impacting
3758 performance (since 1.4.5.8). Improve the readahead algorithm
3759 to avoid this, and also improve some other common cases (read
3760 of small files in particular, where "small" is files smaller than
3761 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3766 Description: MDS crash when receiving packet with unknown intent.
3767 Details : Do not LBUG in unknown intent case, just return -EFAULT
3769 Severity : enhancement
3770 Bugzilla : 9293, 9385
3771 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3772 Details : Do not serialize getattr (non-intent version) and statfs.
3775 Frequency : occasional, when OST network is overloaded/intermittent
3777 Description: client evicted by OST after bulk IO timeout
3778 Details : If a client sends a bulk IO request (read or write) the OST
3779 may evict the client if it is unresposive to its data GET/PUT
3780 request. This is incorrect if the network is overloaded (takes
3781 too long to transfer the RPC data) or dropped the OST GET/PUT
3782 request. There is no need to evict the client at all, since
3783 the pinger and/or lock callbacks will handle this, and the
3784 client can restart the bulk request.
3787 Frequency : Always when mmapping file with no objects
3789 Description: client crashes when mmapping file with no objects
3790 Details : Check that we actually have objects in a file before doing any
3791 operations on objects in ll_vm_open, ll_vm_close and
3797 Description: Request leak when working with deleted CWD
3798 Details : Introduce advanced request refcount tracking for requests
3799 referenced from lustre intent.
3801 Severity : Enhancement
3803 Description: Cache open file handles on client.
3804 Details : MDS now will return special lock along with openhandle, if
3805 requested and client is allowed to hold openhandle, even if unused,
3806 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3807 closing files for every read/write openration.
3809 Severity : Enhancement
3811 Description: Cache open negative dentries on client when possible.
3812 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3813 negative dentries on lock revocation.
3818 Description: Remounting a client read-only wasn't possible with a zconf mount
3819 Details : It wasn't possible to remount a client read-only with llmount.
3821 Severity : enhancement
3822 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3823 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3824 MPICH is included to add Lustre-specific ADIO interfaces.
3825 This is based closely on the UFS ADIO layer and only differs
3826 in file creation, in order to allow the OST striping to be set.
3827 This is user-contributed code and not supported by CFS.
3832 Description: extended inode attributes (immutable, append-only) work improperly
3833 when 2.4 and 2.6 kernels are used on client/server or vice versa
3834 Details : Introduce kernel-independent values for these flags.
3836 Severity : enhancement
3839 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3840 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3841 granularity. Now, Fractional MB granularity is be supported,
3842 this is very useful for low memory system.
3844 Severity : enhancement
3846 Description: Getattr by fid
3847 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3848 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3849 have locks and avoids getattr rpc requests for it on every stat.
3852 Frequency : Always, for filesystems larger than 2TB
3854 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3855 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3856 filesystems larger than 2TB to be mounted with mballoc enabled.
3859 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3861 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3862 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3863 filesystems larger than 2TB is mounted then the kernel will
3864 silently corrupt the start of the filesystem. CONFIG_LBD is
3865 enabled for all CFS-supported kernels, but the possibility of
3866 this happening with a modified kernel config exists.
3868 Severity : enhancement
3870 Description: add client O_DIRECT support for 2.6 kernels
3871 Details : It is now possible to do O_DIRECT reads and writes to files
3872 in the Lustre client mountpoint on 2.6 kernel clients.
3874 Severity : enhancement
3876 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3877 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3878 parallel, not waiting for response from every OST before sending
3879 a rpc to the next OST.
3884 Description: setattr vs write race when updating file timestamps
3885 Details : Client processes that update a file timestamp into the past
3886 right after writing to the file (e.g. tar) it is possible that
3887 the updated file modification time can be reset to the current
3888 time due to a race between processing the setattr and write RPC.
3890 Severity : enhancement
3892 Description: Bring 'lfs find' closer in line with regular Linux find.
3893 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3894 -print0 options and obtains all the needed info through the lustre
3897 Severity : enhancement
3899 Description: support up to 1024 configured devices on one node
3900 Details : change obd_dev array from statically allocated to dynamically
3901 allocated structs as they are first used to reduce memory usage
3906 Description: Flush dirty partially truncated pages during truncate
3907 Details : Immediatelly flush partially truncated pages in filter_setattr,
3908 this way we completely avoid having any pages in page cache on OST
3909 and can retire ugly workarounds during writes to flush such pages.
3914 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3915 Details : protect inode from truncation within vfs_unlink() context
3916 just take a reference before calling vfs_unlink() and release it
3917 when parent's i_sem is free.
3922 Description: last_id value checked outside lock on OST caused LASSERT failure
3923 Details : If there were multiple MDS->OST object precreate requests in
3924 flight, it was possible that the OST's last object id was checked
3925 outside a lock and incorrectly tripped an assertion. Move checks
3926 inside locks, and discard old precreate requests.
3929 Frequency : always, if extents are used on OSTs
3931 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3932 Details : OSTs using the extents format would not zero the high 16 bits of
3933 the index physical block number. This is not a problem for any
3934 OST filesystems smaller than 16TB, and no kernels support ext3
3935 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3936 new/modified files) and can be fixed for existing filesystems
3937 with e2fsprogs-1.39-cfs1.
3942 Description: import connection selection may be incorrect if timer wraps
3943 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3944 import connections to be ignored if the 32-bit jiffies counter
3945 wraps. Use a 64-bit jiffies counter.
3948 Frequency : very large clusters immediately after boot
3950 Description: LNET request buffers exhausted under heavy short-term load
3951 Details : If a large number of client requests are generated on a service
3952 that has previously never seen so many requests it is possible
3953 that the request buffer growth cannot keep up with the spike in
3954 demand. Instead of dropping incoming requests, they are held in
3955 the LND until the RPC service can accept more requests.
3958 Frequency : Sometimes during replay
3960 Description: Assertion failure in ll_local_open after replay.
3961 Details : If replay happened on an open request reply before we were able
3962 to set replay handler, reply will become not swabbed tripping the
3963 assertion in ll_local_open. Now we set the handler right after
3964 recognising of open request
3967 Frequency : very rare
3969 Description: kernel reports "badness in vsnprintf"
3970 Details : Reading from the "recovery_status" /proc file in small chunks
3971 may cause a negative length in lprocfs_obd_rd_recovery_status()
3972 call to vsnprintf() (which is otherwise harmless). Exit early
3973 if there is no more space in the output buffer.
3975 Severity : enhancement
3977 Description: clear OBD RPC statistics by writing to them
3978 Details : It is now possible to clear the OBD RPC statistics by writing
3979 to the "stats" file.
3984 Description: Client mtime is not the same on different clients after utimes
3985 Details : In some cases, the client was using the utimes() syscall on
3986 a file cached on another node. The clients now validate the
3987 ctime from the MDS + OSTs to determine which one is right.
3992 Description: Inability to activate failout mode
3993 Details : lconf script incorrectly assumed that in pythong string's numeric
3994 value is used in comparisons.
3997 Frequency : always with multiple stripes per file
3999 Description: Inefficient object allocation for mutli-stripe files
4000 Details : When selecting which OSTs to stripe files over, for files with
4001 a stripe count that divides evenly into the number of OSTs,
4002 the MDS is always picking the same starting OST for each file.
4003 Return the OST selection heuristic to the original design.
4008 Description: mount failures may take full timeout to return an error
4009 Details : Under some heavy load conditions it is possible that a
4010 failed mount can wait for the full obd_timeout interval,
4011 possibly several minutes, before reporting an error.
4012 Instead return an error as soon as the status is known.
4014 Frequency : quota enabled and large files being deleted
4016 Description: releasing more than 4GB of quota at once hangs OST
4017 Details : If a user deletes more than 4GB of files on a single OST it
4018 will cause the OST to spin in an infinite loop. Release
4019 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4024 Description: statfs data retrieved from /proc may be stale or zero
4025 Details : When reading per-device statfs data from /proc, in the
4026 {kbytes,files}_{total,free,avail} files, it may appear
4027 as zero or be out of date.
4030 Frequency : systems with MD RAID1 external journal devices
4032 Description: lconf's call to blkid is confused by RAID1 journal devices
4033 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4034 of returning all block devices that match the journal UUID.
4037 Frequency : always, for aggregate stripe size over 4GB
4039 Description: assertion fails when trying to use 4GB stripe size
4040 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4041 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4044 Frequency : always on ppc64
4046 Description: the first write on an ext3 filesystem with mballoc got stuck
4047 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4048 perform endianness conversion.
4050 ------------------------------------------------------------------------------
4052 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4054 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4055 INTEROPERATE with older versions automatically. Please read the
4056 user documentation before upgrading any part of a live system.
4057 * WARNING: Lustre networking configuration changes are required with
4058 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4061 * Support for kernels:
4062 2.6.9-22.0.2.EL (RHEL 4)
4063 2.6.5-7.244 (SLES 9)
4064 2.6.12.6 vanilla (kernel.org)
4067 Severity : enhancement
4068 Bugzilla : 7981/8208
4069 Description: Introduced Lustre Networking (LNET)
4070 Details : LNET is new networking infrastructure for Lustre, it includes
4071 a reorganized network configuration mode (see the user
4072 documentation for full details) as well as support for routing
4073 between different network fabrics. Lustre Networking Devices
4074 (LNDs) for the supported network fabrics have also been
4075 created for this new infrastructure.
4077 Severity : enhancement
4078 Description: Introduced Access control lists
4079 Details : clients can set ACLs on files and directories in order to have
4080 more fine-grained permissions than the standard Unix UGO+RWX.
4081 The MDS must be started with the "-o acl" mount option.
4083 Severity : enhancement
4084 Description: Introduced filesystem quotas
4085 Details : Administrators may now establish per-user quotas on the
4088 Severity : enhancement
4090 Description: Configuration change for the XT3
4091 The PTLLND is now used to run Lustre over Portals on the XT3
4092 The configure option(s) --with-cray-portals are no longer used.
4093 Rather --with-portals=<path-to-portals-includes> is used to
4094 enable building on the XT3. In addition to enable XT3 specific
4095 features the option --enable-cray-xt3 must be used.
4100 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4101 Details : A race between allocating a new llcd and re-getting the llcd_lock
4102 allowed another thread to grab newly-allocated llcd.
