1 tbd Sun Microsystems, Inc.
3 * Server support for kernels:
4 2.6.18-164.11.1.el5 (RHEL 5)
5 2.6.18-164.11.1.0.1.el5 (OEL 5)
6 * Client support for unpatched kernels:
7 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
8 2.6.18-164.11.1.el5 (RHEL 5),
9 2.6.18-164.11.1.0.1.el5 (OEL 5)
10 2.6.16.60-0.42.8 (SLES 10),
12 2.6.29.4-167.fc11 (FC11)
13 * Recommended e2fsprogs version: 1.41.10-sun2
16 Bugzilla : 21829, 22312
17 Description: Fix llobdstat.
18 Details : 1. Now updating the write_bytes counter
19 2. Port script fixes from 1.8.
23 Description: decrease the usage of memory on clients.
24 Details : 1. On clients, recycle dentries and inodes unused.
25 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
30 Description: llobdstat fix and enhancement
31 Details : add a counter to set a limit to how many samples will be returned
32 fix a wildcard in the path to limit to obdfilter stats only
36 Description: lfs check only allowed for root
37 Details : Code cleanups and added a regresion test of non-root use of
42 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
43 Details : Fix hr->hr_index overload.
47 Description: Fix FIEMAP functionality.
51 Description: FC11 patchless client support added
55 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
56 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
60 Description: New acc-sm script and various test-framework enhancments.
64 Description: Previously landed patch was reverted.
68 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
72 Description: Add ldiskfs maxdirsize mount option.
76 Description: Fix mkfs.lustre support for 16TB LUNs.
80 Description: Print mmp_check_interval
81 Details : print mmp_check_interval and make it possible to abort
82 mount operation in case it takes too long.
86 Description: Add cascading_rw to lustre mpi tests.
90 Description: Store SOM data in EA.
91 Details : Also several SOM related cleanups.
95 Description: Add refcount to md_open_data
99 Description: Use refcount on lov_request_set to prevent memory corruption.
103 Description: Statahead: use parent dir's i_mutex for synchronization.
107 Description: loadgen improvements.
111 Description: Enable peer health detection in ksocklnd and ko2iblnd.
115 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
119 Description: Fix race on obd_devcie:md_stats
123 Description: Fix SLES 11 client builds.
124 Details : Resolved uninitalized pointer gcc warnings.
128 Description: Allow non-root access for "lfs check".
129 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
133 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
135 Severity : enhancement
137 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
138 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
142 Description: Sanity test_27p fails.
143 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
144 Sanity tests 27 were cleaned up and improved.
148 Description: Unify req format on client/servers
149 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
150 buffers and swabbers approach..
155 Description: drop join file stuff.
157 Severity : enhancement
159 Description: Add 2.6.27 and SLES11 (patchless client) support
163 Description: Fix in ptlrpc_expire_one_request() to print the signed time
166 Severity : enhancement
168 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
171 Severity : enhancement
173 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
175 Severity : enhancement
177 Description: correctly handle big reply message.
178 Details : send LNet event if reply is bigger then buffer and adjust this buffer
183 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
185 Severity : enhancement
187 Description: Add support for OEL5.
191 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
195 Description: client could not reconnect to OST because of an active request.
196 Details : The client now retries to reconnect to the same server, if a
197 connect request failed with EBUSY or -EAGAIN. Test case 35b
198 conf-sanity.sh has been created.
202 Description: Don't allow make backward step on assignin osc next id.
203 Details : race between allocation next id and ll_sync thread can be cause
204 of set wrong osc next id and can be kill valid ost objects.
208 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
212 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
216 Description: Error handling in osc_statfs_interpret() has been improved.
217 Details : Check in osc_statfs_interpret() for EBADR.
221 Description: Do not send statfs() requests to OSTs disabled by administrator.
222 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
224 Severity : enhancement
226 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
228 Severity : enhancement
230 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
232 Severity : enhancement
234 Description: Enhancement for lfs(1) command to use numeric uid/gid.
238 Description: Improved cleanup procedure for conf-sanity test_46a.
242 Description: Do not start bulk transfer, if client tried to reconnect.
243 Details : Do not start bulk transfer, if client tried to reconnect,
244 but failed due to active rpcs. Check for exp_abort_active_req
245 has been implemented.
249 Description: Prevent inconsistences between linux and lustre mount structures.
250 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
251 Make the sleep interruptible.
253 Severity : enhancement
255 Description: provide server to client comms path
256 Details : server to client communications path via new LDLM_SET_INFO rpc
260 Description: truncate starts GFP_FS allocation under transaction causing deadlock
261 Details : ldiskfs_truncate calls grab_cache_page which may start page
262 allocation under an open transaction. This may lead to
263 calling prune_icache with consequent lustre reentrance.
267 Description: send correctly lsm on open replay
268 Details : MDS is trust to LSM size on replay open, but client can set wrong size
273 Description: enable client interface failover
274 Details : When a child reconnects from another NID, properly update export
275 nid hash position and ldlm reverse import.
277 Severity : enhancement
279 Description: Communicate OST degraded/readonly state via statfs to MDS
280 Details : Flags in the statfs returned from OSTs indicate whether the
281 OST is in a degraded RAID state, or if the filesystem has
282 turned read-only after a filesystem error is detected.
287 Description: don't panic if EPROTO was hit when reading symlink
288 Details : correctly handling request reference in error cases.
290 Severity : enhancement
292 Description: Add LustreNetLink, a kernel-userspace communcation path.
294 Severity : enhancement
296 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
301 Description: access to llog context before init.
302 Details : move handling CATALOGS file at osc layer and forbid access to llog
306 Frequency : rare, only when client is writing to OST on local node
308 Description: Avoid deadlock for local client writes
309 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
310 memory freeing context. This allows OST threads to set the
311 PF_MEMALLOC flag on task structures in order to allocate memory
312 from reserved pools and complete IO.
313 Use GFP_HIGHUSER for OST allocations for non-local client writes,
314 so that the OST threads generate memory pressure and allow
315 inactive pages to be reclaimed.
317 Severity : enhancement
319 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
324 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
325 Details : move ldlm namespace creation in setup phase to avoid grab
326 _lprocfs_lock with cli_sem held.
330 Description: Temporarily disable grant shrink.
331 Details : Disable the feature for debugging.
334 Frequency : only when formatting test filesystems
336 Description: Unable to run several mkfs.lustre on loop devices at the same time
337 Details : mkfs.lustre returns error 256 on the concurrent loop devices
338 formatting. The solution is to proper handle the error.
340 Severity : enhancement
342 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
344 Severity : enhancement
346 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
351 Description: resolve race between obd_disconnect and class_disconnect_exports
352 Details : if obd_disconnect will be called to already disconnected export he
353 forget release one reference and osc module can't unloaded.
355 Severity : enhancement
357 Description: Allow tuning service thread via /proc
358 Details : For each service a new
359 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
360 created that can be used to set min/max thread counts, and get the
361 current number of running threads.
366 Descriptoin: don't return error if have partially created objects for file.
367 Details : lov_update_create_set uses set->set_success as index for created
368 objects, so if some requests will be failed, they will have hole
369 at end of array and qos_shrink_lsm for allocate correct lsm.
371 Severity : enhancement
373 Description: Update OFED support to 1.4.1
378 Description: Reduce small size read RPC
379 Details : Set read-ahead limit for every file and only do read-ahead when
380 available read-ahead pages are bigger than 1M to avoid small size
383 Severity : enhancement
385 Description: move AT tunable parameters for more consistent usage
386 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
388 Severity : enhancement
390 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
391 Details : allow skip disconnected ost for send statfs request and hide error
397 Description: fix racy locking of mballoc block bitmaps causing BUG
398 Details : The locking of the mballoc buddy bitmap and the in-memory
399 block bitmap was using two different spin locks in some
400 cases. This made it possible to incorrectly access the
401 mballoc bitmap while another process was modifying it,
402 causing a sanity assertion to fail. While no on-disk corruption
403 was reported, there was some risk of this happening.
406 Frequency : rare, on llog test 6
408 Descriptoin: don't allow connect to already connected import
409 Details : allowing connect to already connected import is hide connecting problem.
414 Descriptoin: allow kill process which wait statahead result
415 Details : for some reasons 'ls' can stick in waiting result from statahead,
416 in this case need way for kill this process.
418 Severity : enhancement
420 Description: Add state history info file, enhance import info file
421 Details : Track import connection state changes in a new osc/mdc proc file;
422 add overview-type data to the osc/mdc import proc file.
424 Severity : enhancement
426 Description: MDS create should not wait for statfs RPC while holding DLM lock.
428 Severity : enhancement
430 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
435 Descriptoin: OST grant too much space to client even there are not enough space.
436 Details : Client will shrink its grant cache to OST if there are no write
437 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
438 this grant cache if there are already not enough avaible space
439 (left_space < total_clients * 32M).
442 Frequency : start MDS on uncleanly shutdowned MDS device
444 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
445 Details : stay in waiting mds<>ost recovery finished produce random bugs
446 due race between two ll_sync thread for one lov target. send
447 ACTIVATE event only if connect realy finished and import have
451 Frequency : rare, connect and disconnect target at same time
453 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
454 Details : don't call obd_disconnect under lov_lock. this long time
455 operation and can block ptlrpcd which answer to connect request.
460 Descriptoin: don't lose wakeup for imp_recovery_waitq
461 Details : recover_import_no_retry or invalidate_import and import_close can
462 both sleep on imp_recovery_waitq, but we was send only one wakeup
466 Frequency : always with long access acl
468 Descriptoin: mds can't pack reply with long acl.
469 Details : mds don't control size of acl but they limited by reint/getattr
472 Severity : enhancement
474 Description: Update to SLES10 kernel-2.6.16.60-0.33.
476 Severity : enhancement
478 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
481 Frequency : start MDS on uncleanly shutdowned MDS device
483 Descriptoin: aborting recovery hang on MDS
484 Details : don't throttle destroy RPCs for the MDT.
487 Frequency : on remount
489 Description: external journal device not working after the remount
490 Details : clear dev_rdonly flag for external journal devices in
496 Description: shutdown vs evict race
497 Details : client_disconnect_export vs connect request race.
498 if client will evicted at this time - we start invalidate
499 thread without referece to import and import can be freed
503 Frequency : rare, need acl's on inode.
505 Description: client can't handle ost additional correctly
506 Details : if ost was added after client connected to mds client can have
507 hit lnet_try_match_md ... to big messages to wide striped files.
508 in this case need teach client to handle config events about add
509 lov target and update client max ea size at that event.
511 Severity : enhancement
513 Description: Changelogs
514 Details : Changelogs are a lightweight mechanism to track filesystem
515 metadata and namespace changes. The changelog is recorded
516 permanently on the MDTs, and is periodically "consumed" / purged
517 when records are no longer needed.
519 Severity : enhancement
521 Description: compact fld format with extents
522 Details : Store range of seq rather than every seq in FLD. Seq
523 controller update FLD rather than clients. In Case of CMD, mdt0
524 has FLD, all other metadata server act as non persistent proxy
525 for FLD queries and cache fld entries in fld cache.
530 Description: don't skip ost target if they assigned to file
531 Details : Drop slow OSCs if we can, but not for requested start idx.
532 This means "if OSC is slow and it is not the requested
533 start OST, then it can be skipped, otherwise skip it only
534 if it is inactive/recovering/out-of-space.
538 Description: more cleanup in mds_lov
539 Details : not send LOV EA under replay, we can't know about they size at this
540 time. Don't allow client connect to mds before any ost connected,
541 for avoid problems with LOV EA size and returning EIO to client.
543 Severity : enhancement
545 Description: Interoperability at server side (Disk interoperability)
547 Severity : enhancement
549 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
551 Severity : enhancement
553 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
555 Severity : enhancement
557 Description: New client IO stack (CLIO).
559 Severity : enhancement
561 Description: Commit on sharing. Eliminate inter-client dependencies between
562 uncommitted transactions by doing transaction commits.
563 Thereby clients may recovery independently.
566 Frequency : Create a symlink file with a very long name
568 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
569 Details : If there is no extra space in the request for early cancels,
570 ldlm_req_handles_avail() returns 0 instead of a negative value.
572 Severity : enhancement
574 Description: Add /proc entry for import status
575 Details : The mdc, osc, and mgc import directories now have
576 an import directory that contains useful import data for debugging
579 Severity : enhancement
581 Description: Re-disable certain /proc logging
582 Details : Enable and disable client's offset_stats, extents_stats and
583 extents_stats_per_process stats logging on the fly.
586 Frequency : Only on FC kernels 2.6.22+
588 Description: oops in statahead
589 Details : Do not drop reference count for the dentry from VFS when lookup,
590 VFS will do that by itself.
592 Severity : enhancement
594 Description: Generic /proc file permissions
595 Details : Set /Proc file permissions in a more generic way to enable non-
596 root users operate on some /proc files.
600 Description: Hitting mdc_commit_close() ASSERTION
601 Details : Properly handle request reference release in
602 ll_release_openhandle().
606 Description: quota recovery deadlock during mds failover
607 Details : This patch includes att18982, att18236, att18237 in bz14840.
609 1. fix osts hang when mds does failover with quotaon
610 2. prevent watchdog storm when osts threads wait for the
615 Frequency : only patchless client
616 Description: add workaround for race between add/remove dentry from hash
618 Severity : enhancement
620 Description: Allow OST glimpses to return PW locks
624 Description: LBUG when llog conf file is full
625 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
629 Description: Prevent import from entering FULL state when server in recovery
633 Description: service mount cannot take device name with ":"
634 Details : Only when device name contains ":/" will mount treat it as
640 Description: replace ptlrpcd with the statahead thread to interpret the async
641 statahead RPC callback
645 Frequency : on recovery
646 Description: I/O failures after umount during fail back
647 Details : if client reconnected to restarted server we need join to recovery
648 instead of find server handler is changed and process self
649 eviction with cancel all locks.
651 Severity : enhancement
653 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
655 Severity : enhancement
657 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
659 Severity : enhancement
661 Description: Upcall on Lustre log has been dumped
662 Details : Allow for a user mode script to be called once a Lustre log has
663 been dumped. It passes the filename of the dumped log to the
664 script, the location of the script can be specified via
665 /proc/sys/lnet/debug_log_upcall.
670 Description: avoid idr_remove called for id which is not allocated.
671 Details : Move assigment s_dev for clustered nfs to end of initialization,
672 to avoid problem with error handling.
677 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
678 Details : When connection is reused this not moved from CONN_UNUSED_HASH
679 into CONN_USED_HASH and this prodice warning when put connection
680 again in unused hash.
683 Severity : enhancement
685 Description: File striping can now be set to use an arbitrary pool of OSTs.
687 Severity : enhancement
689 Description: Export bytes_read/bytes_write count on OSC/OST.
693 Description: Early reply size mismatch, MGC loses connection
694 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
695 the connect flags are properly negotiated.
700 Description: quotas are not honored with O_DIRECT
701 Details : all writes with the flag O_DIRECT will use grants which leads to
702 this problem. Now using OBD_BRW_SYNC to guard this.
706 Description: add quota statistics
707 Details : 1. sort out quota proc entries and proc code.
708 2. add quota statistics
710 Severity : enhancement
712 Description: enable quota support for HEAD.
716 Description: Properly propagate oinfo flags from lov to osc for statfs
717 Details : restore missing copy oi_flags to lov requests.
719 Severity : enhancement
721 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
722 : l_getgroups(8), lst(8), routerstat(8)
723 Details : included man pages for llobdstat(8), llstat(8),
724 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
726 Severity : enhancement
728 Description: configure's --enable-quota should check the
729 : kernel .config for CONFIG_QUOTA
730 Details : configure is terminated if --enable-quota is passed but
731 : no quota support is in kernel
735 Description: Remove portals compatibility
736 Details : Remove portals compatibility, not interoperable with releases
741 Description: Resolve device initialization race
742 Details : Prevent proc handler from accessing devices added to the
743 obd_devs array but yet be intialized.
745 Severity : enhancement
747 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
749 Severity : enhancement
751 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
756 Description: Using wrong pointer in osc_brw_prep_request
757 Details : Access to array[-1] can produce panic if kernel compiled with
758 CONFIG_PAGE_ALLOC enabled
762 Description: Client runs out of low memory
763 Details : Consider only lowmem when counting initial number of llap pages
767 Description: *optional* service tags registration
768 Details : if the "service tags" package is installed on a Lustre node
769 When the filesystem is mounted, a local-node service tag will
770 be created. See http://inventory.sun.com/ for more information
771 about the Service Tags asset management system.
775 Description: Kernel BUG tries to release flock
776 Details : Lustre does not destroy flock lock before last reference goes
777 away. So always drop flock locks when client is evicted and
778 perform unlock regardless of successfulness of speaking to MDS.
782 Description: add refcount protection for osc callbacks, avoid panic on shutdown
786 Description: sanity test 65a fails if stripecount of -1 is set
787 Details : handle -1 striping on filesystem in ll_dirstripe_verify
792 Description: ASSERTION(CheckWriteback(page,cmd)) failed
793 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
796 Severity : enhancement
798 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
802 Description: do not process already freed flock
803 Details : flock can possibly be freed by another thread before it reaches
804 to ldlm_flock_completion_ast.
808 Description: LBUG during stress test
809 Details : Need properly lock accesses the flock deadlock detection list.
813 Description: oops in page fault handler
814 Details : kernel page fault handler can return two special 'pages' in error
815 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
819 Description: timeout with invalidate import.
820 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
821 should be handled by ptlrpcd. This produce long age waiting and
822 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
824 Severity : enhancement
826 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
830 Description: enable MGS and MDT services start separately
831 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
832 a co-located MGS without starting the MGS, which is a complement
833 to 'nosvc' mount option.
837 Frequency : after recovery
838 Description: precreate to many object's after del orphan.
839 Details : del orphan st in oscc last_id == next_id and this triger growing
840 count of precreated objects. Set flag LOW to skip increase count
841 of precreated objects.
845 Frequency : rare, on clear nid stats
846 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
847 Details : when clean nid stats sometimes try destroy live entry,
848 and this produce panic in free.
852 Description: Stack overflow during MDS log replay
853 ease stack pressure by using a thread dealing llog_process.
857 Description: wait until IO finished before start new when do lock cancel.
858 Details : VM protocol want old IO finished before start new, in this case
859 need wait until PG_writeback is cleared until check dirty flag and
860 call writepages in lock cancel callback.
862 Severity : enhancement
864 Description: using special macro for print time and cleanup in includes.
868 Description: mds_mfd_close() ASSERTION(rc == 0)
869 Details : In mds_mfd_close(), we need protect inode's writecount change
870 within its orphan write semaphore to prevent possible races.
874 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
878 Frequency : rare, on shutdown ost
879 Description: don't hit live lock with umount ost.
880 Details : shrink_dcache_parent can be in long loop with destroy dentries,
881 use shrink_dcache_sb instead.
885 Description: don't panic with use echo client
886 Details : echo client pass NULL as client nid pointer and this produce null
891 Description: fix build on ppc32
892 Details : compile code with -m64 flag produce wrong object file for ppc32.
896 Description: add message levels for liblustreapi
900 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
901 Details : Move the 'good_osts' check before the 'total_bavail' check. This
902 will result in an -EAGAIN and in the exit call path we call
903 alloc_rr() which will with increasing aggressiveness attempt to
904 aquire precreated objects on the minimum number of required OSCs.
908 Description: Use old size assignment to avoid deadlock
909 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
910 the scheduling while holding a spinlock. We do not need locking
911 for size in ll_update_inode() because size is only updated from
912 the MDS for directories or files without objects, so there is no
913 other place to do the update, and concurrent access to such inodes
914 are protected by the inode lock.
918 Description: resolve "_IOWR redefined" build error on SLES10
922 Description: dump the memory debugging after all modules are unloaded to
923 suppress false negative in conf_sanity test 39
925 Severity : enhancement
927 Description: build kernel-ib packages for OFED 1.3 in our release cycle
932 Description: fix SLES kernel versioning
933 Details : the kernel version for our SLES 10 kernel did not include a "-"
934 before the "smp" at the end. while this was not a problem in
935 general, it did mean that software trying to use the kernel
936 version to try to detect a vendor specific kernel would fail.
937 this was most evident by the OFED build scripts.
941 Description: Don't update lov_desc members until making sure they are valid
942 Details : When updating lov_desc members via proc fs, need fix their
943 validities before doing the real update.
947 Description: don't put request into delay list while invalidate in flight.
948 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
949 import in flight. this produce timeout for invalidate and sometimes
950 can cause stale data.
954 Frequency : on ppc only
955 Description: not convert ost objects for directory because it's not exist.
956 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
960 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
962 Severity : enhancement
964 Description: Update to support 2.6.22.14 vanilla kernel.
968 Frequency : rare, on recovery
969 Description: read procfs can produce deadlock in some situation
970 Details : Holding lprocfs lock which send rpc can produce block for destroy
971 obd objects and this also block reconnect with -EALREADY. This
972 isn't fix all lprocfs bugs - but make it rare.
974 Severity : enhancement
976 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
979 Frequency : frequent on X2 node
981 Description: mdc_set_open_replay_data LBUG
982 Details : Set replay data for requests that are eligible for replay.
986 Description: lustre_mgs: operation 101 on unconnected MGS
987 Details : When MGC is disconnected from MGS long enough, MGS will evict the
988 MGC, and late on MGC cannot successfully connect to MGS and a lot
989 of the error messages complaining that MGS is not connected.
993 Frequency : on network error
994 Description: panic with double free request if network error
995 Details : mdc_finish_enqueue is finish request if any network error ocuring,
996 but it's true only for synchronus enqueue, for async enqueue
997 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1000 Severity : enhancement
1002 Description: client-side metadata stat-ahead during readdir(directory readahead)
1003 Details : perform client-side metadata stat-ahead when the client detects
1004 readdir and sequential stat of dir entries therein
1007 Frequency : on start mds
1009 Description: Implement get_info(last_id) in obdfilter.
1012 Frequency : occasional
1014 Description: Correctly check stale fid, not start epoch if ost not support SOM
1015 Details : open with flag O_CREATE need set old fid in op_fid3 because
1016 op_fid2 was overwritten with new generated fid, but mds can answer
1017 with one of these two fids and both is not stale. Setattr
1018 incorrectly started an epoch and assume will be called
1019 done_writing, but without SOM done_writing ever being called.