4104 Severity : enhancement
4106 Description: 2.6 OST async journal commit and locking fix to improve performance
4107 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4108 2.6 kernels are now async as they already were in 2.4 kernels so
4109 that they can commit concurrently with the network bulk transfer.
4110 For block-allocated files the filter allocation semaphore is held
4111 to avoid filesystem fragmentation during allocation. BKL lock
4112 removed for 2.6 xattr operations where it is no longer needed.
4117 Description: lconf incorrectly determined whether two IP networks could talk
4118 Details : In some more complicated routing and multiple-network
4119 configurations, lconf will avoid trying to make a network
4120 connection to a disjoint part of the IP space. It was doing the
4121 math incorrectly for one set of cases.
4126 Description: Fix for potential infinite loop processing records in an llog.
4127 Details : If an llog record is corrupted/zeroed, it is possible to loop
4128 forever in llog_process(). Validate the llog record length
4129 and skip the remainder of the block on error.
4132 Frequency : occasional (liblustre only)
4134 Description: liblustre could not open files whose last component is a symlink
4135 Details : sysio_path_walk() would incorrectly pass the open intent to
4136 intermediate path components.
4139 Frequency : rare (liblustre only with non-standard tuning)
4140 Bugzilla : 7201 (7350)
4141 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4142 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4143 after it has been released, only do this for referenced locks
4145 Severity : enhancement
4147 Description: specifying an (invalid) directory default stripe_size of -1
4148 would reset the directory default striping
4149 Details : stripe_size -1 was used internally to signal directory stripe
4150 removal, now use "all default" to signal dir stripe removal
4151 as a directory striping of "all default" is not useful
4154 Frequency : common for large clusters running liblustre clients
4156 Description: doing an ls when liblustre clients are running is slow
4157 Details : sending a glimpse AST to a liblustre client waits for every AST
4158 to time out, as liblustre clients will not respond. Since they
4159 cannot cache data we refresh the OST lock LVB from disk instead.
4161 Severity : enhancement
4163 Description: doing an ls at the same time as file IO can be slow
4164 Details : enqueue and other "small" requests can be blocked behind many
4165 large IO requests. Create a new OST IO portal for non-IO
4166 requests so they can be processed faster.
4169 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4171 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4172 created file with mode 000
4173 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4174 allow an ftruncate() as a non-root user to a file with mode 000.
4175 Lustre now respects this flag to disable mode checks when
4176 truncating a file owned by the user
4179 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4182 Description: Revoking locks from clients that went dead or catatonic might take
4184 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4185 cancellation of such locks instant on servers without waiting for
4186 any reply from clients. Clients drops these locks when cancel
4187 notification from server is received without replying.
4190 Frequency : liblustre-only, when liblustre client dies or becomes busy
4192 Description: Doing ls on Linux clients can take a long time with active
4194 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4195 granting such locks to it in the first place if possible. Locks
4196 are taken by proxy on the OST during the read or write and
4197 dropped immediately afterward. Add connect flags handling, do
4198 not grant locks to liblustre clients for glimpse ASTs.
4200 Severity : enhancement
4202 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4203 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4204 sequential reads and behaves badly with applications doing random
4205 reads. Improve it by reading ahead at least read region, and
4206 avoiding excessive large RPC for small reads.
4208 Severity : enhancement
4210 Description: Creating more than 1000 files for a single job may cause a load
4211 imbalance on the OSTs if there are also a large number of OSTs.
4212 Details : qos_prep_create() uses an OST index reseed value that is an
4213 even multiple of the number of available OSTs so that if the
4214 reseed happens in the middle of the object allocation it will
4215 still utilize the OSTs as uniformly as possible.
4220 Description: OST or MDS may oops in ping_evictor_main()
4221 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4222 but doesn't restart at beginning of obd_exports_timed list
4225 Severity : enhancement
4227 Description: improve by-nid export eviction on the MDS and OST
4228 Details : allow multiple exports with the same NID to be evicted at one
4229 time without re-searching the exports list.
4232 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4234 Description: MDS may oops in groups_free()
4235 Details : in rare race conditions a newly allocated group_info struct is
4236 freed again, and this can be NULL. The 2.4 compatibility code
4237 for groups_free() checked for a NULL pointer, but 2.6 did not.
4240 Frequency : common for liblustre clients doing little filesystem IO
4241 Bugzilla : 9352, 7313
4242 Description: server may evict liblustre clients accessing contended locks
4243 Details : if a client is granted a lock or receives a completion AST
4244 with a blocking AST already set it would not reply to the AST
4245 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4248 Frequency : lfs setstripe, only systems with more than 160 OSTs
4250 Description: unable to set striping with a starting offset beyond OST 160
4251 Details : llapi_create_file() incorrectly limited the starting stripe
4252 index to the maximum single-file stripe count.
4255 Frequency : LDAP users only
4257 Description: lconf did not handle in-kernel recovery with LDAP properly
4258 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4260 Severity : enhancement
4262 Description: bind OST threads to NUMA nodes to improve performance
4263 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4264 node and do their allocations there to localize memory access
4266 Severity : enhancement
4268 Description: llmount can determine client NID directly from Myrinet (GM)
4269 Details : the client NID code from gmnalnid was moved directly into
4270 llmount, removing the need to use this or specifying the
4271 client NID explicitly when mounting GM clients with zeroconf
4274 Frequency : if client is started with down MDS
4276 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4277 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4278 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4279 ptlrpc_import_delayed_req() because we didn't check if the
4280 request was interrupted, and we also didn't break out of the
4281 event loop if there was no timeout
4286 Description: data loss during non-page-aligned writes to a single file from
4287 both multiple nodes and multiple threads on one node at same time
4288 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4289 inconsistency led to false short-reads, that were cached and later
4290 used by ->prepare_write() to fill in partially written page,
4291 leading to data loss.
4294 Frequency : always, if lconf --abort_recovery used
4296 Description: lconf --abort_recovery fails with 'Operation not supported'
4297 Details : lconf was attempting to abort recovery on the MDT device and not
4300 Severity : enhancement
4302 Description: remove cleanup logs
4303 Details : replace lconf-generated cleanup logs with lustre internal
4304 cleanup routines. Eliminates the need for client-cleanup and
4307 Severity : enhancement
4309 Description: add support for EAs (user and system) on lustre filesystems
4310 Details : it is now possible to store extended attributes in the Lustre
4311 client filesystem, and with the user_xattr mount option it
4312 is possible to allow users to store EAs on their files also
4314 Severity : enhancement
4316 Description: Add possibility (config option) to show minimal available OST free
4318 Details : When compiled with --enable-mindf configure option, statfs(2)
4319 (and so, df) will return least minimal free space available from
4320 all OSTs as amount of free space on FS, instead of summary of
4321 free spaces of all OSTs.
4323 Severity : enhancement
4325 Description: do not expand extent locks acquired on OST-side
4326 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4327 by server: they are not cached anyway.
4330 Frequency : when mmap is used/binaries executed from Lustre
4332 Description: Unmmap pages before throwing them away from read cache.
4333 Details : llap_shrink cache now attempts to unmap pages before discarding
4334 them (if unmapping failed - do not discard). SLES9 kernel has
4335 extra checks that trigger if this unmapping is not done first.
4340 Description: lconf didn't resolve symlinks before checking to see whether a
4341 given mountpoint was already in use
4344 Frequency : when migrating failover services
4345 Bugzilla : 6395, 9514
4346 Description: When migrating a subset of services from a node (e.g. failback
4347 from a failover service node) the remaining services would
4348 time out and evict clients.
4349 Details : lconf --force (implied by --failover) sets the global obd_timeout
4350 to 5 seconds in order to quickly disconnect, but this caused
4351 other RPCs to time out too quickly. Do not change the global
4352 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4354 Severity : enhancement
4355 Frequency : if MDS is started with down OST
4356 Bugzilla : 9439,5706
4357 Description: Allow startup/shutdown of an MDS without depending on the
4358 availability of the OSTs.
4359 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4360 Add appropriate locking and lov-osc refcounts for safe
4361 cleaning. Add osc abort_inflight calls in case the
4362 synchronize never started.
4365 Frequency : occasional (Cray XT3 only)
4367 Description: root not authorized to access files in CRAY_PORTALS environment
4368 Details : The client process capabilities were not honoured on the MDS in
4369 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4370 been accessed by an authorized user then root was able to access
4371 the file on the local client also. The root user capabilities
4372 are now allowed on the MDS, as this environment has secure UID.
4375 Frequency : occasional
4377 Description: ldiskfs "too long searching" message happens too often
4378 Details : A debugging message (otherwise harmless) prints too often on
4379 the OST console. This has been reduced to only happen when
4380 there are fragmentation problems on the filesystem.
4385 Description: Division by zero in statfs when all OSCs are inactive
4386 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4387 lov_statfs divides by value returned by lov_get_stripecnt().
4391 Bugzilla : 9489, 3273
4392 Description: First write from each client to each OST was only 4kB in size,
4393 to initialize client writeback cache, which caused sub-optimal
4394 RPCs and poor layout on disk for the first writen file.
4395 Details : Clients now request an initial cache grant at (re)connect time
4396 and so that they can start streaming writes to the cache right
4397 away and always do full-sized RPCs if there is enough data.
4398 If the OST is rebooted the client also re-establishes its grant
4399 so that client cached writes will be honoured under the grant.
4404 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4405 Details : Now I/O RPCs go to different portal number and (presumably) fast
4406 lock requests (and glimses) and other RPCs get their own service
4407 threads pool that should be able to service those RPCs
4410 Severity : enhancement
4412 Description: Ability to exchange lustre version between client and servers and
4413 issue warnings at client side if client is too old. Also for
4414 liblustre clients there is ability to refuse connection of too old
4416 Details : New 'version' field is added to connect data structure that is
4417 filled with version info. That info is later checked by server and
4421 Frequency : rare, liblustre only.
4422 Bugzilla : 9296, 9581
4423 Description: Two simultaneous writes from liblustre at offset within same page
4424 might proceed at the same time overwriting eachother with stale
4426 Details : I/O lock withing llu_file_prwv was released too early, before data
4427 actually was hitting the wire. Extended lock-holding time until
4428 server acknowledges receiving data.
4431 Frequency : extremely rare. Never observed in practice.
4433 Description: avoid generating lustre_handle cookie of 0.
4434 Details : class_handle_hash() generates handle cookies by incrementing
4435 global counter, and can hit 0 occasionaly (this is unlikely, but
4436 not impossible, because initial value of cookie counter is
4437 selected randonly). Value of 0 is used as a sentinel meaning
4438 "unassigned handle" --- avoid it. Also coalesce two critical
4439 sections in this function into one.
4441 Severity : enhancement
4443 Description: allow liblustre clients to delegate truncate locking to OST
4444 Details : To avoid overhead of locking, liblustre client instructs OST to
4445 take extent lock in ost_punch() on client's behalf. New connection
4446 flag is added to handle backward compatibility.
4448 Severity : enhancement
4449 Bugzilla : 4928, 7341, 9758
4450 Description: allow number of OST service threads to be specified
4451 Details : a module parameter allows the number of OST service threads
4452 to be specified via "options ost ost_num_threads={N}" in the
4453 OSS's /etc/modules.conf or /etc/modprobe.conf.
4457 Bugzilla : 6146, 9635, 9895
4458 Description: servers crash with bad pointer in target_handle_connect()
4459 Details : In rare cases when a client is reconnecting it was possible that
4460 the connection request was the last reference for that export.
4461 We would temporarily drop the export reference and get a new
4462 one, but this may have been the last reference and the export
4463 was just destroyed. Get new reference before dropping old one.
4465 Severity : enhancement
4466 Frequency : if client is started with failover MDS
4468 Description: Allow multiple MDS hostnames in the mount command
4469 Details : Try to read the configuration from all specified MDS
4470 hostnames during a client mount in case the "primary"
4473 Severity : enhancement
4475 Description: Stop sending data to evicted clients as soon as possible.
4476 Details : Check if the client we are about to send or are sending data to
4477 was evicted already. (Check is done every second of waiting,
4478 for which l_wait_event interface was extended to allow checking
4479 of exit condition at specified intervals).
4482 Frequency : rare, normally only when NFS exporting is done from client
4484 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4486 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4489 Frequency : always, if srand() is called before liblustre initialization
4491 Description: Liblustre uses system PRNG disturbing its usage by user application
4492 Details : Introduce internal to lustre fast and high-quality PRNG for
4493 lustre usage and make liblustre and some other places in generic
4494 lustre code to use it.
4496 Severity : enhancement
4497 Bugzilla : 9477, 9557, 9870
4498 Description: Verify that the MDS configuration logs are updated when xml is
4499 Details : Check if the .xml configuration logs are newer than the config
4500 logs stored on the MDS and report an error if this is the case.
4501 Request --write-conf, or allow starting with --old_conf.
4503 Severity : enhancement
4505 Description: Handle symlinks in the path when checking if Lustre is mounted.
4506 Details : Resolve intermediate symlinks when checking if a client has
4507 mounted a filesystem to avoid duplicate client mounts.
4512 Description: lconf can hit an error exception but still return success.
4513 Details : The lconf command catches the Command error exception at the top
4514 level script context and will exit with the associated exit
4515 status, but doesn't ensure that this exit status is non-zero.
4520 Description: failure of ptlrpc thread startup can cause oops
4521 Details : Starting a ptlrpc service thread can fail if there are a large
4522 number of threads or the server memory is very fragmented.
4523 Handle this without oopsing.
4526 Frequency : always, only if liblustre and non-default acceptor port was used
4528 Description: liblustre cannot connect to servers with non-default acceptor port
4529 Details : tcpnal_set_default_params() was not called and was therefore
4530 ignoring the environment varaible TCPNAL_PORT, as well as other
4531 TCPNAL_ environment variables
4536 Description: two objects could be created on the same OST for a single file
4537 Details : If an OST is down, in some cases it was possible to create two
4538 objects on a single OST for a single file. No problems other
4539 than potential performance impact and spurious error messages.
4543 Bugzilla : 5681, 9562
4544 Description: Client may oops in ll_unhash_aliases
4545 Details : Client dcache may become inconsistent in race condition.
4546 In some cases "getcwd" can fail if the current directory is
4552 Description: Inode refcounting problems in NFS export code
4553 Details : link_raw functions used to call d_instantiate without obtaining
4554 extra inode reference first.
4558 Bugzilla : 9942, 9903
4559 Description: Referencing freed requests leading to crash, memleaks with NFS.
4560 Details : We used to require that call to ll_revalidate_it was always
4561 followed by ll_lookup_it. Also with revalidate_special() it is
4562 possible to call ll_revalidate_it() twice for the same dentry
4563 even if first occurence returned success. This fix changes semantic
4564 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4565 reference on a request referred from the intent.
4566 ll_intent_release() then releases such a request.
4569 Frequency : rare, normally benchmark loads only
4571 Description: unlinked inodes were kept in memory on the client
4572 Details : If a client is repeatedly creating and unlinking files it
4573 can accumulate a lot of stale inodes in the inode slab cache.
4574 If there is no other client load running this can cause the
4575 client node to run out of memory. Instead flush old inodes
4576 from client cache that have the same inode number as a new inode.
4579 Frequency : SLES9 2.6.5 kernel and long filenames only
4580 Bugzilla : 9969, 10379
4581 Description: utime reports stale NFS file handle
4582 Details : SLES9 uses out-of-dentry names in some cases, which confused
4583 the lustre dentry revalidation. Change it to always use the
4587 Frequency : rare, unless heavy write-truncate concurrency is continuous
4588 Bugzilla : 4180, 6984, 7171, 9963, 9331
4589 Description: OST becomes very slow and/or deadlocked during object unlink
4590 Details : filter_destroy() was holding onto the parent directory lock
4591 while truncating+unlinking objects. For very large objects this
4592 may block other threads for a long time and slow overall OST
4593 responsiveness. It may also be possible to get a lock ordering
4594 deadlock in this case, or run out of journal credits because of
4595 the combined truncate+unlink. Solution is to do object truncate
4596 first in one transaction without parent lock, and then do the
4597 final unlink in a new transaction with the parent lock. This
4598 reduces the lock hold time dramatically.
4601 Frequency : rare, 2.4 kernels only
4603 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4604 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4605 before unmounting their filesystem in order to avoid deadlock.
4606 The kernel_locked() function in 2.4 kernels only checks whether
4607 the kernel lock is held, not whether it is this process that is
4608 holding it as 2.6 kernels do.
4613 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4614 Details : The client ptlrpc code may be trying to reconnect to a down
4615 server before a previous connection attempt has timed out.
4616 Increase the reconnect interval to be longer than the connection
4617 timeout interval to avoid sending duplicate connections to
4621 Frequency : echo_client brw_test command
4623 Description: fix echo_client to work with OST preallocated code
4624 Details : OST preallocation code (5137) didn't take echo_client IO path
4625 into account: echo_client calls filter methods outside of any
4626 OST thread and, hence, there is no per-thread preallocated
4627 pages and buffers to use. Solution: hijack pga pages for IO. As
4628 a byproduct, this avoids unnecessary data copying.
4632 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4633 Description: Client can oops in mdc_commit_close() after open replay
4634 Details : It was possible for the MDS to return an open request with no
4635 transaction number in mds_finish_transno() if the client was
4636 evicted, but without actually returning an error. Clients
4637 would later try to replay that open and may trip an assertion
4638 Simplify the client close codepath, and always return an error
4639 from the MDS in case the open is not successful.
4642 Frequency : rare, 2.6 OSTs only
4644 Description: OST may deadlock under high load on fragmented files
4645 Details : If there was a heavy load and highly-fragmented OST filesystems
4646 it was possible to have all the OST threads deadlock waiting on
4647 allocation of biovecs, because the biovecs were not released
4648 until the entire RPC IO was completed. Instead, release biovecs
4649 as soon as they are complete to ensure forward IO progress.
4651 Severity : enhancement
4653 Description: Support for specifying external journal device at mount
4654 Details : If an OST or MDS device is formatted with an external journal
4655 device, this device major/minor is stored in the ext3 superblock
4656 and may not be valid for failover. Allow detecting and
4657 specifying the external journal at mount time.
4662 Description: Mounting an MDS with pending unlinked files may cause oops
4663 Details : target_finish_recovery() calls mds_postrecov() which returned
4664 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4665 considers this an error and immediately begins cleaning up the
4666 lov, just after starting the mds_lov process
4668 Severity : enhancement
4670 Description: Implement 'lfs df' to report actual free space on per-OST basis
4671 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4672 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4673 usage of objects; '-h' to report in human readable format.
4675 ------------------------------------------------------------------------------
4677 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4684 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4685 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4686 at startup during orphan unlinking. Instead, skip buddy update
4687 before setup, as it will be regenerated after recovery is complete.
4692 Description: If an OST is inactive, its locks might reference stale inodes.
4693 Details : lov_change_cbdata() must iterate over all namespaces, even if
4694 they are inactive to clear inode references from the lock.
4696 Severity : enhancement
4697 Frequency : occasional, if non-standard max_dirty_mb used
4699 Description: Client will block write RPCs if not enough grant
4700 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4701 then the client will block writes while waiting for another RPC
4702 to complete instead of consuming its dirty limit. With change
4703 we get improved performance when max_dirty_mb is small.
4705 Severity : enhancement
4706 Bugzilla : 3389, 6253
4707 Description: Add support for supplementary groups on the MDS.
4708 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4709 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4710 lookups for user supplementary groups into a cache.
4714 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4715 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4716 for liblustre O_EXCL is enforced by MDS.
4720 Description: readdir never returns NULL in liblustre.
4721 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4722 dentry in struct dirent.
4726 Frequency : liblustre only, depends on application IO pattern
4727 Description: liblustre clients evicted if not contacting servers
4728 Details : Don't put liblustre clients into the ping_evictor list, so
4729 they will not be evicted by the pinger ever.
4731 Severity : enhancement
4733 Description: Add ability to evict clients by NID from MDS.
4734 Details : By echoing "nid:$NID" string into
4735 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4736 $NID would be instantly evicted from this MDS and from all active
4737 OSTs connected to it.
4741 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4742 Details : lookup_it_finish() used to query file size from OSTs that was not
4747 Description: service threads change working directory to that of init
4748 Details : Starting lustre service threads may pin the working directory
4749 of the parent thread, making that filesystem busy. Threads
4750 now change to the working directory of init to avoid this.
4754 Frequency : during shutdown only
4755 Description: shutdown with a failed MDS or OST can cause unmount to hang
4756 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4761 Frequency : 2.6 only
4762 Description: chmod/chown may include an extra supplementary group
4763 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4764 supplementary group and if none is specified this is used.