1022 Frequency : rare, depends on device drivers and load
1024 Description: MDS or OSS nodes crash due to stack overflow
1025 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1026 In some cases, in conjunction with device drivers that use a lot
1027 of stack the MDS (or possibly OSS) service threads could overflow
1028 the stack. One change which was identified to consume additional
1029 stack has been reworked to avoid the extra stack usage.
1032 Frequency : occasional
1034 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1035 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1036 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1037 oscc_has_objects and see OSC has no objests and call
1038 oscc_internal_create to resend create request.
1040 Severity : enhancement
1042 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1044 Severity : enhancement
1046 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1049 Frequency : very rare
1051 Description: Fix replay if there is an un-replied request and open
1052 Details : In some cases, older replay request will revert the
1053 mcd->mcd_last_xid on MDS which is used to record the client's
1054 latest sent request.
1056 Severity : enhancement
1058 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1060 Severity : enhancement
1062 Description: Add rhel5 support to HEAD.
1064 Serverity : enhancement
1066 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1071 Description: Don't allow skipping OSTs if index has been specified.
1072 Details : Don't allow skipping OSTs if index has been specified, make
1073 locking in internal create lots better.
1077 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1078 Details : access to bitfield in structure is always rounded to long
1079 and this produce problem with not atomic change any bit.
1083 Description: Lustre make rpms failed.
1084 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1085 builds Lustre rpms from tarball.
1088 Frequency : rare on shutdown ost
1090 Description: If llog cancel was not send before clean_exports phase, this can
1091 produce deadlock in llog code.
1092 Details : If llog thread has last reference to obd and call class_import_put
1093 this produce deadlock because llog_cleanup_commit_master wait when
1094 last llog_commit_thread exited, but this never success because was
1095 called from llog_commit_thread.
1099 Description: allow userland application know is lost one of stripes.
1100 Details : fill lvb_blocks with error code on ost and return it to
1101 application if error flag found.
1105 Description: NULL lov_tgts causing MDS oops
1106 Details : more safe checks for NULL lov_tgts for avoid oops.
1108 Severity : enhancement
1110 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1114 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1115 Details : If an OST is down the MDS will hang indefinitely in
1116 obd_statfs_rqset() waiting for the statfs data. While for
1117 MDS QOS usage of statfs, it should not stuck in waiting.
1119 Severity : enhancement
1121 Description: remote_acl support
1122 Details : Support ACL-based permission check for remote user.
1123 Support setfacl/getfacl for remote user with the utils
1124 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1125 the system "{s,g}etfacl" utils.
1127 Severity : enhancement
1129 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1131 Severity : enhancement
1133 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1137 Description: make mgs_setparam() handle fsname containing dash
1138 Details : fsname containing a dash does not work with lctl conf_param
1140 Severity : enhancement
1142 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1144 Severity : enhancement
1146 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1150 Frequency : rare, at shutdown
1151 Description: access already free / zero obd_namespace.
1152 Details : if client_disconnect_export was called without force flag set,
1153 and exist connect request in flight, this can produce access to
1154 NULL pointer (or already free pointer) when connect_interpret
1155 store ocd flags in obd_namespace.
1159 Frequency : only at startup
1160 Description: not alloc memory with spinlock held.
1161 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1162 if any spinlock held.
1164 Severity : enhancement
1166 Description: make lustre randomly fail allocating memory
1167 Details : Make lustre randomly failed allocating memory for testing purpose.
1169 Severity : enhancement
1171 Description: lost problems with lov objid file
1172 Details : Fixes some scability and access to not inited memory problems
1173 in work with lov objdid file.
1178 Description: lfs find does not continue on file error
1179 Details : Continue other files processing when a file/dir is absent.
1183 Description: Inconsistent usage of lustre_pack_reply()
1184 Details : Standardize the usage of lustre_pack_reply() such that it
1185 always generate a CERROR on failure.
1188 Frequency : occasional
1190 Description: MDS hang or stay in waiting lock
1191 Details : If client receive lock with CBPENDING flag ldlm need send lock
1192 cancel as separate rpc, to avoid situation when cancel request
1193 can't processed due all i/o threads stay in wait lock.
1197 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1201 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1205 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1209 Description: Update to RHEL4 latest kernel.
1211 Severity : enhancement
1213 Description: Build SLES10 patchless client fails
1214 Details : The configure was broken by run ./configure with
1215 --with-linux-obj=.... argument for patchless client. When the
1216 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1217 can't search header adequately. Use absolute path such as
1218 -I($LINUX)/include instead.
1222 Description: interrupt oig_wait produce painc on resend.
1223 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1224 private set, and this produce situation when rq_ptlrpcd_data not
1225 copyed to new allocated request and triggered LBUG on assert
1226 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1227 ptlrpcd set if request is changed and can be safety changed to use
1232 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1233 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1234 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1235 these macros and replace them with their difinitions instead.
1239 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1240 Details : Update which_patch & target file for SLES10 latest kernel.
1243 Description: Remove client patches from SLES 10 kernel.
1244 Details : This causes SLES 10 clients to behave as patchless clients
1245 even on a Lustre-patched (server) kernel.
1247 Severity : enhancement
1249 Description: self-adjustable client's lru lists
1250 Details : use adaptive algorithm for managing client cached locks lru
1251 lists according to current server load, other client's work
1252 pattern, memory activities, etc. Both, server and client
1253 side namespaces provide number of proc tunables for controlling
1256 Severity : enhancement
1258 Description: light-weight GSS support
1259 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1260 and reduce performance overhead.
1262 Severity : enhancement
1264 Description: Linux keyring support
1265 Details : Support using service of Linux keyring for Lustre GSS internal
1266 context refresh/cache mechanism.
1270 Description: Fix errors in lfs documentation
1271 Details : Fixes man pages
1275 Description: don't use GFP_* in generic Lustre code.
1276 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1280 Description: obdclass is limited by single OBD_ALLOC(idarray)
1281 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1285 Description: warnings with build patchless client with vanila 2.6.19 and up
1286 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1287 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1291 Frequency : Only for SLES
1292 Description: sanity_quota fail test_1
1293 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1294 file, which makes TSTID[2] inunique.
1298 Frequency : Only for SLES9
1299 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1300 Details : Update target/ChangeLog/which_patch .
1304 Frequency : Only for test_10 in sanity-quota.sh
1305 Description: a bug in quota test code
1306 Details : checking a test flag in a wrong place
1310 Frequency : Only for test_13 in sanity-quota.sh
1311 Description: a bug in quota test script
1312 Details : shouldn't check overfull allocated quota
1316 Description: jbd statistics
1317 Details : Port older jbd statistics patch for sles10
1319 Severity : enhancement
1321 Description: Add printing inode info into message about error in writepage.
1326 Description: reply_lock_interpret crash due to race with it and lock cancel.
1327 Details : Do not replay locks that are being cancelled. Do not reference
1328 locks by their address during replay, just by their handle.
1332 Frequency : When flocks are used.
1333 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1334 Details : Flock locks might destroy just granted lock if it could be merged
1335 with another existing flock, this is done in completion handler,
1336 so teach ldlm_cli_enquque_fini that this is a valid case for
1342 Description: Oops in read and write path when failing to allocate lock.
1343 Details : Check if lock allocation failed and return error back.
1347 Description: lstripe command fails for valid OST index
1348 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1349 instead of lov->desc.ld_active_tgt_count.
1353 Description: sanity.sh failed test 103
1354 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1359 Description: df doesn't work properly if diskfs blocksize != 4K
1360 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1364 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1365 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1367 Severity : enhancement
1369 Description: Adaptive timeouts
1370 Details : RPC timeouts adapt to changing server load and network
1371 conditions to reduce resend attempts and improve recovery time.
1375 Description: llapi_file_create() does not allow some changes
1376 Details : add llapi_file_open() that allows specifying the mode and
1377 open flags, and also returns an open file handle.
1381 Description: merge and cleanup kernel patches.
1382 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1386 Description: Add journal checksum support.(Kernel part)
1387 Details : The journal checksum feature adds two new flags i.e
1388 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1389 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1390 indicates that the commit block contains the checksum for
1391 the blocks described by the descriptor blocks. Now commit
1392 record can be sent to disk without waiting for descriptor
1393 blocks to be written to disk. This behavior is controlled
1394 using JBD2_FEATURE_ASYNC_COMMIT flag.
1398 Description: OSS needs mutliple precreate threads
1399 Details : Add ability to start more than one create thread per OSS.
1403 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1404 Details : execute lfs setstripe on client
1408 Description: mds_obd_create error creating tmp object
1409 Details : When the user sets quota on root, llog will be affected and can't
1410 create files and write files.
1413 Frequency : Always on ia64 patchless client, and possibly others.
1415 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1416 Details : This allows the patchless client to be loaded on architectures
1417 without this export.
1421 Description: RedHat Update kernel for RHEL5
1422 Details : Add the kernel config file for RHEL5.
1426 Description: RedHat Update kernel for RHEL5
1427 Details : Modify the kernel config file more closer RHEL5.
1431 Description: Build failure against Centos5 (RHEL5)
1432 Details : Define PAGE_SIZE when it isn't present.
1436 Description: test_7 on sanity-quota.sh can't be used on separate machines
1437 Details : add facet to handle it
1441 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1442 Details : with NFS, the anon dentry's parent was set to itself in
1443 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1444 corresponding dentry other than use rec->ur_name.
1446 Severity : enhancement
1448 Description: lfs setstripe enhancement
1449 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1453 Description: enable data checksumming by default
1454 Details : enable checksum by default, allow --disable-checksum
1455 configure option and "-o nochecksum" mount option. Checksums
1456 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1460 Description: lustre support for RHEL5
1461 Details : Add support for RHEL5.
1465 Description: Client eviction due to failover config
1466 Details : after a connection loss, the lustre client should attempt to
1467 reconnect to the last active server first before trying the
1468 other potential connections.
1472 Description: when mds and osts use different quota unit(32bit and 64bit),
1473 quota will be released repeatly.
1474 Details : void sending multiple quota reqs to mds, which will keep the
1475 status between the reqs.
1479 Description: osts not allocated evenly to files
1480 Details : change the condition to increase offset_idx
1483 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1485 Description: Data corruption for OSTs that are formatted larger than 2TB
1487 Details : When generating the bio request for lustre file writes the
1488 sector number would overflow a temporary variable before being
1489 used for the IO. The data reads correctly from Lustre (which
1490 will overflow in a similar manner) but other file data or
1491 filesystem metadata may be corrupted in some cases.
1495 Description: Tune the kernel for good SCSI performance.
1496 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1497 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1502 Description: rewrite ext2-derived code in obdclass/uuid.c
1503 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1504 from scratch preserving functionality.
1508 Description: "lfs find -obd UUID" prints directories
1509 Details : "lfs find -obd UUID" will return all directory names instead
1510 of just file names. It is incorrect because the directories
1511 do not reside on the OSTs.
1515 Description: use i_size_read and i_size_write in 2.6 port
1516 Details : replace inode->i_size access with i_size_read/write()
1519 Frequency : when removing large files
1521 Description: scheduling issue during removal of large Lustre files
1522 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1523 It causes scheduling issues when removing large files (17TB in the
1527 Frequency : only with liblustre clients on XT3
1529 Description: evictions taking too long
1530 Details : allow llrd to evict clients directly on OSTs
1533 Frequency : only on ppc
1535 Description: /proc/fs/lustre/devices broken on ppc
1536 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1537 We should make sure the type of 'index' is loff_t and then cast
1538 explicitly as needed below. Do not assign an explicitly cast
1542 Frequency : only for rhel5
1544 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1545 Details : Modify the target file & which_kernel.
1547 Severity : enhancement
1549 Description: omit set fsid for export NFS
1550 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1551 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1555 Frequency : Always, for kernels after 2.6.16
1556 Description: Fix warning idr_remove called for id=.. which is not allocated.
1557 Details : Last kernels save old s_dev before kill super and not allow
1558 to restore from callback - restore it before call kill_anon_super.
1562 Description: buffer overruns could theoretically occur
1563 Details : llapi_semantic_traverse() modifies the "path" argument by
1564 appending values to the end of the origin string, and a
1565 overrun may occur. Adding buffer overrun check in liblustreapi.
1569 Description: Fix error on 'ls .' at the top of the Lustre mount.
1570 Details : Don't revalidate dentry if it is a root dentry.
1574 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1575 Details : Modify vm-tunables-rhel4.patch.
1579 Description: Kernel config for 2.6.18-vanilla.
1580 Details : Modify targets/2.6-vanilla.target.in.
1581 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1582 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1583 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1584 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1588 Description: improve handling recoverable errors
1589 Details : if request processig with error which can be recoverable on server
1590 request should be resend, otherwise page released from cache and
1595 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1596 Details : Modify target file & which_patch.
1597 A flaw was found in the IA32 system call emulation provided
1598 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1599 value could be stored in the %RAX register, which could trigger an
1600 out-of-bounds system call table access. An untrusted local user
1601 could exploit this flaw to run code in the kernel
1602 (ie a root privilege escalation). (CVE-2007-4573).
1606 Description: O_DIRECT bypasses client statistics.
1607 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1608 read_ahead_stats were updated. copied the stats section from
1609 osc_send_oap_rpc() into async_internal().
1613 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1617 Description: change order of libsysio includes
1618 Details : '#include sysio.h' should always come before '#include xtio.h'
1622 Description: handle "serious error: objid * already exists" more gracefully
1623 Details : If LAST_ID value on disk is smaller than the objects existing in
1624 the O/0/d* directories, it indicates disk corruption and causes an
1625 LBUG(). If the object is 0-length, then we should use the existing
1626 object. This will help to avoid a full fsck in most cases.
1628 Severity : enhancement
1630 Description: adapt the lustre_config script to support the upgrade case
1631 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1632 server targets to 1.6 in parallel.
1636 Description: To avoid grant space > avaible space when the disk is almost
1637 full. Without this patch you might see the error "grant XXXX >
1638 available" or some LBUG about grant, when the disk is almost
1640 Details : In filter_check_grant, for non_grant cache write, we should
1641 check the left space by if (*left > ungranted + bytes), instead
1642 of (*left > ungranted), because only we are sure the left space
1643 is enough for another "bytes", then the ungrant space should be
1644 increase. In client, we should update cl_avail_grant only there
1645 is OBD_MD_FLGRANT in the reply.
1649 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1650 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1651 described in CVE-2007-4573. This problem could allow untrusted
1652 local users to gain root access.
1655 Frequency : when using O_DIRECT and quotas
1657 Description: Incorrect file ownership on O_DIRECT output files
1658 Details : block usage reported by 'lfs quota' does not take into account
1659 files that have been written with O_DIRECT.
1664 Description: touch file failed when fs is not full
1665 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1666 otherwise we can get ENOSP while fs is not full.
1670 Description: parallel lock callbacks
1671 Details : Instead of sending blocking and completion callbacks as separated
1672 requests, adding them to a set and sending in parallel.
1675 Frequency : only for Cray XT3
1676 Bugzilla : 12829/13455
1677 Description: Changing primary group doesn't change the group lustre assigns to
1679 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1680 overridden with the primary group provided by the group upcall,
1681 whereas the supplied fsgid can be trusted if it is in the list of
1682 supplementary groups returned by the group upcall.
1684 Severity : enhancement
1686 Description: Allow masking D_WARNING, D_ERROR messages from console
1687 Details : Console messages can now be disabled via lnet.printk.
1691 Description: User code with malformed file open parameter crashes client node
1692 Details : Before packing join_file req, all the related reference should be
1693 checked carefully in case some malformed flags cause fake
1694 join_file req on client.
1698 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1699 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1700 mode is equal to the granted mode. However, it should also check
1701 that LDLM_FL_CP_REQD is not set.
1705 Description: Heavy nfs access might result in deadlocks
1706 Details : After ELC code landed, it is now improper to enqueue any mds
1707 locks under och_sem, because enqueue might want to decide to
1708 cancel open locks for same inode we are holding och_sem for.
1712 Description: Client eviction while running blogbench
1713 Details : A lot of unlink operations with concurrent I/O can lead to a
1714 deadlock causing evictions. To address the problem, the number of
1715 oustanding OST_DESTROY requests is now throttled to
1716 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1717 callbacks are processed in priority.
1721 Description: enable ACLs on MDS by default
1722 Details : ACLs must be enabled on MDS by default.
1725 Frequency : PPC/PPC64 only
1727 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1728 Details : fix duplicated definitions between asm-ppc64/types.h and
1729 lustre_types.h on PPC.
1732 Frequency : PPC/PPC64 only
1734 Description: asm-ppc/segment.h does not exist
1735 Details : fix compile issue on PPC.
1739 Description: better handle error messages in extents code
1742 Frequency : RHEL4 only
1744 Description: mkfs is very slow on IA64/RHEL4
1745 Details : A performance regression has been discovered in the MPT Fusion
1746 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1747 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1748 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1750 Severity : enhancement
1752 Description: SNMP support enhancement
1753 Details : Adding total number of sampled request for an MDS node in snmp
1756 Severity : enhancement
1758 Description: Optimize ldlm waiting list processing for PR extent locks
1759 Details : When processing waiting list for read extent lock and meeting read
1760 lock that is same or wider to it that is not contended, skip
1761 processing rest of the list and immediatelly return current
1762 status of conflictness, since we are guaranteed there are no
1763 conflicting locks in the rest of the list.
1767 Description: Time out and refuse to reconnect
1768 Details : When the failover node is the primary node, it is possible
1769 to have two identical connections in imp_conn_list. We must
1770 compare not conn's pointers but NIDs, otherwise we can defeat
1771 connection throttling.
1775 Description: port llog fixes from b1_6 into HEAD
1776 Details : Port llog reference couting and some llog cleanups from b1_6
1777 (bug 10800) into HEAD, for protect from panic and access to already
1778 free llog structures.
1782 Description: Detect stride IO mode in read-ahead
1783 Details : When a client does stride read, read-ahead should detect that and
1784 read-ahead pages according to the detected stride pattern.
1788 Description: data checksumming impacts single node performance
1789 Details : add support for several checksum algorithm. Currently, only CRC32
1790 and Adler-32 are supported. The checksum type can be changed on
1791 the fly via /proc/fs/lustre/osc/*/checksum_type.
1795 Description: use adler32 for page checksums
1796 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1801 Description: build for x2 fails
1802 Details : fix compile issue on Cray systems.
1806 Description: Properly match for duplicate locks
1807 Details : Due to different lock order from skiplists code, we need to
1808 traverse entire list for now
1811 Frequency : only on PPC/SLES10
1813 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1814 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1815 be defined. Add a hack in mkfs_lustre.c to work around this compile
1820 Description: LASSERT on MDS when client holding flock lock dies
1821 Details : ldlm pool logic depends on number of granted locks equal to
1822 number of released locks which is not true for flock locks, so
1823 just exclude such locks from consideration.
1827 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1828 Details : Use fsfilt_sync() for both the whole filesystem sync and
1829 individual file sync to eliminate dangerous inode locking
1830 with I_LOCK that can lead to a deadlock.
1834 Description: performance in 1.6.3
1835 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1836 sure that 1MB requests are not fragmented by the block layer.
1838 Severity : enhancement
1840 Description: organize the server-side client stats on per-nid basis
1841 Details : Change the structure of stats under obdfilter and mds to
1849 The "uuid"s file would list the uuids of _active_ exports.
1850 And the clear entry is to clear all stats and stale nids.
1852 Severity : enhancement
1854 Description: eliminate client locks in face of contention
1855 Details : file contention detection and lockless i/o implementation
1856 for contended files.
1860 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1863 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1865 Description: MDS may lose file striping (and hence file data) in some cases
1866 Details : If there are additional extended attributes stored on the MDS,
1867 in particular ACLs, SELinux, or user attributes (if user_xattr
1868 is specified for the client mount options) then there is a risk
1869 of attribute loss. Additionally, the Lustre file striping
1870 needs to be larger than default (e.g. striped over all OSTs),
1871 and an additional attribute must be stored initially in the
1872 inode and then increase in size enough to be moved to the
1873 external attribute block (e.g. ACL growing in size) for file
1878 Description: skiplist implementation simplification
1879 Details : skiplists are used to group compatible locks on granted list
1880 that was implemented as tracking first and last lock of each
1881 lock group the patch changes that to using doubly linked lists
1885 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1886 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1887 because LASSERT on both the data supplied by a client, and the
1888 data on disk is dangerous and incorrect.
1890 Severity : enhancement
1892 Description: Slow truncate/writes to huge files at high offsets.
1893 Details : Directly associate cached pages to lock that protect those pages,
1894 this allows us to quickly find what pages to write and remove
1895 once lock callback is received.
1899 Description: more ldlm soft lockups
1900 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1901 starve other threads from the resource lock for a long time in
1902 case of long waiting queue, so change the debug level from
1903 D_OTHER to the less frequently used D_INFO.
1905 Severity : enhancement
1907 Description: add -gid, -group, -uid, -user options to lfs find
1911 Description: Hung threads in invalidate_inode_pages2_range
1912 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1913 once one is completed. As a result, some RPCs are stuck in the
1914 queue and are never sent.
1918 Description: filter threads hungs on waiting journal commit
1919 Details : Cleanup filter group llog code, then only filter group llog will
1920 be only created in the MDS/OST syncing process.
1924 Description: Procfs and llog threads access destoryed import sometimes.
1925 Details : Sync the import destoryed process with procfs and llog threads by
1926 the import refcount and semaphore.
1928 Severity : enhancement
1930 Description: openlock cache of b1_6 port to HEAD
1935 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1936 Details : If insertion of an extent fails, then discard the inode
1937 preallocation and free data blocks else it can lead to duplicate
1942 Description: don't always update ctime in ext3_xattr_set_handle()
1943 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1944 In some cases the ctime should not be updated, for example for
1945 2.0->1.8 compatibility it is necessary to delete an xattr and it
1946 should not update the ctime.
1950 Bugzilla : 15713/16362
1951 Description: Assertion in iopen_connect_dentry in 1.6.3
1952 Details : looking up an inode via iopen with the wrong generation number can
1953 populate the dcache with a disconneced dentry while the inode
1954 number is in the process of being reallocated. This causes an
1955 assertion failure in iopen since the inode's dentry list contains
1956 both a connected and disconnected dentry.