4767 Bugzilla : 5479 (6816)
4769 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4770 Details : If lookup is in progress on a file that is unlinked we might try
4771 to revalidate the inode and fail in revalidate after lookup is
4772 complete and ll_file_open() enqueues the open again but
4773 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4776 Frequency : always, if lconf --abort_recovery used
4778 Description: lconf --abort_recovery fails with 'Operation not supported'
4779 Details : lconf was attempting to abort recovery on the MDT device and not
4782 ------------------------------------------------------------------------------
4784 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4789 Frequency : rare (only unsupported configurations with a node running as an
4791 Bugzilla : 6514, 5137
4792 Description: Mounting a Lustre file system on a node running as an OST could
4794 Details : OSTs now preallocates memory needed to write out data at
4795 startup, instead of when needed, to avoid having to
4796 allocate memory in possibly low memory situations.
4797 Specifically, if the file system is mounted on on OST,
4798 memory pressure could force it to try to write out data,
4799 which it needed to allocate memory to do. Due to the low
4800 memory, it would be unable to do so and the node would
4801 become unresponsive.
4803 Severity : enhancement
4805 Description: Addition of lconf --service command line option
4806 Details : lconf now accepts a '--service <arg>' option, which is
4807 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4809 Severity : enhancement
4811 Description: Failover mode is now the default for OSTs.
4812 Details : By default, OSTs will now run in failover mode. To return to
4813 the old behaviour, add '--failout' to the lmc line for OSTs.
4815 Severity : enhancement
4817 Description: Health checks are now provided for MDS and OSTs
4818 Details : Additional detailed health check information on MSD and OSTs
4819 is now provided through the procfs health_check value.
4822 Frequency : occasional, depends on IO load
4824 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4825 after numerous create/delete cycles
4826 Details : The ext3 inode allocation policy would not allocate new inodes
4827 very well on the OSTs because there are no new directories
4828 being created. Instead we look for groups with free space if
4829 the parent directories are nearly full.
4834 Description: Network or server problems during mount may cause partially
4835 mounted clients instead of returning an error.
4836 Details : The config llog parsing code may overwrite the error return
4837 code during mount error handling, returning success instead
4843 Description: MDS can fail to allocate large reply buffers
4844 Details : After long uptimes the MDS can fail to allocate large reply
4845 buffers (e.g. zconf client mount config records) due to memory
4846 fragmentation or consumption by the buffer cache. Preallocate
4847 some large reply buffers so that these replies can be sent even
4848 under memory pressure.
4852 Frequency : rare (liblustre)
4853 Description: fsx running with liblustre complained that using truncate() to
4854 extend the file doesn't work. This patch corrects that issue.
4855 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4856 ATTR_SIZE and lsm use in llu_setattr_raw.
4860 Frequency : rare, only 2.6 kernels
4861 Description: Unusual file access patterns on the MDS may result in inode
4862 data being lost in very rare circumstances.
4863 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4864 code in the 2.6 kernel caused the inode and/or EA data not to
4865 be read from disk, causing single-file corruption.
4869 Frequency : rare, only 2.6 filesystems using extents
4870 Description: Heavy concurrent write and delete load may cause data corruption.
4871 Details : It was possible under high-load situations to have an extent
4872 metadata block in the block device cache from a just-unlinked
4873 file overwrite a newly-allocated data block. We now unmap any
4874 metadata buffers that alias just-allocated data blocks.
4878 Frequency : filesystems with default stripe_count larger than 77
4879 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4880 Details : lconf specifies an inode size of 4096 bytes when the default
4881 stripe_count is larger than 77. This conflicts with the default
4882 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4883 this case to avoid pinning too much memory for large EAs.
4885 ------------------------------------------------------------------------------
4887 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4892 Frequency : rare (extremely heavy IO load with hundreds of clients)
4894 Description: Client is evicted, gets IO error writing to file
4895 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4896 caused clients to be evicted for AST timeouts. The fixes in
4897 bug 5192 mean we no longer need to have such short AST timeouts
4898 so ldlm_timeout has been increased.
4901 Frequency : occasional during --force or --failover shutdown under load
4902 Bugzilla : 5949, 4834
4903 Description: Server oops/LBUG if stopped with --force or --failover under load
4904 Details : a collection of import/export refcount and cleanup ordering
4905 issues fixed for safer force cleanup
4908 Frequency : only filesystems larger than 120 OSTs
4909 Bugzilla : 5990, 6223
4910 Description: lfs getstripe would oops on a very large filesystem
4911 Details : lov_getconfig used kfree on vmalloc'd memory
4914 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4915 Bugzilla : 6242, 6243
4916 Description: reading from files that had been truncated to a non-zero size
4917 but never opened returned no data
4918 Details : ll_file_read() reads zeros from no-object files to EOF
4923 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4924 Details : There's little harm in aborting MDS/OSS recovery and letting it
4925 try again, so I removed the LASSERT and return an error instead.
4927 Severity : enhancement
4929 Description: New debugging infrastructure for tracking down data corruption
4930 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4931 (b) cover more of the client-side code path, and (c) try to narrow
4932 down where problems occurred
4936 Bugzilla : 3819, 4364, 4397, 6313
4937 Description: Racing close and eviction MDS could cause assertion in mds_close
4938 Details : It was possible to get multiple mfd references during close and
4939 client eviction, leading to one thread referencing a freed mfd.
4941 Severity: : enhancement
4942 Bugzilla : 3262, 6359
4943 Description: Attempts to reconnect to servers are now more aggressive.
4944 Details : This builds on the enhanced upcall-less recovery that was added
4945 in 1.4.2. When trying to reconnect to servers, clients will
4946 now try each server in the failover group every 10 seconds. By
4947 default, clients would previously try one server every 25 seconds.
4952 Description: After recovery, certain operations trigger a failed
4953 assertion on a client.
4954 Details : Failing over an mds, using lconf -d --failover, while a
4955 client was doing a readdir() call would cause the client to
4956 LBUG after recovery completed and the readdir() was resent.
4958 Severity : enhancement
4960 Description: Default groups are now added by lconf
4961 Details : You can now run lconf --group <servicename> without having to
4962 manually add groups with lmc.
4965 Frequency : occasional
4967 Description: Nodes with an elan id of 0 trigger a failed assertion
4970 Frequency : always when accessing e.g. tty/console device nodes
4972 Description: tty and some other devices nodes cannot be used on lustre
4973 Details : file's private_data field is used by device data and lustre
4974 values in there got lost. New field was added to struct file to
4975 store fs-specific private data.
4978 Frequency : when exporting Lustre via NFS
4980 Description: NFSD failed occasionally when looking up a path component
4981 Details : NFSD is looking up ".." which was broken in ext3 directories
4982 that had grown large enough to become hashed.
4985 Frequency : Clusters with multiple interfaces not on the same subnet
4987 Description: Nodes will repeatedly try to reconnect to an interface which it
4988 cannot reach and report an error to the log.
4989 Details : Extra peer list entries will be created by lconf with some peers
4990 unreachable. lconf now validates the peer before adding it.
4993 Frequency : Only if a default stripe is set on the filesystem root.
4995 Description: Setting a default stripe on the filesystem root prevented the
4996 filesystem from being remounted.
4997 Details : The client was sending extra request flags in the root getattr
4998 request and did not allocate a reply buffer for the dir EA.
5001 Frequency : occasional, higher if lots of files are accessed by one client
5002 Bugzilla : 6159, 6097
5003 Description: Client trips assertion regarding lsm mismatch/magic
5004 Details : While revalidating inodes the VFS looks up inodes with ifind()
5005 and in rare cases can find an inode that is being freed.
5006 The ll_test_inode() code will free the lsm during ifind()
5007 when it finds an existing inode and then the VFS later attaches
5008 this free lsm to a new inode.
5012 Bugzilla : 6422, 7030
5013 Description: MDS deadlock between mkdir and client eviction
5014 Details : Creating a new file via mkdir or mknod (starting a transaction
5015 and getting the ns lock) can deadlock with client eviction
5016 (gets ns lock and trying to finish a synchronous transaction).
5019 Frequency : occasional
5020 Description: While starting a server, the fsfilt_ext3 module could not be
5022 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5023 kernels. Previously, lconf would still use the ext3 name
5024 when trying to load modules. Now, it will correctly use
5025 ext3 on 2.4 and ldiskfs on 2.6.
5027 Severity : enhancement
5028 Description: The default stripe count has been changed to 1
5029 Details : The interpretation of the default stripe count (0, to lfs
5030 or lmc) has been changed to mean striping across a single
5031 OST, rather than all available. For general usage we have
5032 found a stripe count of 1 or 2 works best.
5034 Severity : enhancement
5035 Description: Add support for compiling against Cray portals.
5036 Details : Conditional compiling for some areas that are different
5040 Frequency : occasional
5041 Bugzilla : 6409, 6834
5042 Description: Creating files with an explicit stripe count may lead to
5043 a failed assertion on the MDS
5044 Details : If some OSTs are full or unavailable, creating files may
5045 trigger a failed assertion on the MDS. Now, Lustre will
5046 try to use other servers or return an error to the
5050 Frequency : occasional
5052 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5053 caused lock timeout during readahead (since 1.4.2).
5054 Details : Processes doing readahead might match a lock that hasn't been
5055 granted yet if there are overlapping and conflicting lock
5056 requests. The readahead process waits on ungranted lock
5057 (original lock is CBPENDING), while OST waits for that process
5058 to cancel CBPENDING read lock and eventually evicts client.
5060 Severity : enhancement
5062 Description: Initial enabling of flock support for clients
5063 Details : Implements fcntl advisory locking and file status functions.
5064 This feature is provided as an optional mount flag (default
5065 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5066 locking are implemented yet, and those that are are not guaranteed
5067 to be completely correct in production environments.
5068 mount -t lustre -o [flock|noflock] ...
5071 Frequency : occasional
5073 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5074 assertion in the ext3 JBD (journaling) layer.
5075 Details : The b_committed_data struct is protected by the big kernel lock
5076 in 2.4 kernels, serializing journal_commit_transaction() and
5077 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5078 access to this struct. In 2.6 kernels there is finer grained
5079 locking to improve SMP performance of the JBD layer.
5083 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5088 Description: merge and cleanup kernel patches.
5091 -----------------------------------------------------------------------------
5093 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5095 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5096 versions of Lustre. Please update all servers and clients to
5097 version 1.4.2 or later at the same time. You must also run
5098 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5099 to update the configuration logs.
5101 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5102 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5103 - swab llog records of type '0' so we get proper header size/idx (5861)
5104 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5105 - fix rename of one directory over another leaking an inode (5953)
5106 - avoid SetPageDirty on 2.6 (5981)
5107 - don't re-add just-being-destroyed locks to the waiting list (5653)
5108 - when creating new directories, inherit the parent's custom
5109 striping settings if present parent (3048)
5110 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5111 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5112 - don't hold client locks on temporary worklist from l_lru (5666)
5113 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5114 - automatically evict dead clients (5921)
5115 - Update file size properly in create+truncate+fstat case (6196)
5116 - Do not unhash mountpoint dentries, do not allow removal of
5118 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5119 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5120 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5121 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5122 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5123 - Correctly handle reads of files with no objects (6243)
5124 - lctl recover will also mark a device active if deactivate used (5933)
5126 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5127 - allow --write-conf on an MDS with different nettype than client (5619)
5128 - don't write config llogs to MDS for mounts not from that MDS (5617)
5129 - lconf should create multiple TCP connections from a client (5201)
5130 - init scripts are now turned off by default; run chkconfig --on
5131 lustre and chkconfig --on lustrefs to use them
5132 - upcalls are no longer needed for clients to recover to failover
5134 - add --abort-recovery option to lconf to abort recovery on device
5136 - add support for an arbitrary number of OSTs (3026)
5137 - Quota support protocol changes.
5138 - forward compatibility changes to wire structs (6007)
5139 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5140 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5141 - improved "lustre status" script
5142 - initialize blocksize for non-regular files (6062)
5143 - added --disable-server and --disable-client configure options (5782)
5144 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5145 - Vanilla 2.4.29 support
5146 - increase maximum number of obd devices to 520 (6242)
5147 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5148 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5149 SLES 9 2.6.5 kernels
5150 - limit stripes per file to 160 (the maximum EA size) (6093)
5152 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5155 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5156 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5157 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5158 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5159 - small ext3 extents cleanups and fixes (5733)
5160 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5161 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5162 - avoid needless client->OST connect, fix handle mismatch (5317)
5163 - fix DLM error path that led to out-of-sync client, long delays (5779)
5164 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5165 - fix several locking issues related to i_size (5492,5624,5654,5672)
5166 - don't move pending lock onto export if it is already evicted (5683)
5167 - fix kernel oops when creating .foo in unlinked directory (5548)
5168 - fix deadlock in obdfilter statistics vs. object create (5811)
5169 - use time_{before,after} to avoid timer jiffies wrap (5882)
5170 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5171 - Do not leak request if server was not able to process it (5154)
5172 - If mds_open unable to find parent dir, make that negative lookup(5154)
5173 - don't create new directories with extent-mapping (5909, 5936)
5175 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5176 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5177 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5178 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5179 - support for RHEL4 kernel on i686 (5773)
5180 - provide error messages when incompatible logs are encountered (5898)
5182 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5183 * version 1.4.0.10 (1.4.1 release candidate 1)
5185 - don't keep a lock reference when lock is not granted (4238)
5186 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5187 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5188 - drop import inflight refcount on signal_completed_replay error (5255)
5189 - unlock page after async write error during send (3677)
5190 - handle missing objects in filter_preprw_read properly (5265)
5191 - no transno return for symlink open, don't save no-trasno open (3440)
5192 - don't try to complete elan receive that already failed (4012)
5193 - free RPC server reply state on error (5406)
5194 - clean up thread from ptlrpc_start_thread() on error (5160)
5195 - readahead could read extra page into cache that wasn't ejected (5388)
5196 - prevent races in class_attach/setup/cleanup/detach (5260)
5197 - don't dereference de->d_inode after l_dput of de (5458)
5198 - use "int" for stripe value returned from lock_to_stripe (5544)
5199 - mballoc allocation and error-checking fixes in 2.6 (5504)
5200 - block device patches to fix I/O request sizes in 2.6 (5482)
5201 - look up hostnames for IB nals (5602)
5202 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5203 - don't start multiple acceptors for the same port (5277)
5204 - fix incorrect LASSERT in mds_getattr_name (5635)
5205 - export a proc file for general "ping" checking (5628)
5206 - fix "lfs check" to not block when the MDS is down (5628)
5208 - service request history (4965)
5209 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5210 - create an "evict_client" /proc entry on OSTs, like the MDS has
5211 - fix mount usage message, return errors per mount(8) (5168)
5212 - change grep [] to grep "[]" in tests so they work in more UMLs
5213 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5214 - remove ancient LOV_MAGIC_V0 EA support (5047)
5215 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5216 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5217 - allow readahead allocations to fail when low on memory (5383)
5218 - mmap locking landed again, after considerable improvement (2828)
5219 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5221 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5224 - send OST transaction number in read/write reply to free req (4966)
5225 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5226 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5227 - the watchdog thread now runs as interruptible (5246)
5228 - flock/lockf fixes (but it's still disabled, pending 5135)
5229 - don't use EXT3 constants in llite code (5094)
5230 - memory shortage at startup could cause assertion (5176)
5232 - reorganization of lov code
5233 - single portals codebase
5235 - add extents/mballoc support (5025)
5236 - direct I/O reads in the obdfilter (4048)
5237 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5239 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5242 - send OST transaction number in read/write reply to free req (4966)
5243 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5244 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5245 - the watchdog thread now runs as interruptible (5246)
5246 - handle missing objects in filter_preprw_read properly (5265)
5247 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5248 - drop import inflight refcount on signal_completed_replay error (5255)
5249 - unlock page after async write error during send (3677)
5250 - return original error code on reconstructed replies (3761)
5251 - no transno return for symlink open, don't save no-trasno open (3440)
5253 - add pid to ldlm debugging output (4922)
5254 - bump the watchdog timeouts -- we can't handle 30sec yet
5255 - extra debugging for orphan dentry/inode bug (5259)
5257 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5260 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5261 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5262 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5263 - fix NULL dereference in /proc/sys/portals/routes (4827)
5264 - allow failed mdc_close() operations to be interrupted (4561)
5265 - stop precreate on OST before MDS would time out on it (4778)
5266 - don't send partial-page writes before EOF from client (4410)
5267 - discard client grant for sub-page writes on large-page clients (4520)
5268 - don't free dentries not owned by NFS code, check generation (4806)
5269 - fix lsm leak if mds_create_objects() fails (4801)
5270 - limit debug_daemon file size, always print CERROR messages (4789)
5271 - use transno after validating reply (3892)
5272 - process timed out requests if import state changes (3754)
5273 - update mtime on OST during writes, return in glimpse (4829)
5274 - add mkfsoptions to LDAP (4679)
5275 - use ->max_readahead method instead of zapping global ra (5039)
5276 - don't interrupt __l_wait_event() during strace
5278 - add software watchdogs to catch hung threads quickly (4941)
5279 - make lustrefs init script start after nfs is mounted
5280 - fix CWARN/ERROR duplication (4930)
5281 - return async write errors to application if possible (2248)
5282 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5283 - print NAL number in %x format (4645)
5284 - update barely-supported suse-2.4.21-171 series (4842)
5285 - support for sles 9 %post scripts
5286 - support for building 2.6 kernel-source packages
5287 - support for sles km_* packages
5289 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5292 - ignore -ENOENT errors in osc_destroy (3639)
5293 - notify osc create thread that OSC is being cleaned up (4600)
5294 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5295 - reconstruct ost_handle() like mds_handle() (4657)
5296 - create a new thread to do import eviction to avoid deadlock (3969)
5297 - let lconf resolve symlinked-to devices (4629)
5298 - don't unlink "objects" from directory with default EA (4554)
5299 - hold socknal file ref over connect in case target is down (4394)
5300 - allow more than 32000 subdirectories in a single directory (3244)
5301 - fix blocks count for O_DIRECT writes (3751)
5302 - OST returns ENOSPC from object create when no space left (4539)
5303 - don't send truncate RPC if file size isn't changing (4410)
5304 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5305 - bind to privileged port in socknal and tcpnal (3689)
5307 - rate limit CERROR/CWARN console message to avoid overload (4519)
5308 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5309 - basic mmap support (3918)
5310 - kernel patch series update from b1_4 (4711)
5312 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5315 - avoid crash during MDS cleanup with OST shut down (2775)
5316 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5317 - don't use bad inodes on the MDS (3744)
5318 - dynamic object preallocation to improve recovery speed (4236)
5319 - don't hold spinlock over lock dumping or change debug flags (4401)
5320 - don't zero obd_dev when it is force cleaned (3651)
5321 - print grants to console if they go negative (4431)
5322 - "lctl deactivate" will stop automatic recovery attempts (3406)
5323 - look for existing locks in ldlm_handle_enqueue() (3764)
5324 - don't resolve lock handle twice in recovery avoiding race (4401)
5325 - revalidate should check working dir is a directory (4134)
5327 - don't always mark "slow" obdfilter messages as errors (4418)
5329 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5332 - don't close LustreDB during write_conf until it is done (3860)
5333 - fix typo in lconf for_each_profile (3821)
5334 - allow dumping logs from multiple threads at one time (3820)
5335 - don't allow multiple threads in OSC recovery (3812)
5336 - fix debug_size parameters (3864)
5337 - fix mds_postrecov to initialize import for llog ctxt (3121)
5338 - replace config semaphore with spinlock (3306)
5339 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5340 - don't allow enqueue to complete on a destroyed export (3822)
5341 - down write_lock before checking llog header bitmap (3825)
5342 - recover from lock replay timeout (3764)