1960 Description: assertion failure in ldlm_handle2lock()
1961 Details : fix a race between class_handle_unhash() and class_handle2object()
1962 introduced in lustre 1.6.5 by bug 13622.
1967 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1968 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1973 Description: oops in sd_iostats_seq_show()
1974 Details : unloading/reloading the scsi low level driver triggers a kernel
1975 bug when trying to access the sd iostat file.
1980 Description: Kernel panics during QLogic driver reload
1981 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1982 patch, causing memory corruption.
1987 Description: journal_dev option does not work in b1_6
1988 Details : pass mount option during pre-mount.
1990 Severity : enhancement
1992 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1993 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1994 extent information of a file. It can be used to map logical blocks
1995 in a file to physical blocks in the block device.
1999 Description: LDLM soft lockups - improvement
2000 Details : It is be possible to send the lock handle along with each read
2001 or write request because the client is already doing a lock match
2002 itself so there isn't any reason the OST should have to re-do that
2006 Frequency : only on Cray X2
2008 Description: X2 build failures
2009 Details : fix build failures on Cray X2.
2013 Description: xid & resent requests
2014 Details : Initialize RPC XID from clock at startup (randomly if clock is
2017 Severity : enhancement
2019 Description: Add lustre_start utility to start or stop multiple Lustre servers
2024 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2025 Details : In case of memory pressure, list_del() can be called twice on
2026 req->rq_history_list, causing a kernel oops.
2030 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2031 Details : fix stack overflow in the distributed lock manager by defering
2032 export eviction after a failed AST to the elt thread instead of
2033 handling it in the dlm interpret routine.
2037 Description: Convert some comments to new format.
2038 Details : Update documenting comments to match doxygen conventions.
2042 Description: Grammar fixes.
2043 Details : A couple of trivial spelling fixes.
2047 Description: OSD_COUNTERS-mandatory
2048 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2049 to simplify the code.
2053 Description: simplify cmm_device freeing logic.
2054 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2058 Description: Add lockdep support to dt_object_operations locking interface.
2059 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2060 parameter indicating lock ordering. Update mdd code to use new
2065 Description: Introduce failloc constants for lockless IO tests.
2066 Details : Add two new failloc constants to test lockless IO. Only one of
2067 them in implemented---another is checked in yet to be landed
2072 Description: Add lockdep support for inode mutex.
2073 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2074 in the situations where Lustre has to lock more than one inode
2079 Description: Add optional invariants checking support.
2080 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2081 configure switch. This macro is to be used for consistency and
2082 sanity checks that are too expensive to be left in `production'
2087 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2088 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2093 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2094 Details : Use spin_lock_nested() in (the only) situation where more than
2095 one ldlm_lock is locked simultaneously. Also, fix possible
2096 dead-lock in ldlm_lock_change_resource() by enforcing particular
2101 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2102 Details : Instead of specifying each ldlm_lock call-back through separate
2103 parameter, wrap them into struct ldlm_callback_suite.
2107 Description: Kill join_lru obd method and its callers.
2108 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2109 in memory---a requirement implemented through ->o_join_lru() obd
2110 method in HEAD. Remove this method and its users.
2114 Description: Add asynchronous ldlm ENQUEUE completion handler.
2115 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2116 case that stock ldlm_completion_ast() cannot handle as it waits
2117 until lock is granted. Introduce new ldlm_completion_ast_async()
2118 for this. Also comment ldlm_completion_ast().
2122 Description: ldlm_error <-> errno conversion.
2123 Details : Add functions to map (rather arbitrary) between LDLM error codes
2124 and standard errno values. CLIO needs this to prevent LDLM specific
2125 constants from escaping ldlm and osc.
2129 Description: Kill unused ldlm_handle2lock_ns() function.
2130 Details : Kill unused ldlm_handle2lock_ns() function.
2134 Description: Add lu_ref support to ldlm_lock
2135 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2136 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2137 and ldlm_resource. LDLM interface has to be changed, because code that
2138 releases a reference on a lock, has to "know" what reference this is.
2139 In the most frequent case
2141 lock = ldlm_handle2lock(handle);
2143 LDLM_LOCK_PUT(lock);
2145 no changes are required. When any other reference (received _not_ from
2146 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2147 instead of LDLM_LOCK_PUT().
2149 Arguably, changes are pervasive, and interface requires some discipline
2150 for proper use. On the other hand, it was very instrumental in finding
2151 a few leaked lock references.
2155 Description: Add ldlm_lock_addref_try().
2156 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2157 attempts to addref a lock that might be being canceled
2162 Description: Add ldlm_weigh_callback().
2163 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2164 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2165 hard-coded `number of pages' logic.
2169 Description: Add lockdep annotations to llog code.
2170 Details : Use appropriately tagged _nested() locking calls in the places
2171 where llog takes more than one ->lgh_lock lock.
2175 Description: Add loi_kms_set().
2176 Details : Wrap kms updates into a helper function.
2180 Description: Constify instances of struct lsm_operations.
2181 Details : Constify instances of struct lsm_operations.
2185 Description: lu_conf support.
2186 Details : On a server, a file system object is uniquely identified
2187 by a fid, which is sufficient to locate and load all object
2188 state (inode). On a client, on the other hand, more data are
2189 necessary instantiate an object. Change lu_object_find() and
2190 friends to take additional `lu_conf' argument describing object.
2191 Typically this includes layout information.
2195 Description: lu_context fixes.
2196 Details : Introduce new lu_context functions that are needed on the client
2197 side, where some system threads (ptlrpcd) are shared by multiple
2198 modules, and so cannot be stopped during module shutdown.
2202 Description: Add start and stop methods to lu_device_type_operations.
2203 Details : Introduce two new methods in lu_device_type_operations, that are
2204 invoked when first instance of a given type is created and last one
2205 is destroyed respectively. This is need by CLIO.
2209 Description: Add lu_ref support to struct lu_device.
2210 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2211 track leaked references.
2215 Description: Introduce lu_kmem_descr.
2216 Details : lu_kmem_descr and its companion interface allow to create
2217 and destroy a number of kmem caches at once.
2221 Description: Fix lu_object finalization race.
2222 Details : Fix a race between lu_object_find() finding an object and its
2223 concurrent finalization. This race is (most likely) not possible
2224 on the server, but might happen on the client.
2228 Description: Introduce lu_ref interface.
2229 Details : lu_ref is a debugging module allowing to track references to
2230 a given object. It is quite cpu expensive, and has to be
2231 explicitly enabled with --enable-lu_ref. See usage description
2236 Description: Factor lu_site procfs stats into a separate function.
2237 Details : Separate lu_site stats printing code into a separate function
2238 that can be reused on a client.
2242 Description: Constify instances of struct {lu,dt,md}_device_operations.
2243 Details : Constify instances of struct {lu,dt,md}_device_operations.
2247 Description: Introduce struct md_site and move meta-data specific parts of
2248 struct lu_site here.
2249 Details : Move md-specific fields out of struct lu_site into special struct
2250 md_site, so that lu_site can be used on a client.
2254 Description: Kill mdd_lov_destroy().
2255 Details : Remove unused mdd code.
2259 Description: Add st_block checking to multistat.c.
2260 Details : Add st_block checking to multistat.c.
2264 Description: Add lu_ref support to struct obd_device.
2265 Details : Add lu_ref tracking to obd_device.
2269 Description: Kill obd_set_fail_loc().
2270 Details : Remove unused code.
2274 Description: Add special type for ptlrpc_request interpret functions.
2275 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2276 there. Actual usage will be in CLIO.
2280 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2281 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2282 doesn't work with lockdep.
2286 Description: Add rwv.c test program.
2287 Details : New testing program exercising readv(2) and writev(2) (Qian).
2291 Description: Add sendfile.c test program.
2292 Details : New testing program exercising sendfile(2) (Jay).
2296 Description: Ratelimit a message that can be very frequent.
2297 Details : Ratelimit a memory allocation failure message that can
2302 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2303 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2307 Description: Liblustre build fixes.
2308 Details : Liblustre build fixes.
2312 Description: libcfs: add cfs_{need,cond}_resched() interface.
2313 Details : libcfs: add cfs_{need,cond}_resched() definition and
2314 implementations for Linux, NT, and liblustre.
2316 Severity : enhancement
2318 Description: More exported tunables for mballoc
2319 Details : Add support for tunable preallocation window and new tunables for
2320 large/small requests
2324 Description: Detect corruption of block bitmap and checking for preallocations
2325 Details : Checks validity of on-disk block bitmap. Also it does better
2326 checking of number of applied preallocations. When corruption is
2327 found, it turns filesystem readonly to prevent further corruptions.
2331 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2332 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2333 window to 0. If it does hit cache miss, and read-pattern is still
2334 stride-io mode, does not reset the stride window, but also does not
2335 increase the stride window length in this case.
2339 Frequency : only for big-endian servers
2340 Description: Check if system is big-endian while mounting fs with extents feature
2341 Details : Mounting a filesystem with extents feature will fail on big-endian
2342 systems since ext3-based ldiskfs is not supported on big-endian
2343 systems. This can be over-riden with "bigendian_extents" mount option.
2345 Severity : enhancement
2347 Description: The root squash functionality
2348 Details : A security feature, which is to prevent users from being able
2349 to mount lustre on their desktop, run as root, and delete
2350 all of the files in the filesystem. The goal is accomplished by
2351 remapping user id (UID) and group id (GID) of the root user to
2352 a UID and GID specified by the system administartor via Lustre
2353 configuration management server (MGS). The functionality also
2354 allows to specify sets of clients for which the remapping does
2359 Description: Excessive recovery window
2360 Details : With AT enabled, the recovery window can be excessively long (6000+
2361 seconds). To address this problem, we no longer use
2362 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2363 timeout no longer depends on the service time, it is set to
2364 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2365 time via pb_service_time.
2369 Description: Watchdog triggered on MDS failover
2370 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2371 the OSTs know that the MDS "UUID" can be reused for the same export
2372 from a different NID, so we do not need to wait for the export to be
2376 Frequency : rare, only if using MMP with Linux RAID
2378 Description: MMP doesn't work with Linux RAID
2379 Details : While using HA for Lustre servers with Linux RAID, it is possible
2380 that MMP will not detect multiple mounts. To make this work we
2381 need to unplug the device queue in RAID when the MMP block is being
2382 written. Also while reading the MMP block, we should read it from
2383 disk and not the cached one.
2385 Severity : enhancement
2387 Description: open file using fid
2388 Details : A file can be opened using just its fid, like
2389 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2392 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2395 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2396 Details : add proper locking for creation of jbd_4k slab cache
2400 Description: MMP check in ext3_remount() fails without displaying any error
2401 Details : When multiple mount protection fails during remount, proper error
2404 Severity : enhancement
2406 Description: Allow stripe size to be up to 4G-64k
2407 Details : Fix math logic to allow large stripe sizes.
2411 Description: add check for >8TB ldiskfs filesystems
2412 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2413 Don't allow >8TB ldiskfs filesystems to be mounted without
2414 force_over_8tb mount option
2419 Descriptoin: OSS complains "too many blocks" and crashes
2420 Details : In filter_commitrw_write():
2421 1. lquota_chkquota();
2422 2. obdo_from_inode();
2423 3. lquota_pending_commit();
2424 "2." has changed uid and gid in this case, then caused
2425 lquota_chkquota() and lquota_pending_commit() to use
2426 different uid and gid. That is the root of the bug.
2430 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2431 Details : Detect and print out the 'stripe count = -1' case properly.
2435 Description: Out or order replies might be lost on replay
2436 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2437 request with tid smaller then one currently being added, add it
2438 to the start, not end of the list.
2440 --------------------------------------------------------------------------------
2442 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2444 * Support for kernels:
2445 2.6.5-7.283 (SLES 9),
2446 2.6.9-55.EL (RHEL 4),
2447 2.6.16.46-0.14 (SLES 10),
2448 2.6.18.8 vanilla (kernel.org)
2449 * Client support for unpatched kernels:
2450 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2451 2.6.16 - 2.6.22 vanilla (kernel.org)
2452 * Due to recently discovered recovery problems, we do not recommend
2453 using patchless RHEL 4 clients with this or any earlier release.
2454 * Recommended e2fsprogs version: 1.39.cfs8
2455 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2456 * Starting with this release, the ldiskfs backing filesystem required
2457 by Lustre is now in its own package, lustre-ldiskfs. This package
2458 should be installed. It is versioned separately from Lustre and
2459 may be released separately in future.
2463 Description: block reactivating mgc import until all deactivates complete
2464 Details : Fix race when failing back MDT/MGS to itself (testing)
2467 Frequency : at statup only
2469 Description: mds_lov_synchronize race leads to various problems
2470 Details : simultaneous MDT->OST connections at startup can cause the
2471 sync to abort, leaving the OSC in a bad state.
2473 Severity : enhancement
2475 Description: add optional extra BUILD_VERSION info
2476 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2477 to override the lustre version.
2480 Frequency : 2.6.18 servers only
2482 Description: ll_kern_mount() doesn't release the module reference
2483 Details : The ldiskfs module reference count never drops down to 0
2484 because ll_kern_mount() doesn't release the module reference.
2489 Description: server LBUG when using old ost_num_threads parameter
2490 Details : Accept the old ost_num_threads parameter but warn that it
2491 is deprecated, and fix an off-by-one error that caused an LBUG.
2496 Description: Transient SCSI error results in persistent IO issue
2497 Details : iobuf->dr_error is not reinitialized to 0 between two
2501 Frequency : sometimes when underlying device returns I/O errors
2503 Description: OSTs not going read-only during write failures
2504 Details : OSTs are not remounted read-only when the journal commit threads
2505 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2506 instead of the ext3 wrappers.
2510 Description: poor connect scaling with increasing client count
2511 Details : Don't run filter_grant_sanity_check for more than 100 exports
2512 to improve scaling for large numbers of clients.
2515 Frequency : SLES10 only
2517 Description: sanity-quota.sh quotacheck failed: rc = -22
2518 Details : Quotas cannot be enabled on SLES10.
2521 Frequency : liblustre clients only
2523 Description: getdirentries does not give error when run on compute nodes
2524 Details : getdirentries does not fail when the size specified as an argument
2525 is too small to contain at least one entry
2527 Severity : enhancement
2529 Description: Add LNET router traceability for debug purposes
2530 Details : If a checksum failure occurs with a router as part of the
2531 IO path, the NID of the last router that forwarded the bulk data
2532 is printed so it can be identified.
2537 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2538 Details : Due to a race condition, liblustre clients were occasionally
2539 evicted incorrectly.
2541 Severity : enhancement
2543 Description: lfs setstripe use optional parameters instead of postional
2546 Severity : enhancement
2548 Description: Nanosecond timestamp support for ldiskfs
2549 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2550 resolution timestamps. There is not yet support for this at
2551 the Lustre filesystem level.
2554 Frequency : during server recovery
2556 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2557 Details : request with rq_no_resend flag not awake l_wait_event if they get
2561 Frequency : nfs export on patchless client
2563 Description: connectathon hang when test nfs export over patchless client
2564 Details : Disconnected dentry cannot be found with lookup, so we do not need
2565 to unhash it or make it invalid
2568 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2569 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2570 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2575 Description: open req refcounting wrong on reconnect
2576 Details : If reconnect happened between getting open reply from server and
2577 call to mdc_set_replay_data in ll_file_open, we will schedule
2578 replay for unreferenced request that we are about to free.
2579 Subsequent close will crash in variety of ways.
2580 Check that request is still eligible for replay in
2581 mdc_set_replay_data().
2586 Description: disable writes to filesystem when reading health_check file
2587 Details : the default for reading the health_check proc file has changed
2588 to NOT do a journal transaction and write to disk, because this
2589 can cause reads of the /proc file to hang and block HA state
2590 checking on a healthy but otherwise heavily loaded system. It
2591 is possible to return to the previous behaviour during configure
2592 with --enable-health-write.
2594 Severity : enhancement
2596 Description: 64-bit inode version
2597 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2598 to the inode. This will be required for version-based recovery.
2603 Description: MDS fails to start if a duplicate client export is detected
2604 Details : in some rare cases it was possible for a client to connect to
2605 an MDS multiple times. Upon recovery the MDS would detect this
2606 and fail during startup. Handle this more gracefully.
2608 Severity : enhancement
2610 Description: Add -o localflock option to simulate old noflock
2612 Details : This will achieve local-only flock/fcntl locks
2618 Description: log_commit_thread vs filter_destroy race leads to crash
2619 Details : Take import reference before releasing llog record semaphore
2624 Description: Wrong request locking in request set processing
2625 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2626 imp_delayed_list, in this place should be used imp_lock.
2629 Frequency : when reconnection
2631 Description: Grant Leak when osc reconnect to OST
2632 Details : When osc reconnect ost, OST(filter) should check whether it
2633 should grant more space to client by comparing fed_grant and
2634 cl_avail_grant, and return the granted space to client instead
2635 of "new granted" space, because client will call osc_init_grant
2636 to update the client grant space info.
2639 Frequency : when client reconnect to OST
2641 Description: Grant Leak when osc do resend and replay bulk write
2642 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2643 bulk write request, because the grant info while be sync between
2644 OSC and OST when reconnect, and we should ignore the grant info
2645 these of resend/replay write req.
2650 Description: Grant space more than avaiable left space sometimes.
2651 Details : When then OST is about to be full, if two bulk writing from
2652 different clients came to OST. Accord the avaliable space of the
2653 OST, the first req should be permitted, and the second one
2654 should be denied by ENOSPC. But if the seconde arrived before
2655 the first one is commited. The OST might wrongly permit second
2656 writing, which will cause grant space > avaiable space.
2659 Frequency : when client is evicted
2661 Description: Grant might be wrongly erased when osc is evicted by OST
2662 Details : when the import is evicted by server, it will fork another
2663 thread ptlrpc_invalidate_import_thread to invalidate the
2664 import, where the grant will be set to 0. While the original
2665 thread will update the grant it got when connecting. So if
2666 the former happened latter, the grant will be wrongly errased
2667 because of this race.
2672 Description: Checking Stale with correct fid
2673 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2674 to check whether it is stale, because sometimes, we want the
2675 enqueue happened anyway, and op_data.fid2 will not be initialized.
2677 Severity : enhancement
2679 Description: update patchless client
2680 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2683 Frequency : only with 2.4 kernel
2685 Description: random memory corruption
2686 Details : size of struct ll_inode_info is to big for union inode.u and this
2687 can be cause of random memory corruption.
2692 Description: Memory leak in recovery
2693 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2694 It should also check obd_fail before fsfilt_start, otherwise if
2695 fsfilt_start return -EROFS,(failover mds during mds recovery).
2696 then the req will return with repmsg->transno = 0 and rc = EROFS.
2697 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2698 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2699 be freed no matter whether fsfilt_commit success or not.
2702 Frequency : only with huge count clients
2704 Description: Prevents from taking the superblock lock in llap_from_page for
2706 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2707 for llap_from_page call in ll_removepage prevents from taking the
2708 superblock lock for a soon died page.
2713 Description: Not check open intent error before release open handle
2714 Details : in some rare cases, the open intent error is not checked before
2715 release open handle, which may cause
2716 ASSERTION(open_req->rq_transno != 0), because it tries to release
2717 the failed open handle.
2722 Description: Set cat log bitmap only after create log success.
2723 Details : in some rare cases, the cat log bitmap is set too early. and it
2724 should be set only after create log success.
2728 Description: Accessing a block bevice can re-enable I/O when Lustre is
2729 tearing down a device.
2730 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2735 Description: service threads may hog cpus when there are a lot of requests
2736 Details : Insert cond_resched to give other threads a chance to use some CPU
2741 Description: the cat log was not initialized in recovery
2742 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2743 unlink log on mds will not be initialized until mds post
2744 recovery. And also in mds post recovery, the unlink log will
2745 initialization will be done asynchronausly, so there will be race
2746 between add unlink log and unlink log initialization.
2750 Description: brw_stats were being printed incorrectly
2751 Details : brw_stats were being printed as log2 but all of them were not
2752 recorded as log2. Also remove some code duplication arising from
2753 filter_tally_{read,write}.
2757 Frequency : rare, only in recovery.
2758 Description: ASSERTION(req->rq_type != LI_POISON) failed
2759 Details : imp_lock should be held while iterating over imp_sending_list for
2760 prevent destroy request after get timeout in ptlrpc_queue_wait.
2764 Description: replay-single.sh test 52 fails
2765 Details : A lock's skiplist need to be cleanup when it being unlinked
2766 from its resource list.
2770 Description: Short directio read returns full requested size rather than
2772 Details : Direct I/O operations should return actual amount of bytes
2773 transferred rather than requested size.
2775 Severity : enhancement
2777 Description: metadata RPC reduction (e.g. for rm performance)
2778 Details : decrease the amount of synchronous RPC between clients and servers
2779 by canceling conflicing lock before the operation on the client
2780 and packing thier handles into the main operation RPC to server.
2782 Severity : enhancement
2784 Description: Async OSC create to avoid the blocking unnecessarily.
2785 Details : If a OST has no remain object, system will block on the creating
2786 when need to create a new object on this OST. Now, ways use
2787 pre-created objects when available, instead of blocking on an
2788 empty osc while others are not empty. If we must block, we block
2789 for the shortest possible period of time.
2793 Description: improve handling recoverable errors
2794 Details : if request processig with error which can be recoverable on server
2795 request should be resend, otherwise page released from cache and
2798 Severity : enhancement
2800 Description: refine locking for avoid write wrong info into lov_objid file
2801 Details : fix possible races with add new target and write/update data in
2804 --------------------------------------------------------------------------------
2806 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2811 Frequency : on some architectures
2813 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2814 Details : Uninitialized flags sometimes cause configuration commands to
2818 Frequency : patchless clients only
2820 Description: missing __iget() symbol export
2821 Details : The __iget() symbol export is missing. To avoid the need for
2822 this on patchless clients the deathrow inode reaper is turned
2823 off, and we depend on the VM to clean up old inodes. This
2824 dependency was during via the fix for bug 12181.
2826 --------------------------------------------------------------------------------
2828 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2830 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2831 INTEROPERATE with older versions automatically. In many cases a
2832 special upgrade step is needed. Please read the
2833 user documentation before upgrading any part of a live system.