5343 - up llog sem before sending rpc (3652)
5344 - reduce ns lock hold times when setting kms (3267)
5345 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5346 - fix NULL deref and obd_dev leak on setup error (3312)
5347 - replace some LBUG about llog ops with error handling (3841)
5348 - don't match INVALID dentries from d_lookup and spin (3784)
5349 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5350 - fix invalid assertion in ptlrpc_set_wait (3880)
5352 - add libwrap support for the TCP acceptor (3996)
5353 - add /proc/sys/portals/routes for non-root route listing (3994)
5354 - allow setting MDS UUID in .xml (2580)
5355 - print the stack of a process that LBUGs (4228)
5357 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5360 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5361 - make sure to unset replay flag from failed open requests (3440)
5362 - if default stripe count is 0, use OST count for inode size (3636)
5363 - update parent mtime/ctime on client for create/unlink (2611)
5364 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5365 - free recovery state on server during a forced cleanup (3571)
5366 - unregister_reply for resent reqs (3063)
5367 - loop back devices mounting and status check on 2.6 (3563)
5368 - fix resource-creation race that can provoke i_size == 0 (3513)
5369 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5370 - more debugging for page-accounting assertion (3746)
5371 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5372 - avoid deadlock after precreation failure (3758)
5373 - fix race and lock order deadlock in orphan handling (3450, 3750)
5374 - add validity checks when grabbing inodes from l_ast_data (3599)
5376 - add /proc/.../recovery_status to obdfilter (3428)
5377 - lightweight CDEBUG infrastructure, debug daemon (3668)
5378 - change default OSC RPC parameters to be better on small clusters
5379 - turn off OST read cache for files smaller than 32MB
5380 - install man pages and include them in rpms (3100)
5381 - add new init script for (un)mounting lustre filesystems (2593)
5382 - run chkconfig in %post for init scripts (3701)
5383 - drop scimac NAL (unmaintained)
5385 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5388 - clean kiobufs before and after use (3485)
5389 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5390 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5391 - change init script to fail more clearly if not run as root (1528)
5392 - allow clients to reconnect during replay (1742)
5393 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5394 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5395 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5396 - fix cond_resched (3554)
5397 - only evict unfinished clients after recovery (3515)
5398 - allow bulk resend, prevent data loss (3570)
5399 - dynamic ptlrpc request buffer allocation (2102)
5400 - don't allow unlinking open directory if it isn't empty (2904)
5401 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5402 - remove extraneous obd dereference causing LASSERT failure (3334)
5403 - don't use get_cycles() when creating temp. files on the mds (3156)
5404 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5405 - handle EEXIST for set-stripe, set proper directory name (3336)
5407 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5408 - fix ksocknal_fmb_callback() error messages (2918)
5410 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5413 - don't copy lvb into (possibly NULL) reply on error (2983)
5414 - don't deref dentry after dput, don't free lvb on error (2922)
5415 - use the kms to determine writeback rpc length (2947)
5416 - increment oti_logcookies when osc is inactive (2948)
5417 - update client's i_blocks count via lvb messages (2543)
5418 - handle intent open/close of special files properly (1557)
5419 - mount MDS with errors=remount-ro, like obdfilter (2009)
5420 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5421 - don't use cancelling-locks' kms values (2947)
5422 - use highest lock extent for kms, not last one (2925)
5423 - don't dereference ERR_PTR() dentry in error handling path (3107)
5424 - fix thread race in portals_debug_dumplog() (3122)
5425 - create lprocfs device entries at setup instead of at attach (1519)
5426 - common AST error handler, don't evict client on completion race (3145)
5427 - zero nameidata in detach_mnt in 2.6 (3118)
5428 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5429 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5430 - initialize RPC timeout timer earlier for 2.6 (3219)
5431 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5432 - print nal/nid for unknown nid (3258)
5433 - additional checks for oscc recovery before doing precreate (3284)
5434 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5435 - don't crash in mdc_close for bad permissions on open (3285)
5436 - zero i_rdev for non-device files (3147)
5437 - clear page->private before handing to FS, better assertion (3119)
5438 - tune the read pipeline (3236)
5439 - fix incorrect decref of invalidated dentry (2350)
5440 - provide read-ahead stats and refine rpc in flight stats (3328)
5441 - don't hold journal transaction open across create RPC (3313)
5442 - update atime on MDS at close time (3265)
5443 - close LDAP connection when recovering to avoid server load (3315)
5444 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5445 - don't leak open file on MDS after open resend (3325)
5446 - serialize filter_precreate and filter_destroy_precreated (3329)
5447 - loop device shouldn't call sync_dev() for nul device (3092)
5448 - clear page cache after eviction (2766)
5449 - resynchronize MDS->OST in background (2824)
5450 - refuse to mount the same filesystem twice on same mountpoint (3394)
5451 - allow llmount to create routes for mounting behind routers (3320)
5452 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5453 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5454 - fix rare problem with rename on htree directories (3417)
5456 - allow default OST striping configuration per directory (1414)
5457 - fix compilation for qswnal for 2.6 kernels (3125)
5458 - increase maximum number of MDS request buffers for large systems
5459 - change liblustreapi to be useful for external progs like lfsck (3098)
5460 - increase local configuration timeout for slow disks (3353)
5461 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5463 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5466 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5467 - don't overwrite extent policy data in reply if lock was blocked (2901)
5468 - drop filter export grants atomically with removal from device (2663)
5469 - del obd_self_export from work_list in class_disconnect_exports (2908)
5470 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5471 - swab reply message in mdc_close, other PPC fixes (2464)
5472 - fix destroying of named logs (2325)
5473 - overwrite old logs when running lconf --write_conf (2264)
5474 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5475 - fix race in target_handle_connect (2898)
5476 - mds_reint_create() should take same inode create lock (2926)
5477 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5478 - don't close files for self_export to avoid uninitialized obd (2936)
5479 - allow MDS with the same name as client node (2939)
5480 - hold dentry reference for closed log files for unlink (2325)
5481 - reserve space for all logs during transactions (2059)
5482 - don't evict page beyond end of stripe extent (2925)
5483 - don't oops on a deleted current working directory (2399)
5484 - handle hard links to targets without a parent properly (2517)
5485 - don't dereference NULL lock when racing during eviction (2867)
5486 - don't grow lock extents when lots of conflicting locks (2919)
5488 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5491 - account for cache space usage on clients to avoid data loss (974)
5492 - lfsck support in lustre kernel code (2349)
5493 - reduce journal credits needed for BRW writes (2370)
5494 - orphan handling to avoid losing space on client/server crashes
5495 - ptlrpcd can be blocked, stopping ALL progress (2477)
5496 - use lock value blocks to assist in proper KMS, faster stat (1021)
5497 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5498 - recovery for initial connections (2355)
5499 - fixes for mds_cleanup_orphans (1934)
5500 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5501 - block all file creations until orphan recovery completes (1901)
5502 - client remove rq_connection from request struct (2423)
5503 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5504 - recovery timer race (2670)
5505 - mdc_close recovey bug (2532)
5506 - ptlrpc cleanup bug (2710)
5507 - mds timeout on local locks (2588)
5508 - namespace lock held during RPCs (2431)
5509 - handle interrupted sync write properly (2503)
5510 - don't try to handle a message that hasn't been replied to (2699)
5511 - client assert failure during cleanup after abort recovery (2701)
5512 - leak mdc device after failed mount (2712)
5513 - ptlrpc_check_set allows timedout requests to complete (2714)
5514 - wait for inflight reqs when ptlrpcd finishes (2710)
5515 - make sure unregistered services are removed from the srv_list
5516 - reset bulk XID's when resending them (caught by 1138 test)
5517 - unregister_bulk after timeout
5518 - fix lconf error (2694)
5519 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5520 - readahead locks pages, leaves pending causing memory pressure (2673)
5521 - increase OST request buffers to 4096 on large machines (2729)
5522 - fix up permission of existing directories in simple_mkdir (2661)
5523 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5524 - don't assign transno to errored transactions (2742)
5525 - don't delete objects on OST if given a bogus objid from MDS (2751)
5526 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5527 - if rq_no_resend, then timeout request after recovery (2432)
5528 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5529 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5530 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5531 - MDS recovery shouldn't delete objects that it has given out (2730)
5532 - if enqueue arrives after completion, don't clobber LVB (2819)
5533 - don't unlock pages twice when trigger_group_io returns error (2814)
5534 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5535 - don't write pages to disk if there was an error (1450)
5536 - don't ping imports that have recovery disabled (2676)
5537 - take buffered bytes into account when balancing socknal conn (2817)
5538 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5539 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5540 - remove little-endian swabbing of llog records (1987)
5541 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5542 - retry reposting request buffers if they fail (1191)
5543 - grow extent at grant time to avoid granting a revoked lock (2809)
5544 - lock revoke doesn't evict page if covered by a second lock (2765)
5545 - disable VM readahead to avoid reading outside lock extents (2805)
5547 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5548 - updated kernel patches for hp-2.4.20 kernel (2681)
5550 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5553 - fix truncated write corruption (2366)
5554 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5556 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5557 - ensure len(uuid) < 37 in lmc (1171)
5558 - fix ia64 OOPS in llog_test (2255)
5559 - zero end of page at obdfilter for partial page writes (2648)
5560 - don't leave stale dentries around after renames (bug 2428)
5561 - fix timeouts when evicting a client with a single lock held (2642)
5562 - set deadline for the initial HELLO message to drain (2634)
5563 - print out dotted-quad IP addresses in the socknal (2302)
5565 - additional debugging for MDS client eviction problem (2443)
5566 - fix mkfsoptions support for osts (2603, 2604)
5568 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5571 - add series for the vanilla 2.6.0 kernel
5572 - add series for the vanilla 2.4.24 kernel
5573 - add series for a cray x86/64 UL kernel drop
5574 - fix xattr patches for the vanilla 2.4.19 series
5576 - generate true UUIDs in lmc (1171)
5577 - have portals stack dumping break in UML (2466)
5578 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5579 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5580 - wake callback waiting threads on client eviction (2460)
5581 - Add --ptldebug and --subsystem to lmc (1719)
5582 - update assertion to allow safe interrupt allocation
5583 - set rq_no_resend for cancel requests (2432)
5584 - recalculate ptlrpcd timeout after resend (2494)
5585 - call vfs_rmdir when removing pending directories (2368)
5586 - fix renaming a file to itself (2429)
5587 - lmc creates a default one-stripe lov (2454)
5588 - expand procfs space to handle large clusters (2326)
5589 - increase UML stack to avoid overflow
5590 - update lconf's list of debug and subsystem masks
5591 - fix lfs find --obd (2510)
5592 - /proc tunable for disabling filter read caching (2591)
5593 - stop rpm packages from altering slapd.conf (2301)
5594 - disable nagle in the socknal under 0conf (2578)
5595 - choose mds inode size based on stripe count (2572)
5596 - fix kernel-source rpm problems (2516)
5598 - add --disable-doc to avoid pdf generation (2421)
5599 - update documentation, tests, type-os, comments
5600 - avoid format warnings on ia64
5601 - remove the TOE NAL
5602 - tiny code cleanups by removing unused fields
5604 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5607 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5608 - remove the most blatant lies from BUILDING file (2371)
5609 - change default debug level to reasonable production setting
5610 - reduce client side cache size to reduce cache flush time
5611 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5612 - make TCP zerocopy and pinger support enabled by default (2476)
5613 - sync writes completed after process exits caused crashes (2319)
5614 - maintain correct mount count on the MDS (2356)
5615 - backout 1557, because 2316 wasn't really fixed
5616 - better file I/O statistics gathering in /proc
5617 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5618 - another kernel patch to fix zero-copy TCP function export
5619 - don't take duplicate lock when processing re-sent getattr (2420)
5620 - lctl uses obd_self_export instead of creating new conn (2353)
5621 - MDS/OST recovery case which requires object creation asserted (2425)
5622 - move lfs from /usr/sbin to /usr/bin in packages
5623 - fix race between mds_client_add and mds_client_free (2417)
5624 - use kmalloc instead of slabs in portals (2430)
5625 - don't create duplicate records when a failover MDS is present (2442)
5626 - remove unnecessary mount age check (2332)
5627 - don't remove directory inodes from locks prematurely (2451)
5628 - don't break if MDS service name is the same as hostname (2103)
5629 - fix races in client write RPC generation when cache full (2482)
5631 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5634 - remove now-unused request->rq_obd (278)
5635 - if an allocation fails, print out how much memory we've used (1933)
5636 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5637 - add the "configurable stack size" patch to most series files (1256)
5638 - ability to write large log records, for 100+ OST configs (2306)
5639 - fix NULL deref when filter_prep fails (2314)
5640 - fix operator precedence error in filter_sync
5641 - dynamic allocation of socknal TX descriptors (2315)
5642 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5643 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5644 - ensure that CWARN messages in Portals always get to the syslog
5645 - __init/__exit are not for prototype decls (ldlm_init/exit)
5646 - x86-64 compile warning fixes
5647 - fix gateway LMC keyword conflict (2318)
5648 - fix MDS lock inversions in getattr/reint paths (1844)
5649 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5650 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5651 - DLM race condition prevented some lock evictions (2328)
5652 - ENOMEM detection and retry on socknal sends (2230)
5653 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5654 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5655 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5656 - better error messages when a client is rejected during recovery (1505)
5657 - avoid cancelling locks which were never granted, after failure (2330)
5658 - fix i_sem/journal inversion in mds_client_add (2333)
5659 - fix truncate/getattr lock cycle deadlock (2334)
5660 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5661 - fix two rare exit paths which could leak an l_lock() ref (2321)
5662 - fencepost error in MDS/OST orphan recovery (2226)
5663 - make log record alignment 8 bytes (1988)
5664 - lstripe now fails when requested offset > ost_count (2237)
5665 - ensure that all kernel series have a complete list.h (1607)
5666 - fix crashes in special-file operations (2316)
5667 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5669 - allow configurable automake binary, for testing new versions
5670 - small update to the lfs documentation
5672 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5674 * fix negative export reference count in fsfilt_sync (2312)
5676 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5677 * release candidate 0.9.1
5679 - orphans are moved into the PENDING directory for possible recovery
5680 - replayed opens now open by fid for orphan/rename safety (1042)
5681 - last close of an orphan inode generates a transno (683)
5682 - chdir() and mount() now pin the directory entry (1020)
5683 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5684 - discard very old requests without processing them (1502)
5685 - remove some common, well-understood CERRORs (1505)
5686 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5687 - clear "grant" flags in OST replies until OST grant code lands (1644)
5688 - fix read performance by not clobbering i_blksize on client (1598)
5689 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5690 - make LRU size a /proc tunable, clears locks when reduced (707)
5691 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5692 - server locks take references on exports now (1558)
5693 - build fixes for 2.4.20-rh trees (1663)
5694 - return an error from lov_create if all OSCs are inactive (1751)
5695 - fix import levels when a reconnect happens without a timeout (1597)
5696 - exit early from mds_open if we get a lookup error (1749)
5697 - partial page read at EOF wouldn't wait for disk before sending (1642)
5698 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5699 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5700 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5701 - write/truncate lock inversion (1639)
5702 - Don't auto-load obdclass, portals modules during cleanup (1495)
5703 - fix timestamps from jumping to "now" (1763)
5704 - extra journal assertions (1648)
5705 - add an extra multiunlink test (1771)
5706 - fix read_record/write_record API (1776)
5707 - fix leak of offset_extent, possible incorrect i_size later (1772)
5708 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5709 - Debugging for the kqswnal_get_idle_tx problems (1820)
5710 - Allow recovery to be attempted multiple times (1536)
5711 - Write out MDS last_rcvd file after it is first created (1600)
5712 - Fix tx_descriptor leak in failed transmit situations (1827)
5713 - ext3 journaling fixes for assertion failure after IO error (1871)
5714 - class_export_put() on freed export after completion AST error (1896)
5715 - Fix revalidate looping in VFS (1322)
5716 - Don't access a freed export during MDS_REINT timeout (1521)
5717 - Add open-unlink recovery support on the MDS (1673,1764)
5718 - Return an error if no MDS data was read from last_rcvd (1946)
5719 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5720 - Don't setup a disk device that doesn't match exported UUID (317)
5721 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5722 - avoid committing NULL handle in force close
5723 - local.sh is now a one-stripe LOV configuration
5724 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5725 - fix ext3 htree duplicate directory entry corruption (1516)
5726 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5727 - update to new LOV EA format (2097)
5728 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5730 2003-06-15 Phil Schwan <phil@clusterfs.com>
5733 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5734 - per-import/export recovery handling (958, 931, 959)
5735 - multiple last-rcvd slots, for serving multiple FSes (949)
5736 - connections are again shared between multiple imp/exports (963, 964)
5737 - "umount -f" would hang if any requests needed to be sent (393, 978)
5738 - avoid pinning large req buffer by copying for queued messages (989)
5739 - add "uuid" to "lctl device" command to help upcalls (991)
5740 - "open" RPCs with transnos would confuse recovery counters (1037)
5741 - do proper endian conversion of all wire messages (288, 340, 891)
5742 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5743 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5744 - fix ptlrpc_connection leak in target_handle_connect (1174)
5745 - fix import refcounting bug in OST and MDS cleanup (1134)
5746 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5747 - fix very unlikely obd_types race condition (501)
5748 - remove osc_open hack for echo_client (1187)
5749 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5750 - a failure in read_inode2 leads to deadlock (1139)
5751 - cancel ack-locks as soon as transaction is committed (1072)
5752 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5753 - make sure to commitrw after any preprw to avoid deadlock (1162)
5754 - failing to execute a file in a lustre FS would lock inode (1203)
5755 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5756 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5757 - obd_uuid2tgt would walk off the end of the list (1255)
5758 - on IA64 the osc would give portals incorrect bulk size (1258)
5759 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5760 - fix lock inversion caused by new llite matching code (1282)
5761 - limit the number of dirty pages on a client to 10MB (1286)
5762 - timed out locks were not being corrected cancelled (1289)
5763 - fix O_DIRECT above 4GB on IA-32 (1292)
5764 * major user-visible changes
5765 - fail out/fail over policy now controlled by the upcall (993)
5767 - add OBD_PING to check server availability and failure (954)
5768 - lustre messages are now sent in sending host order (288, 340, 891)
5769 - add eadatalen to MDS getattr reply (340)
5770 - OST read replies may contain second buffer, with per-page status (593)
5772 2003-03-11 Phil Schwan <phil@clusterfs.com>
5775 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5776 - failed open()s could cause deadlock; fixed (867, 869)
5777 - stop cancelling OST locks when files are closed (481)
5778 - overlapping XID spaces caused network corruption (851, 853)
5779 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5780 - setattr_raw added, to do single-RPC, server-side setattrs
5781 - lmc/lconf syntax change for OST UUIDs
5782 - fix crashy race condition between ptlrpc_free_req and osc_close
5783 - don't use request in mdc_enqueue if we hit a timeout (889)
5784 - don't set the inode i_size for regular files from the MDS (896)
5785 - handle out of order completion AST (842)
5786 - don't LBUG if a lock request times out after receiving AST (913)
5787 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5788 - if a bad lock AST arrives, send an error instead of dropping entirely
5789 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5790 - fix leak in bulk IO when only partially completed (899, 900, 926)
5791 - fix O_DIRECT for ia64 (55)
5792 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5793 - C-z after timeout could hang a process forever; fixed (977)
5795 - client-side I/O cache (678, 924, 929, 941, 970)
5797 - READPAGE and SETATTRs which don't take server-side locks get
5800 2003-02-11 Phil Schwan <phil@clusterfs.com>
5803 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5804 - Fix fsfilt_extN_readpage() to read a full page of directory
5805 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5806 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5807 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5808 - Most intent operations were removed, in favour of a new RPC mode
5809 that does a single RPC to the server and bypasses most of the VFS
5810 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5811 - Aggressively cancel local locks on DLM servers
5812 - mds_reint_unlink sends EA to the client if it's the last nlink.
5813 client uses that EA to unlink OST objects.