2834 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2835 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2836 * WARNING: Lustre configuration and startup changes are required with
2837 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2841 Severity : enhancement
2843 Description: MountConf
2844 Details : Lustre configuration is now managed via mkfs and mount
2845 commands instead of lmc and lconf. New obd types (MGS, MGC)
2846 are added for dynamic configuration management. See
2847 https://mail.clusterfs.com/wikis/lustre/MountConf for
2850 Severity : enhancement
2852 Description: dynamic OST addition
2853 Details : OSTs can now be added to a live filesystem
2855 Severity : enhancement
2857 Description: startup order invariance
2858 Details : MDTs and OSTs can be started in any order. Clients only
2859 require the MDT to complete startup.
2861 Severity : enhancement
2863 Description: parallel, asynchronous orphan cleanup
2864 Details : orphan cleanup is now performed in separate threads for each
2865 OST, allowing parallel non-blocking operation.
2867 Severity : enhancement
2869 Description: optimized stripe assignment
2870 Details : stripe assignments are now made based on ost space available,
2871 ost previous usage, and OSS previous usage, in order to try
2872 to optimize storage space and networking resources.
2874 Severity : enhancement
2876 Description: Permanently set tunables
2877 Details : All writable /proc/fs/lustre tunables can now be permanently
2878 set on a per-server basis, at mkfs time or on a live system.
2880 Severity : enhancement
2882 Description: Lustre message v2
2883 Details : Add lustre message format v2.
2885 Severity : enhancement
2887 Description: client OST exclusion list
2888 Details : Clients can be started with a list of OSTs that should be
2889 declared "inactive" for known non-responsive OSTs.
2893 Description: SPEC SFS validation failure on NFS v2 over lustre.
2894 Details : Changes the blocksize for regular files to be 2x RPC size,
2895 and not depend on stripe size.
2897 Severity : enhancement
2899 Description: Multiple MD RPCs in flight.
2900 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2901 To avoid overly-overloading MDS, introduce a limit on number of
2902 MDS RPCs in flight for a single client and add /proc controls
2903 to adjust this limit.
2905 Severity : enhancement
2907 Description: client read/write statistics
2908 Details : Add client read/write call usage stats for performance
2909 analysis of user processes.
2910 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2911 file access. extents_stats shows chunk size distribution.
2912 extents_stats_per_process show chunk size distribution per
2915 Severity : enhancement
2917 Description: mds statistics
2918 Details : Add detailed mds operations statistics in
2919 /proc/fs/lustre/mds/*/stats.
2923 Description: Failure of copying files with lustre special EAs.
2924 Details : Client side always return success for setxattr call for lustre
2925 special xattr (currently only "trusted.lov").
2930 Description: Refcount LNET uuids
2931 Details : The global LNET uuid list grew linearly with every startup;
2932 refcount repeated list entries instead of always adding to
2935 Severity : enhancement
2937 Description: Dynamic service threads
2938 Details : Within a small range, start extra service threads
2939 automatically when the request queue builds up.
2942 Frequency : mixed-endian client/server environments
2944 Description: mixed-endian crashes
2945 Details : The new msg_v2 system had some failures in mixed-endian
2948 Severity : enhancement
2950 Description: Easy OST removal
2951 Details : OSTs can be permanently deactivated with e.g. 'lctl
2952 conf_param lustre-OST0001.osc.active=0'
2954 Severity : enhancement
2956 Description: MGS proc entries
2957 Details : Added basic proc entries for the MGS showing what filesystems
2960 Severity : enhancement
2962 Description: provide MGS failover
2963 Details : Added config lock reacquisition after MGS server failover.
2965 Severity : enhancement
2967 Description: add Linux 2.4 support
2968 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2972 Description: a large application tries to do I/O to the same resource and dies
2973 in the middle of it.
2974 Details : Check the req->rq_arrival time after the call to
2975 ost_brw_lock_get(), but before we do anything about
2976 processing it & sending the BULK transfer request. This
2977 should help move old stale pending locks off the queue as
2978 quickly as obd_timeout.
2981 Frequency : when an incorrect nid is specified during startup
2983 Description: ptlrpc connect to non-existant node causes kernel crash
2984 Details : LNET can't be re-entered from an event callback, which
2985 happened when we expire a message after the export has been
2986 cleaned up. Instead, hand the zombie cleanup off to another
2989 Severity : enhancement
2991 Description: plain/inodebits lock performance improvement
2992 Details : Grouping plain/inodebits in granted list by their request modes
2993 and bits policy, thus improving the performance of search through
2997 Frequency : only if OST filesystem is corrupted
2999 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3000 Details : for a short time RPCs with bulk IO are in the replay list,
3001 but replay of bulk IOs is unimplemented. If the OST filesystem
3002 is corrupted due to disk cache incoherency and then replay is
3003 started it is possible to trip an assertion. Avoid putting
3004 committed RPCs into the replay list at all to avoid this issue.
3007 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3009 Description: System hang on startup
3010 Details : This bug allowed the liblustre (e.g. catamount) client to
3011 return to the app before handling all startup RPCs. This
3012 could leave the node unresponsive to lustre network traffic
3013 and manifested as a server ptllnd timeout.
3015 Severity : enhancement
3017 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3018 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3019 boolean to control whether to print peer debug info when a
3020 client's RPC times out.
3023 Frequency : only for kernels with patches from Lustre below 1.4.3
3025 Description: Remove old rdonly API
3026 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3029 Frequency : only for devices with external journals
3031 Description: Set external device read-only also
3032 Details : During a commanded failover stop, we set the disk device
3033 read-only while the server shuts down. We now also set any
3034 external journal device read-only at the same time.
3037 Frequency : when upgrading from 1.4 while trying to change parameters
3039 Description: The wrong (new) MDC name was used when setting parameters for
3040 upgraded MDT's. Also allows changing of OSC (and MDC)
3041 parameters if --writeconf is specified at tunefs upgrade time.
3044 Frequency : when setting specific ost indicies
3046 Description: QOS code breaks on skipped indicies
3047 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3048 created with --index need not be sequential.
3050 Severity : enhancement
3052 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3053 Details : The uninit_groups feature works in conjunction with the kernel
3054 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3055 up the pass1 processing of e2fsck. This is a read-only feature
3056 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3057 mount filesystems that have had this feature enabled.
3059 Severity : enhancement
3061 Description: Improve multi-block allocation algorithm to avoid fragmentation
3062 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3063 allocation locality and avoid filesystem fragmentation.
3065 ------------------------------------------------------------------------------
3067 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3069 * Support for kernels:
3070 2.4.21-47.0.1.EL (RHEL 3)
3071 2.6.5-7.283 (SLES 9)
3072 2.6.9-42.0.10.EL (RHEL 4)
3073 2.6.12.6 vanilla (kernel.org)
3074 2.6.16.27-0.9 (SLES 10)
3075 * Recommended e2fsprogs version: 1.39.cfs5
3077 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3081 Frequency : occasional, depends on client load and configuration
3082 Bugzilla : 12181, 12203
3083 Description: data loss for recently-modified files
3085 Details : In some cases it is possible that recently written or created
3086 files may not be written to disk in a timely manner (this should
3087 normally be within 30s unless client IO load is very high).
3088 The problem appears as zero-length files or files that are a
3089 multiple of 1MB in size after a client crash or client eviction
3090 that are missing data at the end of the file.
3092 This problem is more likely to be hit on clients where files are
3093 repeatedly created and unlinked in the same directory, clients
3094 have a large amount of RAM, have many CPUs, the filesystem has
3095 many OSTs, the clients are rebooted frequently, and/or the files
3096 are not accessed by other nodes after being written.
3098 The presence of the problem can be detected by looking at
3099 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3100 smaller than the second (nr_unused) then dirty files will not
3101 be flushed automatically to disk. "sync; sleep 10" should be
3102 run several times on the node before unmounting it to update
3103 Lustre (this is also safe to run on nodes without this problem).
3105 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3106 that can cause this same problem, so customers using that kernel
3107 also need to update the kernel in addition to Lustre. In order
3108 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3110 It is normal that files written just before a client crash (less
3111 than 30s) may not yet have been flushed to disk, even for local
3115 Frequency : frequent on thin XT3 nodes
3117 Description: UUID collision on thin XT3 Linux nodes
3118 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3119 randomly, since we relied on an insufficiently-seeded PRNG.
3124 Description: OSS hangs after "All ost request buffers busy"
3125 Details : A deadlock between quota and journal operations caused OSS
3126 hangs after printing "All ost request buffers busy."
3129 Frequency : always on liblustre builds
3131 Description: Cleanup compiler warnings on liblustre
3134 Frequency : always on liblustre builds on XT3
3136 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3140 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3141 > 32000 subdirectories
3142 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3143 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3144 subdirectory count crosses 32000. This will aid e2fsck to
3145 correctly handle more than 32000 subdirectories.
3148 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3150 Description: System hang on startup
3151 Details : This bug allowed the liblustre (e.g. catamount) client to
3152 return to the app before handling all startup RPCs. This
3153 could leave the node unresponsive to lustre network traffic
3154 and manifested as a server ptllnd timeout.
3156 Severity : enhancement
3158 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3159 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3160 boolean to control whether to print peer debug info when a
3161 client's RPC times out.
3166 Description: make O_SYNC working on 2.6 kernels
3167 Details : 2.6 kernels use different method for mark pages for write,
3168 so need add a code to lustre for O_SYNC work.
3173 Description: Failure to close file and release space on NFS
3174 Details : Put inode details into lock acquired in ll_intent_file_open.
3175 Use mdc_intent_lock in ll_intent_open to properly
3176 detect all kind of errors unhandled by mdc_enqueue.
3181 Description: proc file read during shutdown sometimes raced obd removal,
3183 Details : Add lock to prevent obd access after proc file removal.
3186 Frequency : Only for files larger than 4GB on 32-bit clients.
3188 Description: improperly doing page alignment of locks
3189 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3190 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3195 Description: RPCs being resent when they shouldn't be
3196 Details : Some RPCs that should not be resent are being resent. This
3197 can cause inconsistencies in the RPC state machine. Do not
3198 resend such requests.
3201 Frequency : rare, only with NFS export
3203 Description: Crash on NFS re-export node
3204 Details : under very unusual load conditions an assertion is hit in
3205 ll_intent_file_open()
3208 Frequency : only if OST filesystem is corrupted
3210 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3211 Details : for a short time RPCs with bulk IO are in the replay list,
3212 but replay of bulk IOs is unimplemented. If the OST filesystem
3213 is corrupted due to disk cache incoherency and then replay is
3214 started it is possible to trip an assertion. Avoid putting
3215 committed RPCs into the replay list at all to avoid this issue.
3220 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3221 Details : Large single O_DIRECT read and write calls can fail to allocate
3222 a sufficiently large buffer to process the request. In case of
3223 allocation failure the allocation is retried with a smaller
3224 buffer and broken into smaller requests.
3226 Severity : enhancement
3228 Description: Add -o localflock option to simulate old noflock behaviour.
3229 Details : This will achieve local-only flock/fcntl locks coherentness.
3234 Description: versioning check is incomplete
3235 Details : Checking the version difference of client vs. server, report
3236 error if the gap is too big.
3241 Description: add support PG_writeback bit
3242 Details : add support for PG_writeback bit for Lustre, for more carefull
3243 work with page cache in 2.6 kernel. This also fix some deadlocks
3244 and remove hack for work O_SYNC with 2.6 kernel.
3246 Severity : enhancement
3248 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3249 Details : The uninit_groups feature works in conjunction with the kernel
3250 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3251 up the pass1 processing of e2fsck. This is a read-only feature
3252 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3253 mount filesystems that have had this feature enabled.
3255 Severity : enhancement
3257 Description: Improve multi-block allocation algorithm to avoid fragmentation
3258 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3259 allocation locality and avoid filesystem fragmentation.
3261 ------------------------------------------------------------------------------
3263 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3265 * Support for kernels:
3266 2.6.9-42.0.3.EL (RHEL 4)
3267 2.6.5-7.276 (SLES 9)
3268 2.4.21-47.0.1.EL (RHEL 3)
3269 2.6.12.6 vanilla (kernel.org)
3270 2.6.16.21-0.8 (SLES10)
3271 * Recommended e2fsprogs version: 1.39.cfs2-0
3273 * The backwards-compatible /proc/sys/portals symlink has been removed
3274 in this release. Before upgrading, please ensure that you change
3275 any configuration scripts or /etc/sysctl.conf files that access
3276 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3277 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3278 in advance of the upgrade on any system running Lustre 1.4.6 or
3279 newer, since /proc/sys/lnet was added in that version.
3280 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3284 Frequency : only when quota is used
3286 Description: avoid scanning export list for quota master
3287 Details : Change the algorithms to avoid scanning export list in order
3288 to improve the efficiency.
3291 Frequency : MDS failover only, very rarely
3293 Description: "went back in time" messages on mds failover
3294 Details : The greatest transno may be lost when the current operation
3295 finishes with an error (transno==0) and the client's last_rcvd
3296 record is over-written. Save the greatest transno in the
3297 mds_last_transno for this case.
3300 Frequency : always for specific kernels and striping counts
3302 Description: client may get "Matching packet too big" without ACL support
3303 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3304 when trying to access files in certain configurations. The
3305 clients should in fact be denied when mounting because they do
3306 not understand ACLs.
3309 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3311 Description: many clients connecting with IO in progress causes connect timeouts
3312 Details : Avoid synchronous journal commits to avoid delays caused by many
3313 clients connecting/disconnecting when bulk IO is in progress.
3314 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3315 OST_IO_PORTAL to avoid delays behind potentially many pending
3319 Frequency : occasionally with multiple writers to a single file
3321 Description: shared writes to file may result in wrong size reported by stat()
3322 Details : Allow growing of kms when extent lock is cancelled
3325 Frequency : always with random mmap IO to multi-striped file
3327 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3328 Details : Only if the hole is at the end of OST object so that kms is too
3329 small. Fix is to increase kms accordingly in ll_nopage.
3332 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3334 Description: writes to a missing object would leak memory on the OST
3335 Details : If there is an inconsistency between the MDS and OST filesystems,
3336 such that the MDS references an object that doesn't exist, writes
3337 to that object will leak memory due to incorrect cleanup in the
3338 error handling path, eventually running out of memory on the OST.
3343 Description: Creating too long symlink causes lustre errors
3344 Details : Check symlink and name lengths before sending requests to MDS.
3347 Frequency : only if flock is enabled (not on by default)
3349 Description: posix locks not released on fd closure on 2.6.9+
3350 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3351 kernels, this caused such locks not to be released on fd close and
3352 then assertions on fs unmount about still used locks.
3355 Frequency : MDS failover only, very rarely
3357 Description: clients may get ASSERTION(granted_lock != NULL)
3358 Details : When request was taking a long time, and a client was resending
3359 a getattr by name lock request. The were multiple lock requests
3360 with the same client lock handle and
3361 mds_getattr_name->fixup_handle_for_resent_request found one of the
3362 lock handles but later failed with ASSERTION(granted_lock != NULL).
3367 Description: handle->h_buffer_credits > 0, assertion failure
3368 Details : h_buffer_credits is zero after truncate, causing assertion
3369 failure. This patch extends the transaction or creates a new
3373 Frequency : NFS re-export or patchless client
3374 Bugzilla : 11179, 10796
3375 Description: Crash on NFS re-export node (__d_move)
3376 Details : We do not want to hash the dentry if we don't have a lock.
3377 But if this dentry is later used in d_move, we'd hit uninitialised
3378 list head d_hash, so we just do this to init d_hash field but
3379 leave dentry unhashed.
3382 Frequency : NFS re-export or patchless client
3384 Description: NFS exports has problem with symbolic link
3385 Details : lustre client didn't properly install dentry when re-exported
3386 to NFS or running patchless client.
3389 Frequency : NFS re-export or patchless client
3391 Description: Various nfs/patchless fixes.
3392 Details : fixes reuse disconected alias for lookup process - this fixes
3393 warning "find_exported_dentry: npd != pd",
3394 fix permission error with open files at nfs.
3395 fix apply umask when do revalidate.
3398 Frequency : occasional
3400 Description: Crash on NFS re-export node
3401 Details : calling clear_page() on the wrong pointer triggered oops in
3402 generic_mapping_read().
3405 Frequency : rarely, using O_DIRECT IO
3407 Description: unaligned directio crashes client with LASSERT
3408 Details : check for unaligned buffers before trying any requests.
3411 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3413 Description: stale data returned from RAID cache
3414 Details : If only a small amount of IO is done to the RAID device before
3415 reading it again it is possible to get stale data from the RAID
3416 cache instead of reading it from disk.
3419 Frequency : always for sles10 kernel
3421 Description: sles10 support
3422 Details : ll_follow_link: compile fixes and using of nd_set_link
3423 under newer kernels.
3426 Frequency : depends on arch, kernel and compiler version, always on sles10
3429 Description: recursive or deep enough symlinks cause stack overflow
3430 Details : getting rid of large stack-allocated variable in
3434 Frequency : depends on hardware
3436 Description: lustre write performance loss in the SLES10 kernel
3437 Details : the performance loss is caused by using of write barriers in the
3438 ext3 code. The SLES10 kernel turns barrier support on by
3439 default. The fix is to undo that change for ldiskfs.
3441 ------------------------------------------------------------------------------
3443 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3445 * Support for kernels:
3446 2.6.9-42.0.3EL (RHEL 4)
3447 2.6.5-7.276 (SLES 9)
3448 2.4.21-47.0.1.EL (RHEL 3)
3449 2.6.12.6 vanilla (kernel.org)
3453 Frequency : quota enabled and large files being deleted
3455 Description: releasing more than 4GB of quota at once hangs OST
3456 Details : If a user deletes more than 4GB of files on a single OST it
3457 will cause the OST to spin in an infinite loop. Release
3458 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3463 Description: statfs data retrieved from /proc may be stale or zero
3464 Details : When reading per-device statfs data from /proc, in the
3465 {kbytes,files}_{total,free,avail} files, it may appear
3466 as zero or be out of date.
3469 Frequency : systems with MD RAID1 external journal devices
3471 Description: lconf's call to blkid is confused by RAID1 journal devices
3472 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3473 of returning all block devices that match the journal UUID.
3476 Frequency : always, for aggregate stripe size over 4GB
3478 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3479 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3480 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3483 Frequency : always if "lfs find" used on a local file/directory
3485 Description: "lfs find" segfaults if used on a local file/directory
3486 Details : The case where a directory component was not specified wasn't
3487 handled correctly. Handle this properly.
3490 Frequency : always on ppc64
3492 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3493 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3494 perform endianness conversion.
3497 Frequency : rarely (truncate to non-zero file size after write under load)
3498 Bugzilla : 10730, 10687
3499 Description: Files padded with zeros to next 4K multiple
3500 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3501 it is possible that files that are truncated to a non-zero size
3502 immediately after being written are filled with zero bytes beyond
3503 the truncated size. No file data is lost.
3505 Severity : enhancement
3507 Description: Allow recovery/failover for liblustre clients.
3508 Details : liblustre clients were unaware of failover configurations until
3511 Severity : enhancement
3513 Description: user file locks should fail when not mounting with flock option
3514 Details : Set up an error-returning stub in ll_file_operations.lock field
3515 to prevent incorrect behaviour when client is mounted without
3516 flock option. Also, set up properly f_op->flock field for
3520 Frequency : always on ia64
3522 Description: "lfs df" loops on printing out MDS statfs information
3523 Details : The obd_ioctl_data was not initialized and in some systems
3524 this caused a failure during the ioctl that did not return
3525 an error. Initialize the struct and return an error on failure.
3528 Frequency : SLES 9 only
3530 Description: Error of copying files with lustre special EAs as root
3531 Details : Client side always return success for setxattr call for lustre
3532 special xattr (currently only "trusted.lov").
3535 Frequency : rarely on clusters with both ia64+i386 clients
3537 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3538 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3539 clients are doing concurrent writes to the same file it is
3540 possible that the ia64 clients may LASSERT because the OST
3541 extent locks are not PAGE_SIZE aligned. Ensure that grown
3542 locks are always aligned on the request boundary.
3545 Frequency : specific use, occasional
3547 Description: Overwriting in use executable truncates on-disk binary image
3548 Details : If one node attempts to overwrite an executable in use by
3549 another node, we now correctly return ETXTBSY instead of
3550 truncating the file.
3552 Severity : enhancement
3554 Description: Async OSC create to avoid the blocking unnecessarily.
3555 Details : If a OST has no remain object, system will block on the creating
3556 when need to create a new object on this OST. Now, ways use
3557 pre-created objects when available, instead of blocking on an
3558 empty osc while others are not empty. If we must block, we block
3559 for the shortest possible period of time.
3564 Description: chmod on Lustre root is propagated to other clients
3565 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3566 invalid by the follow_mount time.
3569 Frequency : liblustre clients only
3571 Description: Race in 'instant cancel' lock handling could lead to such locks
3572 never to be granted in case of SMP MDS
3573 Details : Do not destroy not yet granted but cbpending locks in
3577 Frequency : replay/resend of open
3579 Description: non null lock assetion failure in mds_intent_policy
3580 Details : Trying to replay/resend lockless open requests resulted in
3581 mds_open() returning 0 with no lock. Now it sets a flag if
3582 a lock is going to be returned.
3584 Severity : enhancement
3586 Description: Checksum enhancements
3587 Details : New checksum enhancements allow for resending RPCs that failed
3590 Severity : enhancement
3592 Description: Tunables on number of dirty pages in cacche
3593 Details : Allow to set limit on number of dirty pages cached.
3598 Description: client crash on unmount - lock still has references
3599 Details : In some error handling cases it was possible to leak a lock
3600 reference on a client while accessing a file. This was not
3601 harmful to the client during operation, but would cause the
3602 client to crash when the filesystem is unmounted.
3605 Frequency : specific case, rare
3607 Description: ETXTBSY on mds though file not in use by client
3608 Details : ETXTBSY is no longer incorrectly returned when attempting to
3609 chmod or chown a directory that the user previously tried to
3610 execute or a currently-executing binary.
3613 Frequency : extremely rare except on liblustre-based clients
3615 Description: Lustre space not freed when files are deleted
3616 Details : Clean up open-unlinked files after client eviction. Previously
3617 the unlink was skipped and the files remained as orphans.
3622 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3623 Details : In some cases it was possible to send a blocking callback to
3624 a client doing a glimpse, even though that client didn't get
3625 a lock granted. When the glimpse lock is cancelled on the OST
3626 the freed lock is left on the waiting list and corrupted the list.
3629 Frequency : all core dumps
3631 Description: Broke core dumps to lustre
3632 Details : Negative dentry may be unhashed if parent does not have UPDATE
3633 lock, but some callers, e.g. do_coredump, expect dentry to be
3634 hashed after successful create, hash it in ll_create_it.
3636 ------------------------------------------------------------------------------
3638 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3640 * Support for kernels:
3641 2.6.9-42.0.2.EL (RHEL 4)
3642 2.6.5-7.276 (SLES 9)
3643 2.4.21-40.EL (RHEL 3)
3644 2.6.12.6 vanilla (kernel.org)
3648 Frequency : always on RHEL 3
3650 Description: Number of open files grows over time
3651 Details : The number of open files grows over time, whether or not
3652 Lustre is started. This was due to a filp leak introduced
3653 by one of our kernel patches.
3655 ------------------------------------------------------------------------------
3657 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3659 * Support for kernels:
3660 2.6.9-42.EL (RHEL 4)
3661 2.6.5-7.276 (SLES 9)
3662 2.4.21-40.EL (RHEL 3)
3663 2.6.12.6 vanilla (kernel.org)
3668 Bugzilla : 5719, 9635, 9792, 9684
3669 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3670 Details : If a server is under heavy load and cannot reply to new
3671 connection requests before the client resends the (re)connect,
3672 the connection handling code can behave badly if two service
3673 threads are concurrently handing separate (re)connections from
3674 the same client. Add better locking to the connection handling
3675 code, and ensure that only a single connection will be processed
3676 for a given client UUID, even if the lock is dropped.
3678 Severity : enhancement
3680 Description: add TCP zero-copy support to kernel
3681 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3682 sends if the hardware supports scatter-gather and checksumming.
3683 This allows socklnd to do client-write and server-read more
3684 efficiently and reduce CPU utilization from skbuf copying.
3687 Frequency : only if NFS exporting from client
3689 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3690 Details : If setting attributes on a file created under NFS that had
3691 never been opened it would be possible to oops the client
3692 if the file had no objects.
3695 Frequency : always for liblustre
3697 Description: liblustre client does MDS+OSTs setattr RPC for each write
3698 Details : When doing a write from a liblustre client, the client
3699 incorrectly issued an RPC to the MDS and each OST the file was
3700 striped over in order to update the timestamps. When writing
3701 with small chunks and many clients this could overwhelm the MDS
3702 with RPCs. In all cases it would slow down the write because
3703 these RPCs are unnecessary.
3705 Severity : enhancement
3707 Description: allow number of MDS service threads to be changed at module load
3708 Details : It is now possible to change the number of MDS service threads
3709 running. Adding "options mds mds_num_threads={N}" to the MDS's
3710 /etc/modprobe.conf will set the number of threads for the next
3711 time Lustre is restarted (assuming the "mds" module is also
3712 reloaded at that time). The default number of threads will
3713 stay the same, 32 for most systems.
3718 Description: OST crash if filesystem is unformatted or corrupt
3719 Details : If an OST is started on a device that has never been formatted
3720 or if the filesystem is corrupt and cannot even mount then the
3721 error handling cleanup routines would dereference a NULL pointer.
3726 Description: NULL pointer deref in llap_from_page.
3727 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3728 incorrect) mapping if the page was truncated/reclaimed while it was
3729 searched for. Check for this condition and skip such pages when
3730 doing readahead. Introduce extra check to llap_from_page() to
3731 verify page->mapping->host is non-NULL (so page is not anonymous).
3734 Frequency : Sometimes when using sys_sendfile
3736 Description: "page not covered by a lock" warnings from ll_readpage
3737 Details : sendfile called ll_readpage without right page locks present.
3738 Now we introduced ll_file_sendfile that does necessary locking
3739 around call to generic_file_sendfile() much like we do in
3743 Frequency : with certain MDS communication failures at client mount time
3745 Description: NULL pointer deref after failed client mount
3746 Details : a client connection request may delayed by the network layer
3747 and not be sent until after the PTLRPC layer has timed out the
3748 request. If the client fails the mount immediately it will try
3749 to clean up before the network times out the request. Add a
3750 reference from the request import to the obd device and delay
3751 the cleanup until the network drops the request.
3754 Frequency : occasionally during client (re)connect
3756 Description: assertion failure during client (re)connect
3757 Details : processing a client connection request may be delayed by the
3758 client or server longer than the client connect timeout. This
3759 causes the client to resend the connection request. If the
3760 original connection request is replied in this interval, the
3761 client may trip an assertion failure in ptlrpc_connect_interpret()
3762 which thought it would be the only running connect process.
3765 Frequency : only with obd_echo servers and clients that are rebooted
3767 Description: kernel BUG accessing uninitialized data structure
3768 Details : When running an obd_echo server it did not start the ping_evictor
3769 thread, and when a client was evicted an uninitialized data
3770 structure was accessed. Start the ping_evictor in the RPC
3771 service startup instead of the OBD startup.
3773 Severity : enhancement
3774 Bugzilla : 10193 (patchless)
3775 Description: Remove dependency on various unexported kernel interfaces.
3776 Details : No longer need reparent_to_init, exit_mm, exit_files,
3777 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3780 Frequency : rare (only users of deprecated and unsupported LDAP config)
3782 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3783 Details : LDAP apparently contains 'lustreName' attributes instead of
3784 'name'. A simple remapping of the name is sufficient.
3787 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3789 Description: waiting_locks_callback trips kernel BUG if client is evicted
3790 Details : Running with the dump_on_timeout debug flag turned on makes
3791 it possible that the waiting_locks_callback() can try to dump
3792 the Lustre kernel debug logs from an interrupt handler. Defer
3793 this log dumping to the expired_lock_main() thread.
3795 Severity : enhancement
3797 Description: Support NFS exporting on 2.6 kernels.
3798 Details : Implement non-rawops metadata methods for NFS server to use without
3799 changing NFS server code.
3802 Frequency : very rare (synthetic metadata workload only)
3804 Description: two racing renames might cause an MDS thread to deadlock
3805 Details : Running the "racer" program may cause one MDS thread to rename
3806 a file from being the source of a rename to being the target of
3807 a rename at exactly the same time that another thread is doing
3808 so, and the second thread has already enqueued these locks after
3809 doing a lookup of the target and is trying to relock them in
3810 order. Ensure that we don't try to re-lock the same resource.
3813 Frequency : only very large systems with liblustre clients
3815 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3816 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3817 all OSTs in parallel. This allows the eviction of stale liblustre
3818 clients to proceed much faster than if they were done in series,
3819 and also offers similar improvements for other set_info RPCs.
3824 Description: excessive CPU usage during initial read phase on client
3825 Details : During the initial read phase on a client, it would agressively
3826 retry readahead on the file, consuming too much CPU and impacting
3827 performance (since 1.4.5.8). Improve the readahead algorithm
3828 to avoid this, and also improve some other common cases (read
3829 of small files in particular, where "small" is files smaller than
3830 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3835 Description: MDS crash when receiving packet with unknown intent.
3836 Details : Do not LBUG in unknown intent case, just return -EFAULT
3838 Severity : enhancement
3839 Bugzilla : 9293, 9385
3840 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3841 Details : Do not serialize getattr (non-intent version) and statfs.
3844 Frequency : occasional, when OST network is overloaded/intermittent
3846 Description: client evicted by OST after bulk IO timeout
3847 Details : If a client sends a bulk IO request (read or write) the OST
3848 may evict the client if it is unresposive to its data GET/PUT
3849 request. This is incorrect if the network is overloaded (takes
3850 too long to transfer the RPC data) or dropped the OST GET/PUT
3851 request. There is no need to evict the client at all, since
3852 the pinger and/or lock callbacks will handle this, and the
3853 client can restart the bulk request.
3856 Frequency : Always when mmapping file with no objects
3858 Description: client crashes when mmapping file with no objects
3859 Details : Check that we actually have objects in a file before doing any
3860 operations on objects in ll_vm_open, ll_vm_close and
3866 Description: Request leak when working with deleted CWD
3867 Details : Introduce advanced request refcount tracking for requests
3868 referenced from lustre intent.
3870 Severity : Enhancement
3872 Description: Cache open file handles on client.
3873 Details : MDS now will return special lock along with openhandle, if
3874 requested and client is allowed to hold openhandle, even if unused,
3875 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3876 closing files for every read/write openration.
3878 Severity : Enhancement
3880 Description: Cache open negative dentries on client when possible.
3881 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3882 negative dentries on lock revocation.
3887 Description: Remounting a client read-only wasn't possible with a zconf mount
3888 Details : It wasn't possible to remount a client read-only with llmount.
3890 Severity : enhancement
3891 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3892 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3893 MPICH is included to add Lustre-specific ADIO interfaces.
3894 This is based closely on the UFS ADIO layer and only differs
3895 in file creation, in order to allow the OST striping to be set.
3896 This is user-contributed code and not supported by CFS.
3901 Description: extended inode attributes (immutable, append-only) work improperly
3902 when 2.4 and 2.6 kernels are used on client/server or vice versa
3903 Details : Introduce kernel-independent values for these flags.
3905 Severity : enhancement
3908 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3909 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3910 granularity. Now, Fractional MB granularity is be supported,
3911 this is very useful for low memory system.
3913 Severity : enhancement
3915 Description: Getattr by fid
3916 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3917 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3918 have locks and avoids getattr rpc requests for it on every stat.
3921 Frequency : Always, for filesystems larger than 2TB
3923 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3924 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3925 filesystems larger than 2TB to be mounted with mballoc enabled.
3928 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3930 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3931 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3932 filesystems larger than 2TB is mounted then the kernel will
3933 silently corrupt the start of the filesystem. CONFIG_LBD is
3934 enabled for all CFS-supported kernels, but the possibility of
3935 this happening with a modified kernel config exists.
3937 Severity : enhancement
3939 Description: add client O_DIRECT support for 2.6 kernels
3940 Details : It is now possible to do O_DIRECT reads and writes to files
3941 in the Lustre client mountpoint on 2.6 kernel clients.
3943 Severity : enhancement
3945 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3946 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3947 parallel, not waiting for response from every OST before sending
3948 a rpc to the next OST.
3953 Description: setattr vs write race when updating file timestamps
3954 Details : Client processes that update a file timestamp into the past
3955 right after writing to the file (e.g. tar) it is possible that
3956 the updated file modification time can be reset to the current
3957 time due to a race between processing the setattr and write RPC.
3959 Severity : enhancement
3961 Description: Bring 'lfs find' closer in line with regular Linux find.
3962 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3963 -print0 options and obtains all the needed info through the lustre
3966 Severity : enhancement
3968 Description: support up to 1024 configured devices on one node
3969 Details : change obd_dev array from statically allocated to dynamically
3970 allocated structs as they are first used to reduce memory usage
3975 Description: Flush dirty partially truncated pages during truncate
3976 Details : Immediatelly flush partially truncated pages in filter_setattr,
3977 this way we completely avoid having any pages in page cache on OST
3978 and can retire ugly workarounds during writes to flush such pages.
3983 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3984 Details : protect inode from truncation within vfs_unlink() context
3985 just take a reference before calling vfs_unlink() and release it
3986 when parent's i_sem is free.
3991 Description: last_id value checked outside lock on OST caused LASSERT failure
3992 Details : If there were multiple MDS->OST object precreate requests in
3993 flight, it was possible that the OST's last object id was checked
3994 outside a lock and incorrectly tripped an assertion. Move checks
3995 inside locks, and discard old precreate requests.
3998 Frequency : always, if extents are used on OSTs
4000 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4001 Details : OSTs using the extents format would not zero the high 16 bits of
4002 the index physical block number. This is not a problem for any
4003 OST filesystems smaller than 16TB, and no kernels support ext3
4004 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4005 new/modified files) and can be fixed for existing filesystems
4006 with e2fsprogs-1.39-cfs1.
4011 Description: import connection selection may be incorrect if timer wraps
4012 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4013 import connections to be ignored if the 32-bit jiffies counter
4014 wraps. Use a 64-bit jiffies counter.
4017 Frequency : very large clusters immediately after boot
4019 Description: LNET request buffers exhausted under heavy short-term load
4020 Details : If a large number of client requests are generated on a service
4021 that has previously never seen so many requests it is possible
4022 that the request buffer growth cannot keep up with the spike in
4023 demand. Instead of dropping incoming requests, they are held in
4024 the LND until the RPC service can accept more requests.
4027 Frequency : Sometimes during replay
4029 Description: Assertion failure in ll_local_open after replay.
4030 Details : If replay happened on an open request reply before we were able
4031 to set replay handler, reply will become not swabbed tripping the
4032 assertion in ll_local_open. Now we set the handler right after
4033 recognising of open request
4036 Frequency : very rare
4038 Description: kernel reports "badness in vsnprintf"
4039 Details : Reading from the "recovery_status" /proc file in small chunks
4040 may cause a negative length in lprocfs_obd_rd_recovery_status()
4041 call to vsnprintf() (which is otherwise harmless). Exit early
4042 if there is no more space in the output buffer.
4044 Severity : enhancement
4046 Description: clear OBD RPC statistics by writing to them
4047 Details : It is now possible to clear the OBD RPC statistics by writing
4048 to the "stats" file.
4053 Description: Client mtime is not the same on different clients after utimes
4054 Details : In some cases, the client was using the utimes() syscall on
4055 a file cached on another node. The clients now validate the
4056 ctime from the MDS + OSTs to determine which one is right.
4061 Description: Inability to activate failout mode
4062 Details : lconf script incorrectly assumed that in pythong string's numeric
4063 value is used in comparisons.
4066 Frequency : always with multiple stripes per file
4068 Description: Inefficient object allocation for mutli-stripe files
4069 Details : When selecting which OSTs to stripe files over, for files with
4070 a stripe count that divides evenly into the number of OSTs,
4071 the MDS is always picking the same starting OST for each file.
4072 Return the OST selection heuristic to the original design.
4077 Description: mount failures may take full timeout to return an error
4078 Details : Under some heavy load conditions it is possible that a
4079 failed mount can wait for the full obd_timeout interval,
4080 possibly several minutes, before reporting an error.
4081 Instead return an error as soon as the status is known.
4083 Frequency : quota enabled and large files being deleted
4085 Description: releasing more than 4GB of quota at once hangs OST
4086 Details : If a user deletes more than 4GB of files on a single OST it
4087 will cause the OST to spin in an infinite loop. Release
4088 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4093 Description: statfs data retrieved from /proc may be stale or zero
4094 Details : When reading per-device statfs data from /proc, in the
4095 {kbytes,files}_{total,free,avail} files, it may appear
4096 as zero or be out of date.
4099 Frequency : systems with MD RAID1 external journal devices
4101 Description: lconf's call to blkid is confused by RAID1 journal devices
4102 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4103 of returning all block devices that match the journal UUID.
4106 Frequency : always, for aggregate stripe size over 4GB
4108 Description: assertion fails when trying to use 4GB stripe size
4109 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4110 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4113 Frequency : always on ppc64
4115 Description: the first write on an ext3 filesystem with mballoc got stuck
4116 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4117 perform endianness conversion.
4119 ------------------------------------------------------------------------------
4121 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4123 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4124 INTEROPERATE with older versions automatically. Please read the
4125 user documentation before upgrading any part of a live system.
4126 * WARNING: Lustre networking configuration changes are required with
4127 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4130 * Support for kernels:
4131 2.6.9-22.0.2.EL (RHEL 4)
4132 2.6.5-7.244 (SLES 9)
4133 2.6.12.6 vanilla (kernel.org)
4136 Severity : enhancement
4137 Bugzilla : 7981/8208
4138 Description: Introduced Lustre Networking (LNET)
4139 Details : LNET is new networking infrastructure for Lustre, it includes
4140 a reorganized network configuration mode (see the user
4141 documentation for full details) as well as support for routing
4142 between different network fabrics. Lustre Networking Devices
4143 (LNDs) for the supported network fabrics have also been
4144 created for this new infrastructure.
4146 Severity : enhancement
4147 Description: Introduced Access control lists
4148 Details : clients can set ACLs on files and directories in order to have
4149 more fine-grained permissions than the standard Unix UGO+RWX.
4150 The MDS must be started with the "-o acl" mount option.
4152 Severity : enhancement
4153 Description: Introduced filesystem quotas
4154 Details : Administrators may now establish per-user quotas on the
4157 Severity : enhancement
4159 Description: Configuration change for the XT3
4160 The PTLLND is now used to run Lustre over Portals on the XT3
4161 The configure option(s) --with-cray-portals are no longer used.
4162 Rather --with-portals=<path-to-portals-includes> is used to
4163 enable building on the XT3. In addition to enable XT3 specific
4164 features the option --enable-cray-xt3 must be used.
4169 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4170 Details : A race between allocating a new llcd and re-getting the llcd_lock
4171 allowed another thread to grab newly-allocated llcd.
4173 Severity : enhancement
4175 Description: 2.6 OST async journal commit and locking fix to improve performance
4176 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4177 2.6 kernels are now async as they already were in 2.4 kernels so
4178 that they can commit concurrently with the network bulk transfer.
4179 For block-allocated files the filter allocation semaphore is held
4180 to avoid filesystem fragmentation during allocation. BKL lock
4181 removed for 2.6 xattr operations where it is no longer needed.
4186 Description: lconf incorrectly determined whether two IP networks could talk
4187 Details : In some more complicated routing and multiple-network
4188 configurations, lconf will avoid trying to make a network
4189 connection to a disjoint part of the IP space. It was doing the
4190 math incorrectly for one set of cases.
4195 Description: Fix for potential infinite loop processing records in an llog.
4196 Details : If an llog record is corrupted/zeroed, it is possible to loop
4197 forever in llog_process(). Validate the llog record length
4198 and skip the remainder of the block on error.
4201 Frequency : occasional (liblustre only)
4203 Description: liblustre could not open files whose last component is a symlink
4204 Details : sysio_path_walk() would incorrectly pass the open intent to
4205 intermediate path components.
4208 Frequency : rare (liblustre only with non-standard tuning)
4209 Bugzilla : 7201 (7350)
4210 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4211 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4212 after it has been released, only do this for referenced locks
4214 Severity : enhancement
4216 Description: specifying an (invalid) directory default stripe_size of -1
4217 would reset the directory default striping
4218 Details : stripe_size -1 was used internally to signal directory stripe
4219 removal, now use "all default" to signal dir stripe removal
4220 as a directory striping of "all default" is not useful
4223 Frequency : common for large clusters running liblustre clients
4225 Description: doing an ls when liblustre clients are running is slow
4226 Details : sending a glimpse AST to a liblustre client waits for every AST
4227 to time out, as liblustre clients will not respond. Since they
4228 cannot cache data we refresh the OST lock LVB from disk instead.
4230 Severity : enhancement
4232 Description: doing an ls at the same time as file IO can be slow
4233 Details : enqueue and other "small" requests can be blocked behind many
4234 large IO requests. Create a new OST IO portal for non-IO
4235 requests so they can be processed faster.
4238 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4240 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4241 created file with mode 000
4242 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4243 allow an ftruncate() as a non-root user to a file with mode 000.
4244 Lustre now respects this flag to disable mode checks when
4245 truncating a file owned by the user
4248 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4251 Description: Revoking locks from clients that went dead or catatonic might take
4253 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4254 cancellation of such locks instant on servers without waiting for
4255 any reply from clients. Clients drops these locks when cancel
4256 notification from server is received without replying.
4259 Frequency : liblustre-only, when liblustre client dies or becomes busy
4261 Description: Doing ls on Linux clients can take a long time with active
4263 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4264 granting such locks to it in the first place if possible. Locks
4265 are taken by proxy on the OST during the read or write and
4266 dropped immediately afterward. Add connect flags handling, do
4267 not grant locks to liblustre clients for glimpse ASTs.
4269 Severity : enhancement
4271 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4272 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4273 sequential reads and behaves badly with applications doing random
4274 reads. Improve it by reading ahead at least read region, and
4275 avoiding excessive large RPC for small reads.
4277 Severity : enhancement
4279 Description: Creating more than 1000 files for a single job may cause a load
4280 imbalance on the OSTs if there are also a large number of OSTs.
4281 Details : qos_prep_create() uses an OST index reseed value that is an
4282 even multiple of the number of available OSTs so that if the
4283 reseed happens in the middle of the object allocation it will
4284 still utilize the OSTs as uniformly as possible.
4289 Description: OST or MDS may oops in ping_evictor_main()
4290 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4291 but doesn't restart at beginning of obd_exports_timed list
4294 Severity : enhancement
4296 Description: improve by-nid export eviction on the MDS and OST
4297 Details : allow multiple exports with the same NID to be evicted at one
4298 time without re-searching the exports list.
4301 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4303 Description: MDS may oops in groups_free()
4304 Details : in rare race conditions a newly allocated group_info struct is
4305 freed again, and this can be NULL. The 2.4 compatibility code
4306 for groups_free() checked for a NULL pointer, but 2.6 did not.
4309 Frequency : common for liblustre clients doing little filesystem IO
4310 Bugzilla : 9352, 7313
4311 Description: server may evict liblustre clients accessing contended locks
4312 Details : if a client is granted a lock or receives a completion AST
4313 with a blocking AST already set it would not reply to the AST
4314 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4317 Frequency : lfs setstripe, only systems with more than 160 OSTs
4319 Description: unable to set striping with a starting offset beyond OST 160
4320 Details : llapi_create_file() incorrectly limited the starting stripe
4321 index to the maximum single-file stripe count.
4324 Frequency : LDAP users only
4326 Description: lconf did not handle in-kernel recovery with LDAP properly
4327 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4329 Severity : enhancement
4331 Description: bind OST threads to NUMA nodes to improve performance
4332 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4333 node and do their allocations there to localize memory access
4335 Severity : enhancement
4337 Description: llmount can determine client NID directly from Myrinet (GM)
4338 Details : the client NID code from gmnalnid was moved directly into
4339 llmount, removing the need to use this or specifying the
4340 client NID explicitly when mounting GM clients with zeroconf
4343 Frequency : if client is started with down MDS
4345 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4346 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4347 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4348 ptlrpc_import_delayed_req() because we didn't check if the
4349 request was interrupted, and we also didn't break out of the
4350 event loop if there was no timeout
4355 Description: data loss during non-page-aligned writes to a single file from
4356 both multiple nodes and multiple threads on one node at same time
4357 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4358 inconsistency led to false short-reads, that were cached and later
4359 used by ->prepare_write() to fill in partially written page,
4360 leading to data loss.
4363 Frequency : always, if lconf --abort_recovery used
4365 Description: lconf --abort_recovery fails with 'Operation not supported'
4366 Details : lconf was attempting to abort recovery on the MDT device and not
4369 Severity : enhancement
4371 Description: remove cleanup logs
4372 Details : replace lconf-generated cleanup logs with lustre internal
4373 cleanup routines. Eliminates the need for client-cleanup and
4376 Severity : enhancement
4378 Description: add support for EAs (user and system) on lustre filesystems
4379 Details : it is now possible to store extended attributes in the Lustre
4380 client filesystem, and with the user_xattr mount option it
4381 is possible to allow users to store EAs on their files also
4383 Severity : enhancement
4385 Description: Add possibility (config option) to show minimal available OST free
4387 Details : When compiled with --enable-mindf configure option, statfs(2)
4388 (and so, df) will return least minimal free space available from
4389 all OSTs as amount of free space on FS, instead of summary of
4390 free spaces of all OSTs.
4392 Severity : enhancement
4394 Description: do not expand extent locks acquired on OST-side
4395 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4396 by server: they are not cached anyway.
4399 Frequency : when mmap is used/binaries executed from Lustre
4401 Description: Unmmap pages before throwing them away from read cache.
4402 Details : llap_shrink cache now attempts to unmap pages before discarding
4403 them (if unmapping failed - do not discard). SLES9 kernel has
4404 extra checks that trigger if this unmapping is not done first.
4409 Description: lconf didn't resolve symlinks before checking to see whether a
4410 given mountpoint was already in use
4413 Frequency : when migrating failover services
4414 Bugzilla : 6395, 9514
4415 Description: When migrating a subset of services from a node (e.g. failback
4416 from a failover service node) the remaining services would
4417 time out and evict clients.
4418 Details : lconf --force (implied by --failover) sets the global obd_timeout
4419 to 5 seconds in order to quickly disconnect, but this caused
4420 other RPCs to time out too quickly. Do not change the global
4421 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4423 Severity : enhancement
4424 Frequency : if MDS is started with down OST
4425 Bugzilla : 9439,5706
4426 Description: Allow startup/shutdown of an MDS without depending on the
4427 availability of the OSTs.
4428 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4429 Add appropriate locking and lov-osc refcounts for safe
4430 cleaning. Add osc abort_inflight calls in case the
4431 synchronize never started.
4434 Frequency : occasional (Cray XT3 only)
4436 Description: root not authorized to access files in CRAY_PORTALS environment
4437 Details : The client process capabilities were not honoured on the MDS in
4438 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4439 been accessed by an authorized user then root was able to access
4440 the file on the local client also. The root user capabilities
4441 are now allowed on the MDS, as this environment has secure UID.
4444 Frequency : occasional
4446 Description: ldiskfs "too long searching" message happens too often
4447 Details : A debugging message (otherwise harmless) prints too often on
4448 the OST console. This has been reduced to only happen when
4449 there are fragmentation problems on the filesystem.
4454 Description: Division by zero in statfs when all OSCs are inactive
4455 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4456 lov_statfs divides by value returned by lov_get_stripecnt().
4460 Bugzilla : 9489, 3273
4461 Description: First write from each client to each OST was only 4kB in size,
4462 to initialize client writeback cache, which caused sub-optimal
4463 RPCs and poor layout on disk for the first writen file.
4464 Details : Clients now request an initial cache grant at (re)connect time
4465 and so that they can start streaming writes to the cache right
4466 away and always do full-sized RPCs if there is enough data.
4467 If the OST is rebooted the client also re-establishes its grant
4468 so that client cached writes will be honoured under the grant.
4473 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4474 Details : Now I/O RPCs go to different portal number and (presumably) fast
4475 lock requests (and glimses) and other RPCs get their own service
4476 threads pool that should be able to service those RPCs
4479 Severity : enhancement
4481 Description: Ability to exchange lustre version between client and servers and
4482 issue warnings at client side if client is too old. Also for
4483 liblustre clients there is ability to refuse connection of too old
4485 Details : New 'version' field is added to connect data structure that is
4486 filled with version info. That info is later checked by server and
4490 Frequency : rare, liblustre only.
4491 Bugzilla : 9296, 9581
4492 Description: Two simultaneous writes from liblustre at offset within same page
4493 might proceed at the same time overwriting eachother with stale
4495 Details : I/O lock withing llu_file_prwv was released too early, before data
4496 actually was hitting the wire. Extended lock-holding time until
4497 server acknowledges receiving data.
4500 Frequency : extremely rare. Never observed in practice.
4502 Description: avoid generating lustre_handle cookie of 0.
4503 Details : class_handle_hash() generates handle cookies by incrementing
4504 global counter, and can hit 0 occasionaly (this is unlikely, but
4505 not impossible, because initial value of cookie counter is
4506 selected randonly). Value of 0 is used as a sentinel meaning
4507 "unassigned handle" --- avoid it. Also coalesce two critical
4508 sections in this function into one.
4510 Severity : enhancement
4512 Description: allow liblustre clients to delegate truncate locking to OST
4513 Details : To avoid overhead of locking, liblustre client instructs OST to
4514 take extent lock in ost_punch() on client's behalf. New connection
4515 flag is added to handle backward compatibility.
4517 Severity : enhancement
4518 Bugzilla : 4928, 7341, 9758
4519 Description: allow number of OST service threads to be specified
4520 Details : a module parameter allows the number of OST service threads
4521 to be specified via "options ost ost_num_threads={N}" in the
4522 OSS's /etc/modules.conf or /etc/modprobe.conf.
4526 Bugzilla : 6146, 9635, 9895
4527 Description: servers crash with bad pointer in target_handle_connect()
4528 Details : In rare cases when a client is reconnecting it was possible that
4529 the connection request was the last reference for that export.
4530 We would temporarily drop the export reference and get a new
4531 one, but this may have been the last reference and the export
4532 was just destroyed. Get new reference before dropping old one.
4534 Severity : enhancement
4535 Frequency : if client is started with failover MDS
4537 Description: Allow multiple MDS hostnames in the mount command
4538 Details : Try to read the configuration from all specified MDS
4539 hostnames during a client mount in case the "primary"
4542 Severity : enhancement
4544 Description: Stop sending data to evicted clients as soon as possible.
4545 Details : Check if the client we are about to send or are sending data to
4546 was evicted already. (Check is done every second of waiting,
4547 for which l_wait_event interface was extended to allow checking
4548 of exit condition at specified intervals).
4551 Frequency : rare, normally only when NFS exporting is done from client
4553 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4555 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4558 Frequency : always, if srand() is called before liblustre initialization
4560 Description: Liblustre uses system PRNG disturbing its usage by user application
4561 Details : Introduce internal to lustre fast and high-quality PRNG for
4562 lustre usage and make liblustre and some other places in generic
4563 lustre code to use it.
4565 Severity : enhancement
4566 Bugzilla : 9477, 9557, 9870
4567 Description: Verify that the MDS configuration logs are updated when xml is
4568 Details : Check if the .xml configuration logs are newer than the config
4569 logs stored on the MDS and report an error if this is the case.
4570 Request --write-conf, or allow starting with --old_conf.
4572 Severity : enhancement
4574 Description: Handle symlinks in the path when checking if Lustre is mounted.
4575 Details : Resolve intermediate symlinks when checking if a client has
4576 mounted a filesystem to avoid duplicate client mounts.
4581 Description: lconf can hit an error exception but still return success.
4582 Details : The lconf command catches the Command error exception at the top
4583 level script context and will exit with the associated exit
4584 status, but doesn't ensure that this exit status is non-zero.
4589 Description: failure of ptlrpc thread startup can cause oops
4590 Details : Starting a ptlrpc service thread can fail if there are a large
4591 number of threads or the server memory is very fragmented.
4592 Handle this without oopsing.
4595 Frequency : always, only if liblustre and non-default acceptor port was used
4597 Description: liblustre cannot connect to servers with non-default acceptor port
4598 Details : tcpnal_set_default_params() was not called and was therefore
4599 ignoring the environment varaible TCPNAL_PORT, as well as other
4600 TCPNAL_ environment variables
4605 Description: two objects could be created on the same OST for a single file
4606 Details : If an OST is down, in some cases it was possible to create two
4607 objects on a single OST for a single file. No problems other
4608 than potential performance impact and spurious error messages.
4612 Bugzilla : 5681, 9562
4613 Description: Client may oops in ll_unhash_aliases
4614 Details : Client dcache may become inconsistent in race condition.
4615 In some cases "getcwd" can fail if the current directory is
4621 Description: Inode refcounting problems in NFS export code
4622 Details : link_raw functions used to call d_instantiate without obtaining
4623 extra inode reference first.
4627 Bugzilla : 9942, 9903
4628 Description: Referencing freed requests leading to crash, memleaks with NFS.
4629 Details : We used to require that call to ll_revalidate_it was always
4630 followed by ll_lookup_it. Also with revalidate_special() it is
4631 possible to call ll_revalidate_it() twice for the same dentry
4632 even if first occurence returned success. This fix changes semantic
4633 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4634 reference on a request referred from the intent.
4635 ll_intent_release() then releases such a request.
4638 Frequency : rare, normally benchmark loads only
4640 Description: unlinked inodes were kept in memory on the client
4641 Details : If a client is repeatedly creating and unlinking files it
4642 can accumulate a lot of stale inodes in the inode slab cache.
4643 If there is no other client load running this can cause the
4644 client node to run out of memory. Instead flush old inodes
4645 from client cache that have the same inode number as a new inode.
4648 Frequency : SLES9 2.6.5 kernel and long filenames only
4649 Bugzilla : 9969, 10379
4650 Description: utime reports stale NFS file handle
4651 Details : SLES9 uses out-of-dentry names in some cases, which confused
4652 the lustre dentry revalidation. Change it to always use the
4656 Frequency : rare, unless heavy write-truncate concurrency is continuous
4657 Bugzilla : 4180, 6984, 7171, 9963, 9331
4658 Description: OST becomes very slow and/or deadlocked during object unlink
4659 Details : filter_destroy() was holding onto the parent directory lock
4660 while truncating+unlinking objects. For very large objects this
4661 may block other threads for a long time and slow overall OST
4662 responsiveness. It may also be possible to get a lock ordering
4663 deadlock in this case, or run out of journal credits because of
4664 the combined truncate+unlink. Solution is to do object truncate
4665 first in one transaction without parent lock, and then do the
4666 final unlink in a new transaction with the parent lock. This
4667 reduces the lock hold time dramatically.
4670 Frequency : rare, 2.4 kernels only
4672 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4673 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4674 before unmounting their filesystem in order to avoid deadlock.
4675 The kernel_locked() function in 2.4 kernels only checks whether
4676 the kernel lock is held, not whether it is this process that is
4677 holding it as 2.6 kernels do.
4682 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4683 Details : The client ptlrpc code may be trying to reconnect to a down
4684 server before a previous connection attempt has timed out.
4685 Increase the reconnect interval to be longer than the connection
4686 timeout interval to avoid sending duplicate connections to
4690 Frequency : echo_client brw_test command
4692 Description: fix echo_client to work with OST preallocated code
4693 Details : OST preallocation code (5137) didn't take echo_client IO path
4694 into account: echo_client calls filter methods outside of any
4695 OST thread and, hence, there is no per-thread preallocated
4696 pages and buffers to use. Solution: hijack pga pages for IO. As
4697 a byproduct, this avoids unnecessary data copying.
4701 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4702 Description: Client can oops in mdc_commit_close() after open replay
4703 Details : It was possible for the MDS to return an open request with no
4704 transaction number in mds_finish_transno() if the client was
4705 evicted, but without actually returning an error. Clients
4706 would later try to replay that open and may trip an assertion
4707 Simplify the client close codepath, and always return an error
4708 from the MDS in case the open is not successful.
4711 Frequency : rare, 2.6 OSTs only
4713 Description: OST may deadlock under high load on fragmented files
4714 Details : If there was a heavy load and highly-fragmented OST filesystems
4715 it was possible to have all the OST threads deadlock waiting on
4716 allocation of biovecs, because the biovecs were not released
4717 until the entire RPC IO was completed. Instead, release biovecs
4718 as soon as they are complete to ensure forward IO progress.
4720 Severity : enhancement
4722 Description: Support for specifying external journal device at mount
4723 Details : If an OST or MDS device is formatted with an external journal
4724 device, this device major/minor is stored in the ext3 superblock
4725 and may not be valid for failover. Allow detecting and
4726 specifying the external journal at mount time.
4731 Description: Mounting an MDS with pending unlinked files may cause oops
4732 Details : target_finish_recovery() calls mds_postrecov() which returned
4733 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4734 considers this an error and immediately begins cleaning up the
4735 lov, just after starting the mds_lov process
4737 Severity : enhancement
4739 Description: Implement 'lfs df' to report actual free space on per-OST basis
4740 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4741 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4742 usage of objects; '-h' to report in human readable format.
4744 ------------------------------------------------------------------------------
4746 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4753 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4754 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4755 at startup during orphan unlinking. Instead, skip buddy update
4756 before setup, as it will be regenerated after recovery is complete.
4761 Description: If an OST is inactive, its locks might reference stale inodes.
4762 Details : lov_change_cbdata() must iterate over all namespaces, even if
4763 they are inactive to clear inode references from the lock.
4765 Severity : enhancement
4766 Frequency : occasional, if non-standard max_dirty_mb used
4768 Description: Client will block write RPCs if not enough grant
4769 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4770 then the client will block writes while waiting for another RPC
4771 to complete instead of consuming its dirty limit. With change
4772 we get improved performance when max_dirty_mb is small.
4774 Severity : enhancement
4775 Bugzilla : 3389, 6253
4776 Description: Add support for supplementary groups on the MDS.
4777 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4778 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4779 lookups for user supplementary groups into a cache.
4783 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4784 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4785 for liblustre O_EXCL is enforced by MDS.
4789 Description: readdir never returns NULL in liblustre.
4790 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4791 dentry in struct dirent.
4795 Frequency : liblustre only, depends on application IO pattern
4796 Description: liblustre clients evicted if not contacting servers
4797 Details : Don't put liblustre clients into the ping_evictor list, so
4798 they will not be evicted by the pinger ever.
4800 Severity : enhancement
4802 Description: Add ability to evict clients by NID from MDS.
4803 Details : By echoing "nid:$NID" string into
4804 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4805 $NID would be instantly evicted from this MDS and from all active
4806 OSTs connected to it.
4810 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4811 Details : lookup_it_finish() used to query file size from OSTs that was not
4816 Description: service threads change working directory to that of init
4817 Details : Starting lustre service threads may pin the working directory
4818 of the parent thread, making that filesystem busy. Threads
4819 now change to the working directory of init to avoid this.
4823 Frequency : during shutdown only
4824 Description: shutdown with a failed MDS or OST can cause unmount to hang
4825 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4830 Frequency : 2.6 only
4831 Description: chmod/chown may include an extra supplementary group
4832 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4833 supplementary group and if none is specified this is used.
4836 Bugzilla : 5479 (6816)
4838 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4839 Details : If lookup is in progress on a file that is unlinked we might try
4840 to revalidate the inode and fail in revalidate after lookup is
4841 complete and ll_file_open() enqueues the open again but
4842 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4845 Frequency : always, if lconf --abort_recovery used
4847 Description: lconf --abort_recovery fails with 'Operation not supported'
4848 Details : lconf was attempting to abort recovery on the MDT device and not
4851 ------------------------------------------------------------------------------
4853 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4858 Frequency : rare (only unsupported configurations with a node running as an
4860 Bugzilla : 6514, 5137
4861 Description: Mounting a Lustre file system on a node running as an OST could
4863 Details : OSTs now preallocates memory needed to write out data at
4864 startup, instead of when needed, to avoid having to
4865 allocate memory in possibly low memory situations.
4866 Specifically, if the file system is mounted on on OST,
4867 memory pressure could force it to try to write out data,
4868 which it needed to allocate memory to do. Due to the low
4869 memory, it would be unable to do so and the node would
4870 become unresponsive.
4872 Severity : enhancement
4874 Description: Addition of lconf --service command line option
4875 Details : lconf now accepts a '--service <arg>' option, which is
4876 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4878 Severity : enhancement
4880 Description: Failover mode is now the default for OSTs.
4881 Details : By default, OSTs will now run in failover mode. To return to
4882 the old behaviour, add '--failout' to the lmc line for OSTs.
4884 Severity : enhancement
4886 Description: Health checks are now provided for MDS and OSTs
4887 Details : Additional detailed health check information on MSD and OSTs
4888 is now provided through the procfs health_check value.
4891 Frequency : occasional, depends on IO load
4893 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4894 after numerous create/delete cycles
4895 Details : The ext3 inode allocation policy would not allocate new inodes
4896 very well on the OSTs because there are no new directories
4897 being created. Instead we look for groups with free space if
4898 the parent directories are nearly full.
4903 Description: Network or server problems during mount may cause partially
4904 mounted clients instead of returning an error.
4905 Details : The config llog parsing code may overwrite the error return
4906 code during mount error handling, returning success instead
4912 Description: MDS can fail to allocate large reply buffers
4913 Details : After long uptimes the MDS can fail to allocate large reply
4914 buffers (e.g. zconf client mount config records) due to memory
4915 fragmentation or consumption by the buffer cache. Preallocate
4916 some large reply buffers so that these replies can be sent even
4917 under memory pressure.
4921 Frequency : rare (liblustre)
4922 Description: fsx running with liblustre complained that using truncate() to
4923 extend the file doesn't work. This patch corrects that issue.
4924 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4925 ATTR_SIZE and lsm use in llu_setattr_raw.
4929 Frequency : rare, only 2.6 kernels
4930 Description: Unusual file access patterns on the MDS may result in inode
4931 data being lost in very rare circumstances.
4932 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4933 code in the 2.6 kernel caused the inode and/or EA data not to
4934 be read from disk, causing single-file corruption.
4938 Frequency : rare, only 2.6 filesystems using extents
4939 Description: Heavy concurrent write and delete load may cause data corruption.
4940 Details : It was possible under high-load situations to have an extent
4941 metadata block in the block device cache from a just-unlinked
4942 file overwrite a newly-allocated data block. We now unmap any
4943 metadata buffers that alias just-allocated data blocks.
4947 Frequency : filesystems with default stripe_count larger than 77
4948 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4949 Details : lconf specifies an inode size of 4096 bytes when the default
4950 stripe_count is larger than 77. This conflicts with the default
4951 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4952 this case to avoid pinning too much memory for large EAs.
4954 ------------------------------------------------------------------------------
4956 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4961 Frequency : rare (extremely heavy IO load with hundreds of clients)
4963 Description: Client is evicted, gets IO error writing to file
4964 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4965 caused clients to be evicted for AST timeouts. The fixes in
4966 bug 5192 mean we no longer need to have such short AST timeouts
4967 so ldlm_timeout has been increased.
4970 Frequency : occasional during --force or --failover shutdown under load
4971 Bugzilla : 5949, 4834
4972 Description: Server oops/LBUG if stopped with --force or --failover under load
4973 Details : a collection of import/export refcount and cleanup ordering
4974 issues fixed for safer force cleanup
4977 Frequency : only filesystems larger than 120 OSTs
4978 Bugzilla : 5990, 6223
4979 Description: lfs getstripe would oops on a very large filesystem
4980 Details : lov_getconfig used kfree on vmalloc'd memory
4983 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4984 Bugzilla : 6242, 6243
4985 Description: reading from files that had been truncated to a non-zero size
4986 but never opened returned no data
4987 Details : ll_file_read() reads zeros from no-object files to EOF
4992 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4993 Details : There's little harm in aborting MDS/OSS recovery and letting it
4994 try again, so I removed the LASSERT and return an error instead.
4996 Severity : enhancement
4998 Description: New debugging infrastructure for tracking down data corruption
4999 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5000 (b) cover more of the client-side code path, and (c) try to narrow
5001 down where problems occurred
5005 Bugzilla : 3819, 4364, 4397, 6313
5006 Description: Racing close and eviction MDS could cause assertion in mds_close
5007 Details : It was possible to get multiple mfd references during close and
5008 client eviction, leading to one thread referencing a freed mfd.
5010 Severity: : enhancement
5011 Bugzilla : 3262, 6359
5012 Description: Attempts to reconnect to servers are now more aggressive.
5013 Details : This builds on the enhanced upcall-less recovery that was added
5014 in 1.4.2. When trying to reconnect to servers, clients will
5015 now try each server in the failover group every 10 seconds. By
5016 default, clients would previously try one server every 25 seconds.
5021 Description: After recovery, certain operations trigger a failed
5022 assertion on a client.
5023 Details : Failing over an mds, using lconf -d --failover, while a
5024 client was doing a readdir() call would cause the client to
5025 LBUG after recovery completed and the readdir() was resent.
5027 Severity : enhancement
5029 Description: Default groups are now added by lconf
5030 Details : You can now run lconf --group <servicename> without having to
5031 manually add groups with lmc.
5034 Frequency : occasional
5036 Description: Nodes with an elan id of 0 trigger a failed assertion
5039 Frequency : always when accessing e.g. tty/console device nodes
5041 Description: tty and some other devices nodes cannot be used on lustre
5042 Details : file's private_data field is used by device data and lustre
5043 values in there got lost. New field was added to struct file to
5044 store fs-specific private data.
5047 Frequency : when exporting Lustre via NFS
5049 Description: NFSD failed occasionally when looking up a path component
5050 Details : NFSD is looking up ".." which was broken in ext3 directories
5051 that had grown large enough to become hashed.
5054 Frequency : Clusters with multiple interfaces not on the same subnet
5056 Description: Nodes will repeatedly try to reconnect to an interface which it
5057 cannot reach and report an error to the log.
5058 Details : Extra peer list entries will be created by lconf with some peers
5059 unreachable. lconf now validates the peer before adding it.
5062 Frequency : Only if a default stripe is set on the filesystem root.
5064 Description: Setting a default stripe on the filesystem root prevented the
5065 filesystem from being remounted.
5066 Details : The client was sending extra request flags in the root getattr
5067 request and did not allocate a reply buffer for the dir EA.
5070 Frequency : occasional, higher if lots of files are accessed by one client
5071 Bugzilla : 6159, 6097
5072 Description: Client trips assertion regarding lsm mismatch/magic
5073 Details : While revalidating inodes the VFS looks up inodes with ifind()
5074 and in rare cases can find an inode that is being freed.
5075 The ll_test_inode() code will free the lsm during ifind()
5076 when it finds an existing inode and then the VFS later attaches
5077 this free lsm to a new inode.
5081 Bugzilla : 6422, 7030
5082 Description: MDS deadlock between mkdir and client eviction
5083 Details : Creating a new file via mkdir or mknod (starting a transaction
5084 and getting the ns lock) can deadlock with client eviction
5085 (gets ns lock and trying to finish a synchronous transaction).
5088 Frequency : occasional
5089 Description: While starting a server, the fsfilt_ext3 module could not be
5091 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5092 kernels. Previously, lconf would still use the ext3 name
5093 when trying to load modules. Now, it will correctly use
5094 ext3 on 2.4 and ldiskfs on 2.6.
5096 Severity : enhancement
5097 Description: The default stripe count has been changed to 1
5098 Details : The interpretation of the default stripe count (0, to lfs
5099 or lmc) has been changed to mean striping across a single
5100 OST, rather than all available. For general usage we have
5101 found a stripe count of 1 or 2 works best.
5103 Severity : enhancement
5104 Description: Add support for compiling against Cray portals.
5105 Details : Conditional compiling for some areas that are different
5109 Frequency : occasional
5110 Bugzilla : 6409, 6834
5111 Description: Creating files with an explicit stripe count may lead to
5112 a failed assertion on the MDS
5113 Details : If some OSTs are full or unavailable, creating files may
5114 trigger a failed assertion on the MDS. Now, Lustre will
5115 try to use other servers or return an error to the
5119 Frequency : occasional
5121 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5122 caused lock timeout during readahead (since 1.4.2).
5123 Details : Processes doing readahead might match a lock that hasn't been
5124 granted yet if there are overlapping and conflicting lock
5125 requests. The readahead process waits on ungranted lock
5126 (original lock is CBPENDING), while OST waits for that process
5127 to cancel CBPENDING read lock and eventually evicts client.
5129 Severity : enhancement
5131 Description: Initial enabling of flock support for clients
5132 Details : Implements fcntl advisory locking and file status functions.
5133 This feature is provided as an optional mount flag (default
5134 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5135 locking are implemented yet, and those that are are not guaranteed
5136 to be completely correct in production environments.
5137 mount -t lustre -o [flock|noflock] ...
5140 Frequency : occasional
5142 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5143 assertion in the ext3 JBD (journaling) layer.
5144 Details : The b_committed_data struct is protected by the big kernel lock
5145 in 2.4 kernels, serializing journal_commit_transaction() and
5146 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5147 access to this struct. In 2.6 kernels there is finer grained
5148 locking to improve SMP performance of the JBD layer.
5152 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5157 Description: merge and cleanup kernel patches.
5160 -----------------------------------------------------------------------------
5162 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5164 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5165 versions of Lustre. Please update all servers and clients to
5166 version 1.4.2 or later at the same time. You must also run
5167 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5168 to update the configuration logs.
5170 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5171 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5172 - swab llog records of type '0' so we get proper header size/idx (5861)
5173 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5174 - fix rename of one directory over another leaking an inode (5953)
5175 - avoid SetPageDirty on 2.6 (5981)
5176 - don't re-add just-being-destroyed locks to the waiting list (5653)
5177 - when creating new directories, inherit the parent's custom
5178 striping settings if present parent (3048)
5179 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5180 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5181 - don't hold client locks on temporary worklist from l_lru (5666)
5182 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5183 - automatically evict dead clients (5921)
5184 - Update file size properly in create+truncate+fstat case (6196)
5185 - Do not unhash mountpoint dentries, do not allow removal of
5187 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5188 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5189 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5190 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5191 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5192 - Correctly handle reads of files with no objects (6243)
5193 - lctl recover will also mark a device active if deactivate used (5933)
5195 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5196 - allow --write-conf on an MDS with different nettype than client (5619)
5197 - don't write config llogs to MDS for mounts not from that MDS (5617)
5198 - lconf should create multiple TCP connections from a client (5201)
5199 - init scripts are now turned off by default; run chkconfig --on
5200 lustre and chkconfig --on lustrefs to use them
5201 - upcalls are no longer needed for clients to recover to failover
5203 - add --abort-recovery option to lconf to abort recovery on device
5205 - add support for an arbitrary number of OSTs (3026)
5206 - Quota support protocol changes.
5207 - forward compatibility changes to wire structs (6007)
5208 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5209 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5210 - improved "lustre status" script
5211 - initialize blocksize for non-regular files (6062)
5212 - added --disable-server and --disable-client configure options (5782)
5213 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5214 - Vanilla 2.4.29 support
5215 - increase maximum number of obd devices to 520 (6242)
5216 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5217 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5218 SLES 9 2.6.5 kernels
5219 - limit stripes per file to 160 (the maximum EA size) (6093)
5221 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5224 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5225 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5226 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5227 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5228 - small ext3 extents cleanups and fixes (5733)
5229 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5230 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5231 - avoid needless client->OST connect, fix handle mismatch (5317)
5232 - fix DLM error path that led to out-of-sync client, long delays (5779)
5233 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5234 - fix several locking issues related to i_size (5492,5624,5654,5672)
5235 - don't move pending lock onto export if it is already evicted (5683)
5236 - fix kernel oops when creating .foo in unlinked directory (5548)
5237 - fix deadlock in obdfilter statistics vs. object create (5811)
5238 - use time_{before,after} to avoid timer jiffies wrap (5882)
5239 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5240 - Do not leak request if server was not able to process it (5154)
5241 - If mds_open unable to find parent dir, make that negative lookup(5154)
5242 - don't create new directories with extent-mapping (5909, 5936)
5244 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5245 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5246 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5247 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5248 - support for RHEL4 kernel on i686 (5773)
5249 - provide error messages when incompatible logs are encountered (5898)
5251 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5252 * version 1.4.0.10 (1.4.1 release candidate 1)
5254 - don't keep a lock reference when lock is not granted (4238)
5255 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5256 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5257 - drop import inflight refcount on signal_completed_replay error (5255)
5258 - unlock page after async write error during send (3677)
5259 - handle missing objects in filter_preprw_read properly (5265)
5260 - no transno return for symlink open, don't save no-trasno open (3440)
5261 - don't try to complete elan receive that already failed (4012)
5262 - free RPC server reply state on error (5406)
5263 - clean up thread from ptlrpc_start_thread() on error (5160)
5264 - readahead could read extra page into cache that wasn't ejected (5388)
5265 - prevent races in class_attach/setup/cleanup/detach (5260)
5266 - don't dereference de->d_inode after l_dput of de (5458)
5267 - use "int" for stripe value returned from lock_to_stripe (5544)
5268 - mballoc allocation and error-checking fixes in 2.6 (5504)
5269 - block device patches to fix I/O request sizes in 2.6 (5482)
5270 - look up hostnames for IB nals (5602)
5271 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5272 - don't start multiple acceptors for the same port (5277)
5273 - fix incorrect LASSERT in mds_getattr_name (5635)
5274 - export a proc file for general "ping" checking (5628)
5275 - fix "lfs check" to not block when the MDS is down (5628)
5277 - service request history (4965)
5278 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5279 - create an "evict_client" /proc entry on OSTs, like the MDS has
5280 - fix mount usage message, return errors per mount(8) (5168)
5281 - change grep [] to grep "[]" in tests so they work in more UMLs
5282 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5283 - remove ancient LOV_MAGIC_V0 EA support (5047)
5284 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5285 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5286 - allow readahead allocations to fail when low on memory (5383)
5287 - mmap locking landed again, after considerable improvement (2828)
5288 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5290 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5293 - send OST transaction number in read/write reply to free req (4966)
5294 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5295 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5296 - the watchdog thread now runs as interruptible (5246)
5297 - flock/lockf fixes (but it's still disabled, pending 5135)
5298 - don't use EXT3 constants in llite code (5094)
5299 - memory shortage at startup could cause assertion (5176)
5301 - reorganization of lov code
5302 - single portals codebase
5304 - add extents/mballoc support (5025)
5305 - direct I/O reads in the obdfilter (4048)
5306 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5308 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5311 - send OST transaction number in read/write reply to free req (4966)
5312 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5313 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5314 - the watchdog thread now runs as interruptible (5246)
5315 - handle missing objects in filter_preprw_read properly (5265)
5316 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5317 - drop import inflight refcount on signal_completed_replay error (5255)
5318 - unlock page after async write error during send (3677)
5319 - return original error code on reconstructed replies (3761)
5320 - no transno return for symlink open, don't save no-trasno open (3440)
5322 - add pid to ldlm debugging output (4922)
5323 - bump the watchdog timeouts -- we can't handle 30sec yet
5324 - extra debugging for orphan dentry/inode bug (5259)
5326 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5329 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5330 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5331 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5332 - fix NULL dereference in /proc/sys/portals/routes (4827)
5333 - allow failed mdc_close() operations to be interrupted (4561)
5334 - stop precreate on OST before MDS would time out on it (4778)
5335 - don't send partial-page writes before EOF from client (4410)
5336 - discard client grant for sub-page writes on large-page clients (4520)
5337 - don't free dentries not owned by NFS code, check generation (4806)
5338 - fix lsm leak if mds_create_objects() fails (4801)
5339 - limit debug_daemon file size, always print CERROR messages (4789)
5340 - use transno after validating reply (3892)
5341 - process timed out requests if import state changes (3754)
5342 - update mtime on OST during writes, return in glimpse (4829)
5343 - add mkfsoptions to LDAP (4679)
5344 - use ->max_readahead method instead of zapping global ra (5039)
5345 - don't interrupt __l_wait_event() during strace
5347 - add software watchdogs to catch hung threads quickly (4941)
5348 - make lustrefs init script start after nfs is mounted
5349 - fix CWARN/ERROR duplication (4930)
5350 - return async write errors to application if possible (2248)
5351 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5352 - print NAL number in %x format (4645)
5353 - update barely-supported suse-2.4.21-171 series (4842)
5354 - support for sles 9 %post scripts
5355 - support for building 2.6 kernel-source packages
5356 - support for sles km_* packages
5358 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5361 - ignore -ENOENT errors in osc_destroy (3639)
5362 - notify osc create thread that OSC is being cleaned up (4600)
5363 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5364 - reconstruct ost_handle() like mds_handle() (4657)
5365 - create a new thread to do import eviction to avoid deadlock (3969)
5366 - let lconf resolve symlinked-to devices (4629)
5367 - don't unlink "objects" from directory with default EA (4554)
5368 - hold socknal file ref over connect in case target is down (4394)
5369 - allow more than 32000 subdirectories in a single directory (3244)
5370 - fix blocks count for O_DIRECT writes (3751)
5371 - OST returns ENOSPC from object create when no space left (4539)
5372 - don't send truncate RPC if file size isn't changing (4410)
5373 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5374 - bind to privileged port in socknal and tcpnal (3689)
5376 - rate limit CERROR/CWARN console message to avoid overload (4519)
5377 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5378 - basic mmap support (3918)
5379 - kernel patch series update from b1_4 (4711)
5381 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5384 - avoid crash during MDS cleanup with OST shut down (2775)
5385 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5386 - don't use bad inodes on the MDS (3744)
5387 - dynamic object preallocation to improve recovery speed (4236)
5388 - don't hold spinlock over lock dumping or change debug flags (4401)
5389 - don't zero obd_dev when it is force cleaned (3651)
5390 - print grants to console if they go negative (4431)
5391 - "lctl deactivate" will stop automatic recovery attempts (3406)
5392 - look for existing locks in ldlm_handle_enqueue() (3764)
5393 - don't resolve lock handle twice in recovery avoiding race (4401)
5394 - revalidate should check working dir is a directory (4134)
5396 - don't always mark "slow" obdfilter messages as errors (4418)
5398 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5401 - don't close LustreDB during write_conf until it is done (3860)
5402 - fix typo in lconf for_each_profile (3821)
5403 - allow dumping logs from multiple threads at one time (3820)
5404 - don't allow multiple threads in OSC recovery (3812)
5405 - fix debug_size parameters (3864)
5406 - fix mds_postrecov to initialize import for llog ctxt (3121)
5407 - replace config semaphore with spinlock (3306)
5408 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5409 - don't allow enqueue to complete on a destroyed export (3822)
5410 - down write_lock before checking llog header bitmap (3825)
5411 - recover from lock replay timeout (3764)
5412 - up llog sem before sending rpc (3652)
5413 - reduce ns lock hold times when setting kms (3267)
5414 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5415 - fix NULL deref and obd_dev leak on setup error (3312)
5416 - replace some LBUG about llog ops with error handling (3841)
5417 - don't match INVALID dentries from d_lookup and spin (3784)
5418 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5419 - fix invalid assertion in ptlrpc_set_wait (3880)
5421 - add libwrap support for the TCP acceptor (3996)
5422 - add /proc/sys/portals/routes for non-root route listing (3994)
5423 - allow setting MDS UUID in .xml (2580)
5424 - print the stack of a process that LBUGs (4228)
5426 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5429 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5430 - make sure to unset replay flag from failed open requests (3440)
5431 - if default stripe count is 0, use OST count for inode size (3636)
5432 - update parent mtime/ctime on client for create/unlink (2611)
5433 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5434 - free recovery state on server during a forced cleanup (3571)
5435 - unregister_reply for resent reqs (3063)
5436 - loop back devices mounting and status check on 2.6 (3563)
5437 - fix resource-creation race that can provoke i_size == 0 (3513)
5438 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5439 - more debugging for page-accounting assertion (3746)
5440 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5441 - avoid deadlock after precreation failure (3758)
5442 - fix race and lock order deadlock in orphan handling (3450, 3750)
5443 - add validity checks when grabbing inodes from l_ast_data (3599)
5445 - add /proc/.../recovery_status to obdfilter (3428)
5446 - lightweight CDEBUG infrastructure, debug daemon (3668)
5447 - change default OSC RPC parameters to be better on small clusters
5448 - turn off OST read cache for files smaller than 32MB
5449 - install man pages and include them in rpms (3100)
5450 - add new init script for (un)mounting lustre filesystems (2593)
5451 - run chkconfig in %post for init scripts (3701)
5452 - drop scimac NAL (unmaintained)
5454 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5457 - clean kiobufs before and after use (3485)
5458 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5459 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5460 - change init script to fail more clearly if not run as root (1528)
5461 - allow clients to reconnect during replay (1742)
5462 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5463 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5464 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5465 - fix cond_resched (3554)
5466 - only evict unfinished clients after recovery (3515)
5467 - allow bulk resend, prevent data loss (3570)
5468 - dynamic ptlrpc request buffer allocation (2102)
5469 - don't allow unlinking open directory if it isn't empty (2904)
5470 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5471 - remove extraneous obd dereference causing LASSERT failure (3334)
5472 - don't use get_cycles() when creating temp. files on the mds (3156)
5473 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5474 - handle EEXIST for set-stripe, set proper directory name (3336)
5476 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5477 - fix ksocknal_fmb_callback() error messages (2918)
5479 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5482 - don't copy lvb into (possibly NULL) reply on error (2983)
5483 - don't deref dentry after dput, don't free lvb on error (2922)
5484 - use the kms to determine writeback rpc length (2947)
5485 - increment oti_logcookies when osc is inactive (2948)
5486 - update client's i_blocks count via lvb messages (2543)
5487 - handle intent open/close of special files properly (1557)
5488 - mount MDS with errors=remount-ro, like obdfilter (2009)
5489 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5490 - don't use cancelling-locks' kms values (2947)
5491 - use highest lock extent for kms, not last one (2925)
5492 - don't dereference ERR_PTR() dentry in error handling path (3107)
5493 - fix thread race in portals_debug_dumplog() (3122)
5494 - create lprocfs device entries at setup instead of at attach (1519)
5495 - common AST error handler, don't evict client on completion race (3145)
5496 - zero nameidata in detach_mnt in 2.6 (3118)
5497 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5498 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5499 - initialize RPC timeout timer earlier for 2.6 (3219)
5500 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5501 - print nal/nid for unknown nid (3258)
5502 - additional checks for oscc recovery before doing precreate (3284)
5503 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5504 - don't crash in mdc_close for bad permissions on open (3285)
5505 - zero i_rdev for non-device files (3147)
5506 - clear page->private before handing to FS, better assertion (3119)
5507 - tune the read pipeline (3236)
5508 - fix incorrect decref of invalidated dentry (2350)
5509 - provide read-ahead stats and refine rpc in flight stats (3328)
5510 - don't hold journal transaction open across create RPC (3313)
5511 - update atime on MDS at close time (3265)
5512 - close LDAP connection when recovering to avoid server load (3315)
5513 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5514 - don't leak open file on MDS after open resend (3325)
5515 - serialize filter_precreate and filter_destroy_precreated (3329)
5516 - loop device shouldn't call sync_dev() for nul device (3092)
5517 - clear page cache after eviction (2766)
5518 - resynchronize MDS->OST in background (2824)
5519 - refuse to mount the same filesystem twice on same mountpoint (3394)
5520 - allow llmount to create routes for mounting behind routers (3320)
5521 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5522 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5523 - fix rare problem with rename on htree directories (3417)
5525 - allow default OST striping configuration per directory (1414)
5526 - fix compilation for qswnal for 2.6 kernels (3125)
5527 - increase maximum number of MDS request buffers for large systems
5528 - change liblustreapi to be useful for external progs like lfsck (3098)
5529 - increase local configuration timeout for slow disks (3353)
5530 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5532 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5535 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5536 - don't overwrite extent policy data in reply if lock was blocked (2901)
5537 - drop filter export grants atomically with removal from device (2663)
5538 - del obd_self_export from work_list in class_disconnect_exports (2908)
5539 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5540 - swab reply message in mdc_close, other PPC fixes (2464)
5541 - fix destroying of named logs (2325)
5542 - overwrite old logs when running lconf --write_conf (2264)
5543 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5544 - fix race in target_handle_connect (2898)
5545 - mds_reint_create() should take same inode create lock (2926)
5546 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5547 - don't close files for self_export to avoid uninitialized obd (2936)
5548 - allow MDS with the same name as client node (2939)
5549 - hold dentry reference for closed log files for unlink (2325)
5550 - reserve space for all logs during transactions (2059)
5551 - don't evict page beyond end of stripe extent (2925)
5552 - don't oops on a deleted current working directory (2399)
5553 - handle hard links to targets without a parent properly (2517)
5554 - don't dereference NULL lock when racing during eviction (2867)
5555 - don't grow lock extents when lots of conflicting locks (2919)
5557 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5560 - account for cache space usage on clients to avoid data loss (974)
5561 - lfsck support in lustre kernel code (2349)
5562 - reduce journal credits needed for BRW writes (2370)
5563 - orphan handling to avoid losing space on client/server crashes
5564 - ptlrpcd can be blocked, stopping ALL progress (2477)
5565 - use lock value blocks to assist in proper KMS, faster stat (1021)
5566 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5567 - recovery for initial connections (2355)
5568 - fixes for mds_cleanup_orphans (1934)
5569 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5570 - block all file creations until orphan recovery completes (1901)
5571 - client remove rq_connection from request struct (2423)
5572 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5573 - recovery timer race (2670)
5574 - mdc_close recovey bug (2532)
5575 - ptlrpc cleanup bug (2710)
5576 - mds timeout on local locks (2588)
5577 - namespace lock held during RPCs (2431)
5578 - handle interrupted sync write properly (2503)
5579 - don't try to handle a message that hasn't been replied to (2699)
5580 - client assert failure during cleanup after abort recovery (2701)
5581 - leak mdc device after failed mount (2712)
5582 - ptlrpc_check_set allows timedout requests to complete (2714)
5583 - wait for inflight reqs when ptlrpcd finishes (2710)
5584 - make sure unregistered services are removed from the srv_list
5585 - reset bulk XID's when resending them (caught by 1138 test)
5586 - unregister_bulk after timeout
5587 - fix lconf error (2694)
5588 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5589 - readahead locks pages, leaves pending causing memory pressure (2673)
5590 - increase OST request buffers to 4096 on large machines (2729)
5591 - fix up permission of existing directories in simple_mkdir (2661)
5592 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5593 - don't assign transno to errored transactions (2742)
5594 - don't delete objects on OST if given a bogus objid from MDS (2751)
5595 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5596 - if rq_no_resend, then timeout request after recovery (2432)
5597 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5598 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5599 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5600 - MDS recovery shouldn't delete objects that it has given out (2730)
5601 - if enqueue arrives after completion, don't clobber LVB (2819)
5602 - don't unlock pages twice when trigger_group_io returns error (2814)
5603 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5604 - don't write pages to disk if there was an error (1450)
5605 - don't ping imports that have recovery disabled (2676)
5606 - take buffered bytes into account when balancing socknal conn (2817)
5607 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5608 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5609 - remove little-endian swabbing of llog records (1987)
5610 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5611 - retry reposting request buffers if they fail (1191)
5612 - grow extent at grant time to avoid granting a revoked lock (2809)
5613 - lock revoke doesn't evict page if covered by a second lock (2765)
5614 - disable VM readahead to avoid reading outside lock extents (2805)
5616 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5617 - updated kernel patches for hp-2.4.20 kernel (2681)
5619 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5622 - fix truncated write corruption (2366)
5623 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5625 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5626 - ensure len(uuid) < 37 in lmc (1171)
5627 - fix ia64 OOPS in llog_test (2255)
5628 - zero end of page at obdfilter for partial page writes (2648)
5629 - don't leave stale dentries around after renames (bug 2428)
5630 - fix timeouts when evicting a client with a single lock held (2642)
5631 - set deadline for the initial HELLO message to drain (2634)
5632 - print out dotted-quad IP addresses in the socknal (2302)
5634 - additional debugging for MDS client eviction problem (2443)
5635 - fix mkfsoptions support for osts (2603, 2604)
5637 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5640 - add series for the vanilla 2.6.0 kernel
5641 - add series for the vanilla 2.4.24 kernel
5642 - add series for a cray x86/64 UL kernel drop
5643 - fix xattr patches for the vanilla 2.4.19 series
5645 - generate true UUIDs in lmc (1171)
5646 - have portals stack dumping break in UML (2466)
5647 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5648 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5649 - wake callback waiting threads on client eviction (2460)
5650 - Add --ptldebug and --subsystem to lmc (1719)
5651 - update assertion to allow safe interrupt allocation
5652 - set rq_no_resend for cancel requests (2432)
5653 - recalculate ptlrpcd timeout after resend (2494)
5654 - call vfs_rmdir when removing pending directories (2368)
5655 - fix renaming a file to itself (2429)
5656 - lmc creates a default one-stripe lov (2454)
5657 - expand procfs space to handle large clusters (2326)
5658 - increase UML stack to avoid overflow
5659 - update lconf's list of debug and subsystem masks
5660 - fix lfs find --obd (2510)
5661 - /proc tunable for disabling filter read caching (2591)
5662 - stop rpm packages from altering slapd.conf (2301)
5663 - disable nagle in the socknal under 0conf (2578)
5664 - choose mds inode size based on stripe count (2572)
5665 - fix kernel-source rpm problems (2516)
5667 - add --disable-doc to avoid pdf generation (2421)
5668 - update documentation, tests, type-os, comments
5669 - avoid format warnings on ia64
5670 - remove the TOE NAL
5671 - tiny code cleanups by removing unused fields
5673 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5676 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5677 - remove the most blatant lies from BUILDING file (2371)
5678 - change default debug level to reasonable production setting
5679 - reduce client side cache size to reduce cache flush time
5680 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5681 - make TCP zerocopy and pinger support enabled by default (2476)
5682 - sync writes completed after process exits caused crashes (2319)
5683 - maintain correct mount count on the MDS (2356)
5684 - backout 1557, because 2316 wasn't really fixed
5685 - better file I/O statistics gathering in /proc
5686 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5687 - another kernel patch to fix zero-copy TCP function export
5688 - don't take duplicate lock when processing re-sent getattr (2420)
5689 - lctl uses obd_self_export instead of creating new conn (2353)
5690 - MDS/OST recovery case which requires object creation asserted (2425)
5691 - move lfs from /usr/sbin to /usr/bin in packages
5692 - fix race between mds_client_add and mds_client_free (2417)
5693 - use kmalloc instead of slabs in portals (2430)
5694 - don't create duplicate records when a failover MDS is present (2442)
5695 - remove unnecessary mount age check (2332)
5696 - don't remove directory inodes from locks prematurely (2451)
5697 - don't break if MDS service name is the same as hostname (2103)
5698 - fix races in client write RPC generation when cache full (2482)
5700 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5703 - remove now-unused request->rq_obd (278)
5704 - if an allocation fails, print out how much memory we've used (1933)
5705 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5706 - add the "configurable stack size" patch to most series files (1256)
5707 - ability to write large log records, for 100+ OST configs (2306)
5708 - fix NULL deref when filter_prep fails (2314)
5709 - fix operator precedence error in filter_sync
5710 - dynamic allocation of socknal TX descriptors (2315)
5711 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5712 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5713 - ensure that CWARN messages in Portals always get to the syslog
5714 - __init/__exit are not for prototype decls (ldlm_init/exit)
5715 - x86-64 compile warning fixes
5716 - fix gateway LMC keyword conflict (2318)
5717 - fix MDS lock inversions in getattr/reint paths (1844)
5718 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5719 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5720 - DLM race condition prevented some lock evictions (2328)
5721 - ENOMEM detection and retry on socknal sends (2230)
5722 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5723 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5724 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5725 - better error messages when a client is rejected during recovery (1505)
5726 - avoid cancelling locks which were never granted, after failure (2330)
5727 - fix i_sem/journal inversion in mds_client_add (2333)
5728 - fix truncate/getattr lock cycle deadlock (2334)
5729 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5730 - fix two rare exit paths which could leak an l_lock() ref (2321)
5731 - fencepost error in MDS/OST orphan recovery (2226)
5732 - make log record alignment 8 bytes (1988)
5733 - lstripe now fails when requested offset > ost_count (2237)
5734 - ensure that all kernel series have a complete list.h (1607)
5735 - fix crashes in special-file operations (2316)
5736 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5738 - allow configurable automake binary, for testing new versions
5739 - small update to the lfs documentation
5741 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5743 * fix negative export reference count in fsfilt_sync (2312)
5745 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5746 * release candidate 0.9.1
5748 - orphans are moved into the PENDING directory for possible recovery
5749 - replayed opens now open by fid for orphan/rename safety (1042)
5750 - last close of an orphan inode generates a transno (683)
5751 - chdir() and mount() now pin the directory entry (1020)
5752 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5753 - discard very old requests without processing them (1502)
5754 - remove some common, well-understood CERRORs (1505)
5755 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5756 - clear "grant" flags in OST replies until OST grant code lands (1644)
5757 - fix read performance by not clobbering i_blksize on client (1598)
5758 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5759 - make LRU size a /proc tunable, clears locks when reduced (707)
5760 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5761 - server locks take references on exports now (1558)
5762 - build fixes for 2.4.20-rh trees (1663)
5763 - return an error from lov_create if all OSCs are inactive (1751)
5764 - fix import levels when a reconnect happens without a timeout (1597)
5765 - exit early from mds_open if we get a lookup error (1749)
5766 - partial page read at EOF wouldn't wait for disk before sending (1642)
5767 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5768 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5769 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5770 - write/truncate lock inversion (1639)
5771 - Don't auto-load obdclass, portals modules during cleanup (1495)
5772 - fix timestamps from jumping to "now" (1763)
5773 - extra journal assertions (1648)
5774 - add an extra multiunlink test (1771)
5775 - fix read_record/write_record API (1776)
5776 - fix leak of offset_extent, possible incorrect i_size later (1772)
5777 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5778 - Debugging for the kqswnal_get_idle_tx problems (1820)
5779 - Allow recovery to be attempted multiple times (1536)
5780 - Write out MDS last_rcvd file after it is first created (1600)
5781 - Fix tx_descriptor leak in failed transmit situations (1827)
5782 - ext3 journaling fixes for assertion failure after IO error (1871)
5783 - class_export_put() on freed export after completion AST error (1896)
5784 - Fix revalidate looping in VFS (1322)
5785 - Don't access a freed export during MDS_REINT timeout (1521)
5786 - Add open-unlink recovery support on the MDS (1673,1764)
5787 - Return an error if no MDS data was read from last_rcvd (1946)
5788 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5789 - Don't setup a disk device that doesn't match exported UUID (317)
5790 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5791 - avoid committing NULL handle in force close
5792 - local.sh is now a one-stripe LOV configuration
5793 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5794 - fix ext3 htree duplicate directory entry corruption (1516)
5795 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5796 - update to new LOV EA format (2097)
5797 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5799 2003-06-15 Phil Schwan <phil@clusterfs.com>
5802 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5803 - per-import/export recovery handling (958, 931, 959)
5804 - multiple last-rcvd slots, for serving multiple FSes (949)
5805 - connections are again shared between multiple imp/exports (963, 964)
5806 - "umount -f" would hang if any requests needed to be sent (393, 978)
5807 - avoid pinning large req buffer by copying for queued messages (989)
5808 - add "uuid" to "lctl device" command to help upcalls (991)
5809 - "open" RPCs with transnos would confuse recovery counters (1037)
5810 - do proper endian conversion of all wire messages (288, 340, 891)
5811 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5812 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5813 - fix ptlrpc_connection leak in target_handle_connect (1174)
5814 - fix import refcounting bug in OST and MDS cleanup (1134)
5815 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5816 - fix very unlikely obd_types race condition (501)
5817 - remove osc_open hack for echo_client (1187)
5818 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5819 - a failure in read_inode2 leads to deadlock (1139)
5820 - cancel ack-locks as soon as transaction is committed (1072)
5821 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5822 - make sure to commitrw after any preprw to avoid deadlock (1162)
5823 - failing to execute a file in a lustre FS would lock inode (1203)
5824 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5825 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5826 - obd_uuid2tgt would walk off the end of the list (1255)
5827 - on IA64 the osc would give portals incorrect bulk size (1258)
5828 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5829 - fix lock inversion caused by new llite matching code (1282)
5830 - limit the number of dirty pages on a client to 10MB (1286)
5831 - timed out locks were not being corrected cancelled (1289)
5832 - fix O_DIRECT above 4GB on IA-32 (1292)
5833 * major user-visible changes
5834 - fail out/fail over policy now controlled by the upcall (993)
5836 - add OBD_PING to check server availability and failure (954)
5837 - lustre messages are now sent in sending host order (288, 340, 891)
5838 - add eadatalen to MDS getattr reply (340)
5839 - OST read replies may contain second buffer, with per-page status (593)
5841 2003-03-11 Phil Schwan <phil@clusterfs.com>
5844 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5845 - failed open()s could cause deadlock; fixed (867, 869)
5846 - stop cancelling OST locks when files are closed (481)
5847 - overlapping XID spaces caused network corruption (851, 853)
5848 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5849 - setattr_raw added, to do single-RPC, server-side setattrs
5850 - lmc/lconf syntax change for OST UUIDs
5851 - fix crashy race condition between ptlrpc_free_req and osc_close
5852 - don't use request in mdc_enqueue if we hit a timeout (889)
5853 - don't set the inode i_size for regular files from the MDS (896)
5854 - handle out of order completion AST (842)
5855 - don't LBUG if a lock request times out after receiving AST (913)
5856 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5857 - if a bad lock AST arrives, send an error instead of dropping entirely
5858 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5859 - fix leak in bulk IO when only partially completed (899, 900, 926)
5860 - fix O_DIRECT for ia64 (55)
5861 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5862 - C-z after timeout could hang a process forever; fixed (977)
5864 - client-side I/O cache (678, 924, 929, 941, 970)
5866 - READPAGE and SETATTRs which don't take server-side locks get
5869 2003-02-11 Phil Schwan <phil@clusterfs.com>
5872 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5873 - Fix fsfilt_extN_readpage() to read a full page of directory
5874 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5875 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5876 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5877 - Most intent operations were removed, in favour of a new RPC mode
5878 that does a single RPC to the server and bypasses most of the VFS
5879 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5880 - Aggressively cancel local locks on DLM servers
5881 - mds_reint_unlink sends EA to the client if it's the last nlink.
5882 client uses that EA to unlink OST objects.
5883 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5884 - recursive symlinks were fixed (439)
5885 - fixed NULL deref in DEBUG_REQ
5886 - filter_update_lastobjid no longer calls sync, which annoyed extN
5887 - fixed multi-client small-writes to a single file problem (445)
5888 - fixed mtime updates during file writes (607)
5889 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5890 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5891 - fixed Linux OST size reporting problem (444, 656)
5892 - OST now updates object mtime with writes or setattr (607, 619)
5893 - client verifies file size before zeroing page past EOF (445)
5894 - OST now writes last allocated objid to disk with allocation (108)
5895 - LOV on echo now works (409)
5897 - mds_reint_unlink sends a new buffer, with the EA included. this
5898 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5899 set if a regular file was being unlinked, and it was the last link
5900 - use PtlGet from the target for bulk writes (315)
5901 - OST now updates object mtime with writes or setattr (607, 619)
5902 - LDLM now has a grant-time callback to revalidate locked items, if
5904 - Many MDS operations were reorganized to combat race conditions
5906 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5907 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5909 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5912 - Fully reactivate OST imports after reconnection (512, others)
5913 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5914 - More graceful error handling for truncating on dead OST (515)
5915 - Don't error out unless we're actually accessing dead stripes (474)
5916 - Fix garbage sizes when stripes are missing (410)
5917 - LRU counters were broken, causing constant lock purge (433, 432)
5918 - garbage on read from stripes with failed OSTs (441)
5919 - mark OSCs as active before reconnecting during recovery (438)
5920 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5921 - lfind did not preserve OST order in output (443)
5922 - symlinks cause hung clients, incorrect data (439)
5923 - stop dereferencing request after dropping refcount (457)
5924 - don't LASSERT(spin_is_locked) on non-SMP (455)
5925 - fixes for many rename() bugs
5926 - fstat didn't correctly synchronize attributes (399)
5927 - server must handle lock cancellation during blocking AST prep (487)
5928 - bulk descriptors were free()d too soon (511)
5929 - fix paths in lconf, which would load incorrect modules (451, 507)
5930 - fix confusing lconf 'host not found' error message (386)
5931 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5932 - fix race condition in mdc_blocking_ast() for inode access (526)
5933 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5934 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5935 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5936 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5937 - fix lstripe in several ways (kernel side) (527)
5938 - fix request leak in ldlm_cli_enqueue (262)
5939 - incorrect OSC was marked inactive after OST failure
5940 - call mds_fs_cleanup before unmounting filesystem (524)
5941 - fix races between taking ns_lock and ldlm_lock_change_resource
5942 - fix races updating LOV export open file list
5943 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5944 - fix recovery NULL deref in ldlm_cli_cancel_unused
5945 - fix some DLM races by using new hash table for lock handles (419)
5946 - permit the client to specify desired inodes, at replay
5947 - duplicate requests when we queue them for replay reintegration
5948 - fix last_rcvd offset calculation
5949 - sync after each recovered transaction, so we always make progress
5950 - never, not always, ERESTART requests without transnos
5951 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5953 - skip replay if the MDS says that the client is already connected
5954 - don't check for a recovery-enabled export to match lctl's UUID
5955 - don't INC_USE_COUNT for phantom exports
5956 - don't crash when cleaning up phantom exports (567)
5957 - don't double-finish or set replay data for errored mdc_open requests
5958 - abort requests when they time out, so we don't get old replies
5959 - send/receive replies for AST messages again
5960 - if the client says that it doesn't have the lock, cancel it on the
5962 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5963 mark it as destroyed, it will all work out when decref is called
5964 - fix module use counts (22, 581)
5966 - ASTs now expect a reply (server cancels lock on error reply)
5968 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5971 - fix many simultaneous client startup (392)
5972 - fix dentry->d_it clobbering
5973 - credentials weren't being shipped for readdir/getattr operations
5974 - remove invalid assertions triggered during some concurrent MD
5976 - proper Lustre versions added (336, 389)
5977 - fix memory leak for create error case (398)
5978 - fix LOV locking bug that would get cli/srv out of sync
5979 - fix echo client over LOV (409)
5980 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5981 - fix double-O_EXCL intent crash (424)
5982 - avoid sending multiple lock CANCELs (352)
5984 - MDS can do multi-client recovery (modulo bugs in new code)
5986 - many updates, edits, cleanups
5988 2002-11-18 Phil Schwan <phil@clusterfs.com>
5991 - fix null d_it dereference (346)
5992 - fix full OST/dbench hang (333)
5993 - fix permission problem with file removal (286)
5994 - fix removal of OSCs from LOV when they fail
5995 - fix NULL deref during bulk timeout (214)
5996 - fix problems related to multiple filesystems on one MDS (241)
5997 - fixed serious subtle metadata locking bugs
5998 - free locks on clients when inodes are removed due to memory
6000 - fix inode pointer in lock data (285)
6001 - partial support for multiple MDS on a single host (241)
6002 - data locks weren't cancelled at clear_inode time (290, 311)
6003 - intent locks could lead to unbounded lock growth (205)
6004 - added a maximum lock count, an LRU list, and a flusher
6005 - fix multiple rename (365)
6006 - properly abstracted the echo client
6007 - OSC locked 1 byte too many; fixed
6008 - rewrote brw callback code:
6009 - fixed recovery bugs related to LOVs (306)
6010 - fixed too-many-pages-in-one-write crash (191)
6011 - fixed (again) crash in sync_io_timeout (214)
6012 - probably fixed callback-related race (385)
6014 - Add capability to MDS protocol
6015 - LDLM cancellations and callbacks on different portals
6017 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6020 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6021 - defer open object destruction to close time (601981, 138)
6022 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6023 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6024 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6025 - don't call obd_setattr in ll_file_release for destroyed objects
6026 * protocol change to lustre_msg: move |version| and add |flags|
6027 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6028 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6029 * added replay of create, unlink, link and rename operations during
6030 MDS failover; recovery should be much more robust now
6031 * remove failed OSCs from LOVs (only lov_create uses this so far)
6032 * the lustre-HOWTO was brought (more) up to date (582544)
6034 2002-10-23 Phil Schwan <phil@clusterfs.com>
6037 - in-use dentries weren't being reused properly (617851)
6038 - prevent multiple LDLM setup (599178)
6039 - fix LOV size calculations for truncate (617853)
6040 - fix client handling of MDS intent errors (POSIX)
6041 - fix permission bug in lovstripe.c test (624321)
6042 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6043 - truncate past end of file could corrupt data
6044 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6045 - a race in recovery could return ETIMEDOUT to apps (623947)
6046 - building outside the source directory was fixed
6047 * the lustre-HOWTO was brought (more) up to date (582544)
6048 * major progress was made on recovery functionality
6050 2002-10-10 Phil Schwan <phil@clusterfs.com>
6053 - recovery deadlock fix
6054 - rm -rf causes LBUG fix (617817)
6055 - file open by multiple tasks fix (618962)
6056 - directory permissions bugs (602707 and 620007)
6057 - journal_stop fixed with locking (611313)
6058 - O_APPEND failures resolved (618273, perhaps 614459)
6059 - lconf PATH fix (619770)
6060 - IA64 build fix (621450)
6061 - RPC buffer sizes scale with amount of memory
6063 2002-10-01 Phil Schwan <phil@clusterfs.com>
6066 - locks would be cancelled without throwing away data pages,
6067 resulting in inconsistent data (605627)
6068 - inode attributes were not always being refreshed (605627, 612449)
6069 - lconf now continues to cleanup after lctl reports an error
6070 - MDS now enforces user permissions (602707)
6071 - lprocfs cleanup fixed, but not yet enabled (614157)
6072 - fixed infinite server hang, should a client not respond to an AST
6073 - avoid going into recovery if user calls readlink() with a buffer
6074 that's too small (613941)
6075 - AST RPCs no longer require replies (614867) -- this may be changed
6076 - don't crash server if client sends an IOV that's too big (611336)
6077 - fixed lock conversion deadlock (611892)
6078 - fixed the following of symlinks (614622)
6079 * recovery: the server can remove locks from a client that dies, other
6080 clients can make progress
6081 * more extN patch fixes
6082 * compile-time configurable ptlrpc buffer allocations
6084 - collaborative read cache document
6085 - Lustre Lite Performance CDR document-in-progress
6087 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6090 - fix typo in patch-2.4.18
6092 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6095 - clear ptlrpc request each time in handle_incoming_request()
6096 - unlink of files now destroys the object on the OST
6098 2002-09-19 Peter Braam <braam@clusterfs.com>
6100 * add hard link support
6101 * change obdfile creation method
6102 * kernel patch changed
6104 2002-09-19 Peter Braam <braam@clusterfs.com>
6107 - stack overflow bug in extN fixed
6109 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6111 * documentation updates
6112 - add man pages for config tools
6113 - update tests/README to describe testing with new config tools
6114 - finish metadata API descriptions
6115 * bug fixes and cleanups
6116 - statfs workaround for 16TB limit
6117 - LOV stripe allocation improved, can stripe on subset of OSTs
6118 - LOV file size/IO offset was wrong for files > 4GB in size
6119 - object EA data was being dropped, caused files to be unreadable
6120 - memory overflow with non-LOV OST caused memory corruption
6121 - fixed regression tests to work with new config tools, obdfilter
6122 - fixed bug when directory size became larger than 1 block
6123 - fixed bug (for single client case) when PWD was deleted
6124 - invalidate local directory pages when doing intent-based ops
6125 - avoid LDLM oops when lock callback contained bad data
6127 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6129 * documentation updates
6130 * bug fixes and cleanups
6131 - configuration tools
6134 - 64-bit compile warnings
6135 - 64-bit internal statfs data
6137 * test_brw on persistent OST devices
6139 * lprocfs (disabled)
6141 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6143 * documentation updates
6144 * bug fixes and cleanups
6145 * configuration tools
6147 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6150 * many small fixes to 0_5_4
6151 * io/network handling
6152 * thinkos in MDS operations
6154 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6157 * crucial basic fixes to 0.5.3
6158 * IOR, Iozone work over Elan
6161 2002-08-07 Phil Schwan <phil@clusterfs.com>
6162 * version 0_5_3, our first alpha
6163 * we use the new Portals iovs
6164 * documentation updates
6165 * bug fixes and cleanups
6166 * small changes in the DLM wire protocol
6168 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6169 * version 0_5_1 with some initial stability,
6170 * locking on MD and file I/O.
6171 * documentation updates
6172 * several bug fixes since 0.5.0
6173 * small changes in wire protocol
6175 2002-07-18 Phil Schwan <phil@clusterfs.com>
6177 * delivered as Lustre Light Alpha
6178 * fixed a crash after handling invalid MDS requests
6179 * fixed directory pages for architectures with non-4k pages sizes
6181 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6182 * release version v0_4_4
6183 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6185 * Fixes a number of interruption problems with OST operations.
6186 * Update documentation for portals header changes
6187 * Move all wire protocol structs/defines to lustre_idl.h
6188 * Fixes symlink length bug.
6189 * Add tcpdump to repository.
6191 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6192 * release version v0_4_3
6193 * Fixes statfs for inodes on extN.
6194 * Fixes bug in runtests which would delete /etc/hosts.
6195 * Use 64-bit object IDs wherever possible (not into VFS though)
6196 Remove ost_get_info, which is unused by lustre, and out of date.
6198 2002-07-03 Peter Braam <braam@clusterfs.com>
6199 * release version v0_4_2 Fixes a lookup error (type not passed)
6200 * move forward to head of Portals
6201 * move forward to latest Lustre kernel
6203 2002-06-25 Peter Braam <braam@clusterfs.com>
6204 * release version v0_4_1. Hopefully stable on single node use.