5814 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5815 - recursive symlinks were fixed (439)
5816 - fixed NULL deref in DEBUG_REQ
5817 - filter_update_lastobjid no longer calls sync, which annoyed extN
5818 - fixed multi-client small-writes to a single file problem (445)
5819 - fixed mtime updates during file writes (607)
5820 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5821 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5822 - fixed Linux OST size reporting problem (444, 656)
5823 - OST now updates object mtime with writes or setattr (607, 619)
5824 - client verifies file size before zeroing page past EOF (445)
5825 - OST now writes last allocated objid to disk with allocation (108)
5826 - LOV on echo now works (409)
5828 - mds_reint_unlink sends a new buffer, with the EA included. this
5829 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5830 set if a regular file was being unlinked, and it was the last link
5831 - use PtlGet from the target for bulk writes (315)
5832 - OST now updates object mtime with writes or setattr (607, 619)
5833 - LDLM now has a grant-time callback to revalidate locked items, if
5835 - Many MDS operations were reorganized to combat race conditions
5837 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5838 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5840 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5843 - Fully reactivate OST imports after reconnection (512, others)
5844 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5845 - More graceful error handling for truncating on dead OST (515)
5846 - Don't error out unless we're actually accessing dead stripes (474)
5847 - Fix garbage sizes when stripes are missing (410)
5848 - LRU counters were broken, causing constant lock purge (433, 432)
5849 - garbage on read from stripes with failed OSTs (441)
5850 - mark OSCs as active before reconnecting during recovery (438)
5851 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5852 - lfind did not preserve OST order in output (443)
5853 - symlinks cause hung clients, incorrect data (439)
5854 - stop dereferencing request after dropping refcount (457)
5855 - don't LASSERT(spin_is_locked) on non-SMP (455)
5856 - fixes for many rename() bugs
5857 - fstat didn't correctly synchronize attributes (399)
5858 - server must handle lock cancellation during blocking AST prep (487)
5859 - bulk descriptors were free()d too soon (511)
5860 - fix paths in lconf, which would load incorrect modules (451, 507)
5861 - fix confusing lconf 'host not found' error message (386)
5862 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5863 - fix race condition in mdc_blocking_ast() for inode access (526)
5864 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5865 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5866 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5867 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5868 - fix lstripe in several ways (kernel side) (527)
5869 - fix request leak in ldlm_cli_enqueue (262)
5870 - incorrect OSC was marked inactive after OST failure
5871 - call mds_fs_cleanup before unmounting filesystem (524)
5872 - fix races between taking ns_lock and ldlm_lock_change_resource
5873 - fix races updating LOV export open file list
5874 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5875 - fix recovery NULL deref in ldlm_cli_cancel_unused
5876 - fix some DLM races by using new hash table for lock handles (419)
5877 - permit the client to specify desired inodes, at replay
5878 - duplicate requests when we queue them for replay reintegration
5879 - fix last_rcvd offset calculation
5880 - sync after each recovered transaction, so we always make progress
5881 - never, not always, ERESTART requests without transnos
5882 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5884 - skip replay if the MDS says that the client is already connected
5885 - don't check for a recovery-enabled export to match lctl's UUID
5886 - don't INC_USE_COUNT for phantom exports
5887 - don't crash when cleaning up phantom exports (567)
5888 - don't double-finish or set replay data for errored mdc_open requests
5889 - abort requests when they time out, so we don't get old replies
5890 - send/receive replies for AST messages again
5891 - if the client says that it doesn't have the lock, cancel it on the
5893 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5894 mark it as destroyed, it will all work out when decref is called
5895 - fix module use counts (22, 581)
5897 - ASTs now expect a reply (server cancels lock on error reply)
5899 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5902 - fix many simultaneous client startup (392)
5903 - fix dentry->d_it clobbering
5904 - credentials weren't being shipped for readdir/getattr operations
5905 - remove invalid assertions triggered during some concurrent MD
5907 - proper Lustre versions added (336, 389)
5908 - fix memory leak for create error case (398)
5909 - fix LOV locking bug that would get cli/srv out of sync
5910 - fix echo client over LOV (409)
5911 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5912 - fix double-O_EXCL intent crash (424)
5913 - avoid sending multiple lock CANCELs (352)
5915 - MDS can do multi-client recovery (modulo bugs in new code)
5917 - many updates, edits, cleanups
5919 2002-11-18 Phil Schwan <phil@clusterfs.com>
5922 - fix null d_it dereference (346)
5923 - fix full OST/dbench hang (333)
5924 - fix permission problem with file removal (286)
5925 - fix removal of OSCs from LOV when they fail
5926 - fix NULL deref during bulk timeout (214)
5927 - fix problems related to multiple filesystems on one MDS (241)
5928 - fixed serious subtle metadata locking bugs
5929 - free locks on clients when inodes are removed due to memory
5931 - fix inode pointer in lock data (285)
5932 - partial support for multiple MDS on a single host (241)
5933 - data locks weren't cancelled at clear_inode time (290, 311)
5934 - intent locks could lead to unbounded lock growth (205)
5935 - added a maximum lock count, an LRU list, and a flusher
5936 - fix multiple rename (365)
5937 - properly abstracted the echo client
5938 - OSC locked 1 byte too many; fixed
5939 - rewrote brw callback code:
5940 - fixed recovery bugs related to LOVs (306)
5941 - fixed too-many-pages-in-one-write crash (191)
5942 - fixed (again) crash in sync_io_timeout (214)
5943 - probably fixed callback-related race (385)
5945 - Add capability to MDS protocol
5946 - LDLM cancellations and callbacks on different portals
5948 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5951 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5952 - defer open object destruction to close time (601981, 138)
5953 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5954 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5955 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5956 - don't call obd_setattr in ll_file_release for destroyed objects
5957 * protocol change to lustre_msg: move |version| and add |flags|
5958 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5959 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5960 * added replay of create, unlink, link and rename operations during
5961 MDS failover; recovery should be much more robust now
5962 * remove failed OSCs from LOVs (only lov_create uses this so far)
5963 * the lustre-HOWTO was brought (more) up to date (582544)
5965 2002-10-23 Phil Schwan <phil@clusterfs.com>
5968 - in-use dentries weren't being reused properly (617851)
5969 - prevent multiple LDLM setup (599178)
5970 - fix LOV size calculations for truncate (617853)
5971 - fix client handling of MDS intent errors (POSIX)
5972 - fix permission bug in lovstripe.c test (624321)
5973 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5974 - truncate past end of file could corrupt data
5975 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5976 - a race in recovery could return ETIMEDOUT to apps (623947)
5977 - building outside the source directory was fixed
5978 * the lustre-HOWTO was brought (more) up to date (582544)
5979 * major progress was made on recovery functionality
5981 2002-10-10 Phil Schwan <phil@clusterfs.com>
5984 - recovery deadlock fix
5985 - rm -rf causes LBUG fix (617817)
5986 - file open by multiple tasks fix (618962)
5987 - directory permissions bugs (602707 and 620007)
5988 - journal_stop fixed with locking (611313)
5989 - O_APPEND failures resolved (618273, perhaps 614459)
5990 - lconf PATH fix (619770)
5991 - IA64 build fix (621450)
5992 - RPC buffer sizes scale with amount of memory
5994 2002-10-01 Phil Schwan <phil@clusterfs.com>
5997 - locks would be cancelled without throwing away data pages,
5998 resulting in inconsistent data (605627)
5999 - inode attributes were not always being refreshed (605627, 612449)
6000 - lconf now continues to cleanup after lctl reports an error
6001 - MDS now enforces user permissions (602707)
6002 - lprocfs cleanup fixed, but not yet enabled (614157)
6003 - fixed infinite server hang, should a client not respond to an AST
6004 - avoid going into recovery if user calls readlink() with a buffer
6005 that's too small (613941)
6006 - AST RPCs no longer require replies (614867) -- this may be changed
6007 - don't crash server if client sends an IOV that's too big (611336)
6008 - fixed lock conversion deadlock (611892)
6009 - fixed the following of symlinks (614622)
6010 * recovery: the server can remove locks from a client that dies, other
6011 clients can make progress
6012 * more extN patch fixes
6013 * compile-time configurable ptlrpc buffer allocations
6015 - collaborative read cache document
6016 - Lustre Lite Performance CDR document-in-progress
6018 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6021 - fix typo in patch-2.4.18
6023 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6026 - clear ptlrpc request each time in handle_incoming_request()
6027 - unlink of files now destroys the object on the OST
6029 2002-09-19 Peter Braam <braam@clusterfs.com>
6031 * add hard link support
6032 * change obdfile creation method
6033 * kernel patch changed
6035 2002-09-19 Peter Braam <braam@clusterfs.com>
6038 - stack overflow bug in extN fixed
6040 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6042 * documentation updates
6043 - add man pages for config tools
6044 - update tests/README to describe testing with new config tools
6045 - finish metadata API descriptions
6046 * bug fixes and cleanups
6047 - statfs workaround for 16TB limit
6048 - LOV stripe allocation improved, can stripe on subset of OSTs
6049 - LOV file size/IO offset was wrong for files > 4GB in size
6050 - object EA data was being dropped, caused files to be unreadable
6051 - memory overflow with non-LOV OST caused memory corruption
6052 - fixed regression tests to work with new config tools, obdfilter
6053 - fixed bug when directory size became larger than 1 block
6054 - fixed bug (for single client case) when PWD was deleted
6055 - invalidate local directory pages when doing intent-based ops
6056 - avoid LDLM oops when lock callback contained bad data
6058 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6060 * documentation updates
6061 * bug fixes and cleanups
6062 - configuration tools
6065 - 64-bit compile warnings
6066 - 64-bit internal statfs data
6068 * test_brw on persistent OST devices
6070 * lprocfs (disabled)
6072 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6074 * documentation updates
6075 * bug fixes and cleanups
6076 * configuration tools
6078 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6081 * many small fixes to 0_5_4
6082 * io/network handling
6083 * thinkos in MDS operations
6085 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6088 * crucial basic fixes to 0.5.3
6089 * IOR, Iozone work over Elan
6092 2002-08-07 Phil Schwan <phil@clusterfs.com>
6093 * version 0_5_3, our first alpha
6094 * we use the new Portals iovs
6095 * documentation updates
6096 * bug fixes and cleanups
6097 * small changes in the DLM wire protocol
6099 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6100 * version 0_5_1 with some initial stability,
6101 * locking on MD and file I/O.
6102 * documentation updates
6103 * several bug fixes since 0.5.0
6104 * small changes in wire protocol
6106 2002-07-18 Phil Schwan <phil@clusterfs.com>
6108 * delivered as Lustre Light Alpha
6109 * fixed a crash after handling invalid MDS requests
6110 * fixed directory pages for architectures with non-4k pages sizes
6112 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6113 * release version v0_4_4
6114 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6116 * Fixes a number of interruption problems with OST operations.
6117 * Update documentation for portals header changes
6118 * Move all wire protocol structs/defines to lustre_idl.h
6119 * Fixes symlink length bug.
6120 * Add tcpdump to repository.
6122 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6123 * release version v0_4_3
6124 * Fixes statfs for inodes on extN.
6125 * Fixes bug in runtests which would delete /etc/hosts.
6126 * Use 64-bit object IDs wherever possible (not into VFS though)
6127 Remove ost_get_info, which is unused by lustre, and out of date.
6129 2002-07-03 Peter Braam <braam@clusterfs.com>
6130 * release version v0_4_2 Fixes a lookup error (type not passed)
6131 * move forward to head of Portals
6132 * move forward to latest Lustre kernel
6134 2002-06-25 Peter Braam <braam@clusterfs.com>
6135 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob