1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.7.1.el5 (RHEL 5),
6 2.6.18-128.7.1.el5 (OEL 5),
8 2.6.27 vanilla (kernel.org).
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.27 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.41.6.sun1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
14 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
15 removed cwd "./" (refer to Bugzilla 14399).
20 Description: drop join file stuff.
22 Severity : enhancement
24 Description: Add 2.6.27 and SLES11 (patchless client) support
26 Severity : enhancement
28 Description: Adjust locks' extents on their first enqueue, so that at the time
29 they get granted, there is no need for another pass through the
30 queues since they are already shaped into the proper forms.
34 Description: Fix in ptlrpc_expire_one_request() to print the signed time
37 Severity : enhancement
39 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
42 Severity : enhancement
44 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
46 Severity : enhancement
48 Description: correctly handle big reply message.
49 Details : send LNet event if reply is bigger then buffer and adjust this buffer
54 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
56 Severity : enhancement
58 Description: Add support for OEL5.
62 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
66 Description: client could not reconnect to OST because of an active request.
67 Details : The client now retries to reconnect to the same server, if a
68 connect request failed with EBUSY or -EAGAIN. Test case 35b
69 conf-sanity.sh has been created.
73 Description: Don't allow make backward step on assignin osc next id.
74 Details : race between allocation next id and ll_sync thread can be cause
75 of set wrong osc next id and can be kill valid ost objects.
79 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
83 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
87 Description: Error handling in osc_statfs_interpret() has been improved.
88 Details : Check in osc_statfs_interpret() for EBADR.
92 Description: Do not send statfs() requests to OSTs disabled by administrator.
93 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
95 Severity : enhancement
97 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
99 Severity : enhancement
101 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
103 Severity : enhancement
105 Description: Enhancement for lfs(1) command to use numeric uid/gid.
109 Description: Improved cleanup procedure for conf-sanity test_46a.
113 Description: Do not start bulk transfer, if client tried to reconnect.
114 Details : Do not start bulk transfer, if client tried to reconnect,
115 but failed due to active rpcs. Check for exp_abort_active_req
116 has been implemented.
120 Description: Prevent inconsistences between linux and lustre mount structures.
121 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
122 Make the sleep interruptible.
124 Severity : enhancement
126 Description: provide server to client comms path
127 Details : server to client communications path via new LDLM_SET_INFO rpc
131 Description: truncate starts GFP_FS allocation under transaction causing deadlock
132 Details : ldiskfs_truncate calls grab_cache_page which may start page
133 allocation under an open transaction. This may lead to
134 calling prune_icache with consequent lustre reentrance.
138 Description: send correctly lsm on open replay
139 Details : MDS is trust to LSM size on replay open, but client can set wrong size
144 Description: enable client interface failover
145 Details : When a child reconnects from another NID, properly update export
146 nid hash position and ldlm reverse import.
148 Severity : enhancement
150 Description: Communicate OST degraded/readonly state via statfs to MDS
151 Details : Flags in the statfs returned from OSTs indicate whether the
152 OST is in a degraded RAID state, or if the filesystem has
153 turned read-only after a filesystem error is detected.
158 Description: don't panic if EPROTO was hit when reading symlink
159 Details : correctly handling request reference in error cases.
161 Severity : enhancement
163 Description: Add LustreNetLink, a kernel-userspace communcation path.
165 Severity : enhancement
167 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
172 Description: access to llog context before init.
173 Details : move handling CATALOGS file at osc layer and forbid access to llog
177 Frequency : rare, only when client is writing to OST on local node
179 Description: Avoid deadlock for local client writes
180 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
181 memory freeing context. This allows OST threads to set the
182 PF_MEMALLOC flag on task structures in order to allocate memory
183 from reserved pools and complete IO.
184 Use GFP_HIGHUSER for OST allocations for non-local client writes,
185 so that the OST threads generate memory pressure and allow
186 inactive pages to be reclaimed.
188 Severity : enhancement
190 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
195 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
196 Details : move ldlm namespace creation in setup phase to avoid grab
197 _lprocfs_lock with cli_sem held.
201 Description: Temporarily disable grant shrink.
202 Details : Disable the feature for debugging.
205 Frequency : only when formatting test filesystems
207 Description: Unable to run several mkfs.lustre on loop devices at the same time
208 Details : mkfs.lustre returns error 256 on the concurrent loop devices
209 formatting. The solution is to proper handle the error.
211 Severity : enhancement
213 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
215 Severity : enhancement
217 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
222 Description: resolve race between obd_disconnect and class_disconnect_exports
223 Details : if obd_disconnect will be called to already disconnected export he
224 forget release one reference and osc module can't unloaded.
226 Severity : enhancement
228 Description: Allow tuning service thread via /proc
229 Details : For each service a new
230 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
231 created that can be used to set min/max thread counts, and get the
232 current number of running threads.
237 Descriptoin: don't return error if have partially created objects for file.
238 Details : lov_update_create_set uses set->set_success as index for created
239 objects, so if some requests will be failed, they will have hole
240 at end of array and qos_shrink_lsm for allocate correct lsm.
242 Severity : enhancement
244 Description: Update OFED support to 1.4.1
249 Description: Reduce small size read RPC
250 Details : Set read-ahead limit for every file and only do read-ahead when
251 available read-ahead pages are bigger than 1M to avoid small size
254 Severity : enhancement
256 Description: move AT tunable parameters for more consistent usage
257 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
259 Severity : enhancement
261 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
262 Details : allow skip disconnected ost for send statfs request and hide error
268 Description: fix racy locking of mballoc block bitmaps causing BUG
269 Details : The locking of the mballoc buddy bitmap and the in-memory
270 block bitmap was using two different spin locks in some
271 cases. This made it possible to incorrectly access the
272 mballoc bitmap while another process was modifying it,
273 causing a sanity assertion to fail. While no on-disk corruption
274 was reported, there was some risk of this happening.
277 Frequency : rare, on llog test 6
279 Descriptoin: don't allow connect to already connected import
280 Details : allowing connect to already connected import is hide connecting problem.
285 Descriptoin: allow kill process which wait statahead result
286 Details : for some reasons 'ls' can stick in waiting result from statahead,
287 in this case need way for kill this process.
289 Severity : enhancement
291 Description: Add state history info file, enhance import info file
292 Details : Track import connection state changes in a new osc/mdc proc file;
293 add overview-type data to the osc/mdc import proc file.
295 Severity : enhancement
297 Description: MDS create should not wait for statfs RPC while holding DLM lock.
299 Severity : enhancement
301 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
306 Descriptoin: OST grant too much space to client even there are not enough space.
307 Details : Client will shrink its grant cache to OST if there are no write
308 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
309 this grant cache if there are already not enough avaible space
310 (left_space < total_clients * 32M).
313 Frequency : start MDS on uncleanly shutdowned MDS device
315 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
316 Details : stay in waiting mds<>ost recovery finished produce random bugs
317 due race between two ll_sync thread for one lov target. send
318 ACTIVATE event only if connect realy finished and import have
322 Frequency : rare, connect and disconnect target at same time
324 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
325 Details : don't call obd_disconnect under lov_lock. this long time
326 operation and can block ptlrpcd which answer to connect request.
331 Descriptoin: don't lose wakeup for imp_recovery_waitq
332 Details : recover_import_no_retry or invalidate_import and import_close can
333 both sleep on imp_recovery_waitq, but we was send only one wakeup
337 Frequency : always with long access acl
339 Descriptoin: mds can't pack reply with long acl.
340 Details : mds don't control size of acl but they limited by reint/getattr
343 Severity : enhancement
345 Description: Update to SLES10 kernel-2.6.16.60-0.33.
347 Severity : enhancement
349 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
352 Frequency : start MDS on uncleanly shutdowned MDS device
354 Descriptoin: aborting recovery hang on MDS
355 Details : don't throttle destroy RPCs for the MDT.
358 Frequency : on remount
360 Description: external journal device not working after the remount
361 Details : clear dev_rdonly flag for external journal devices in
367 Description: shutdown vs evict race
368 Details : client_disconnect_export vs connect request race.
369 if client will evicted at this time - we start invalidate
370 thread without referece to import and import can be freed
374 Frequency : rare, need acl's on inode.
376 Description: client can't handle ost additional correctly
377 Details : if ost was added after client connected to mds client can have
378 hit lnet_try_match_md ... to big messages to wide striped files.
379 in this case need teach client to handle config events about add
380 lov target and update client max ea size at that event.
382 Severity : enhancement
384 Description: Changelogs
385 Details : Changelogs are a lightweight mechanism to track filesystem
386 metadata and namespace changes. The changelog is recorded
387 permanently on the MDTs, and is periodically "consumed" / purged
388 when records are no longer needed.
390 Severity : enhancement
392 Description: compact fld format with extents
393 Details : Store range of seq rather than every seq in FLD. Seq
394 controller update FLD rather than clients. In Case of CMD, mdt0
395 has FLD, all other metadata server act as non persistent proxy
396 for FLD queries and cache fld entries in fld cache.
401 Description: don't skip ost target if they assigned to file
402 Details : Drop slow OSCs if we can, but not for requested start idx.
403 This means "if OSC is slow and it is not the requested
404 start OST, then it can be skipped, otherwise skip it only
405 if it is inactive/recovering/out-of-space.
409 Description: more cleanup in mds_lov
410 Details : not send LOV EA under replay, we can't know about they size at this
411 time. Don't allow client connect to mds before any ost connected,
412 for avoid problems with LOV EA size and returning EIO to client.
414 Severity : enhancement
416 Description: Interoperability at server side (Disk interoperability)
418 Severity : enhancement
420 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
422 Severity : enhancement
424 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
426 Severity : enhancement
428 Description: New client IO stack (CLIO).
430 Severity : enhancement
432 Description: Commit on sharing. Eliminate inter-client dependencies between
433 uncommitted transactions by doing transaction commits.
434 Thereby clients may recovery independently.
437 Frequency : Create a symlink file with a very long name
439 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
440 Details : If there is no extra space in the request for early cancels,
441 ldlm_req_handles_avail() returns 0 instead of a negative value.
443 Severity : enhancement
445 Description: Add /proc entry for import status
446 Details : The mdc, osc, and mgc import directories now have
447 an import directory that contains useful import data for debugging
450 Severity : enhancement
452 Description: Re-disable certain /proc logging
453 Details : Enable and disable client's offset_stats, extents_stats and
454 extents_stats_per_process stats logging on the fly.
457 Frequency : Only on FC kernels 2.6.22+
459 Description: oops in statahead
460 Details : Do not drop reference count for the dentry from VFS when lookup,
461 VFS will do that by itself.
463 Severity : enhancement
465 Description: Generic /proc file permissions
466 Details : Set /Proc file permissions in a more generic way to enable non-
467 root users operate on some /proc files.
471 Description: Hitting mdc_commit_close() ASSERTION
472 Details : Properly handle request reference release in
473 ll_release_openhandle().
477 Description: quota recovery deadlock during mds failover
478 Details : This patch includes att18982, att18236, att18237 in bz14840.
480 1. fix osts hang when mds does failover with quotaon
481 2. prevent watchdog storm when osts threads wait for the
486 Frequency : only patchless client
487 Description: add workaround for race between add/remove dentry from hash
489 Severity : enhancement
491 Description: Allow OST glimpses to return PW locks
495 Description: LBUG when llog conf file is full
496 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
500 Description: Prevent import from entering FULL state when server in recovery
504 Description: service mount cannot take device name with ":"
505 Details : Only when device name contains ":/" will mount treat it as
511 Description: replace ptlrpcd with the statahead thread to interpret the async
512 statahead RPC callback
516 Frequency : on recovery
517 Description: I/O failures after umount during fail back
518 Details : if client reconnected to restarted server we need join to recovery
519 instead of find server handler is changed and process self
520 eviction with cancel all locks.
522 Severity : enhancement
524 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
526 Severity : enhancement
528 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
530 Severity : enhancement
532 Description: Upcall on Lustre log has been dumped
533 Details : Allow for a user mode script to be called once a Lustre log has
534 been dumped. It passes the filename of the dumped log to the
535 script, the location of the script can be specified via
536 /proc/sys/lnet/debug_log_upcall.
541 Description: avoid idr_remove called for id which is not allocated.
542 Details : Move assigment s_dev for clustered nfs to end of initialization,
543 to avoid problem with error handling.
548 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
549 Details : When connection is reused this not moved from CONN_UNUSED_HASH
550 into CONN_USED_HASH and this prodice warning when put connection
551 again in unused hash.
554 Severity : enhancement
556 Description: File striping can now be set to use an arbitrary pool of OSTs.
558 Severity : enhancement
560 Description: Export bytes_read/bytes_write count on OSC/OST.
564 Description: Early reply size mismatch, MGC loses connection
565 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
566 the connect flags are properly negotiated.
571 Description: quotas are not honored with O_DIRECT
572 Details : all writes with the flag O_DIRECT will use grants which leads to
573 this problem. Now using OBD_BRW_SYNC to guard this.
577 Description: add quota statistics
578 Details : 1. sort out quota proc entries and proc code.
579 2. add quota statistics
581 Severity : enhancement
583 Description: enable quota support for HEAD.
587 Description: Properly propagate oinfo flags from lov to osc for statfs
588 Details : restore missing copy oi_flags to lov requests.
590 Severity : enhancement
592 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
593 : l_getgroups(8), lst(8), routerstat(8)
594 Details : included man pages for llobdstat(8), llstat(8),
595 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
597 Severity : enhancement
599 Description: configure's --enable-quota should check the
600 : kernel .config for CONFIG_QUOTA
601 Details : configure is terminated if --enable-quota is passed but
602 : no quota support is in kernel
606 Description: Remove portals compatibility
607 Details : Remove portals compatibility, not interoperable with releases
612 Description: Resolve device initialization race
613 Details : Prevent proc handler from accessing devices added to the
614 obd_devs array but yet be intialized.
616 Severity : enhancement
618 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
620 Severity : enhancement
622 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
627 Description: Using wrong pointer in osc_brw_prep_request
628 Details : Access to array[-1] can produce panic if kernel compiled with
629 CONFIG_PAGE_ALLOC enabled
633 Description: Client runs out of low memory
634 Details : Consider only lowmem when counting initial number of llap pages
638 Description: *optional* service tags registration
639 Details : if the "service tags" package is installed on a Lustre node
640 When the filesystem is mounted, a local-node service tag will
641 be created. See http://inventory.sun.com/ for more information
642 about the Service Tags asset management system.
646 Description: Kernel BUG tries to release flock
647 Details : Lustre does not destroy flock lock before last reference goes
648 away. So always drop flock locks when client is evicted and
649 perform unlock regardless of successfulness of speaking to MDS.
653 Description: add refcount protection for osc callbacks, avoid panic on shutdown
657 Description: sanity test 65a fails if stripecount of -1 is set
658 Details : handle -1 striping on filesystem in ll_dirstripe_verify
663 Description: ASSERTION(CheckWriteback(page,cmd)) failed
664 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
667 Severity : enhancement
669 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
673 Description: do not process already freed flock
674 Details : flock can possibly be freed by another thread before it reaches
675 to ldlm_flock_completion_ast.
679 Description: LBUG during stress test
680 Details : Need properly lock accesses the flock deadlock detection list.
684 Description: oops in page fault handler
685 Details : kernel page fault handler can return two special 'pages' in error
686 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
690 Description: timeout with invalidate import.
691 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
692 should be handled by ptlrpcd. This produce long age waiting and
693 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
695 Severity : enhancement
697 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
701 Description: enable MGS and MDT services start separately
702 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
703 a co-located MGS without starting the MGS, which is a complement
704 to 'nosvc' mount option.
708 Frequency : after recovery
709 Description: precreate to many object's after del orphan.
710 Details : del orphan st in oscc last_id == next_id and this triger growing
711 count of precreated objects. Set flag LOW to skip increase count
712 of precreated objects.
716 Frequency : rare, on clear nid stats
717 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
718 Details : when clean nid stats sometimes try destroy live entry,
719 and this produce panic in free.
723 Description: Stack overflow during MDS log replay
724 ease stack pressure by using a thread dealing llog_process.
728 Description: wait until IO finished before start new when do lock cancel.
729 Details : VM protocol want old IO finished before start new, in this case
730 need wait until PG_writeback is cleared until check dirty flag and
731 call writepages in lock cancel callback.
733 Severity : enhancement
735 Description: using special macro for print time and cleanup in includes.
739 Description: mds_mfd_close() ASSERTION(rc == 0)
740 Details : In mds_mfd_close(), we need protect inode's writecount change
741 within its orphan write semaphore to prevent possible races.
745 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
749 Frequency : rare, on shutdown ost
750 Description: don't hit live lock with umount ost.
751 Details : shrink_dcache_parent can be in long loop with destroy dentries,
752 use shrink_dcache_sb instead.
756 Description: don't panic with use echo client
757 Details : echo client pass NULL as client nid pointer and this produce null
762 Description: fix build on ppc32
763 Details : compile code with -m64 flag produce wrong object file for ppc32.
767 Description: add message levels for liblustreapi
771 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
772 Details : Move the 'good_osts' check before the 'total_bavail' check. This
773 will result in an -EAGAIN and in the exit call path we call
774 alloc_rr() which will with increasing aggressiveness attempt to
775 aquire precreated objects on the minimum number of required OSCs.
779 Description: Use old size assignment to avoid deadlock
780 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
781 the scheduling while holding a spinlock. We do not need locking
782 for size in ll_update_inode() because size is only updated from
783 the MDS for directories or files without objects, so there is no
784 other place to do the update, and concurrent access to such inodes
785 are protected by the inode lock.
789 Description: resolve "_IOWR redefined" build error on SLES10
793 Description: dump the memory debugging after all modules are unloaded to
794 suppress false negative in conf_sanity test 39
796 Severity : enhancement
798 Description: build kernel-ib packages for OFED 1.3 in our release cycle
803 Description: fix SLES kernel versioning
804 Details : the kernel version for our SLES 10 kernel did not include a "-"
805 before the "smp" at the end. while this was not a problem in
806 general, it did mean that software trying to use the kernel
807 version to try to detect a vendor specific kernel would fail.
808 this was most evident by the OFED build scripts.
812 Description: Don't update lov_desc members until making sure they are valid
813 Details : When updating lov_desc members via proc fs, need fix their
814 validities before doing the real update.
818 Description: don't put request into delay list while invalidate in flight.
819 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
820 import in flight. this produce timeout for invalidate and sometimes
821 can cause stale data.
825 Frequency : on ppc only
826 Description: not convert ost objects for directory because it's not exist.
827 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
831 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
833 Severity : enhancement
835 Description: Update to support 2.6.22.14 vanilla kernel.
839 Frequency : rare, on recovery
840 Description: read procfs can produce deadlock in some situation
841 Details : Holding lprocfs lock which send rpc can produce block for destroy
842 obd objects and this also block reconnect with -EALREADY. This
843 isn't fix all lprocfs bugs - but make it rare.
845 Severity : enhancement
847 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
850 Frequency : frequent on X2 node
852 Description: mdc_set_open_replay_data LBUG
853 Details : Set replay data for requests that are eligible for replay.
857 Description: lustre_mgs: operation 101 on unconnected MGS
858 Details : When MGC is disconnected from MGS long enough, MGS will evict the
859 MGC, and late on MGC cannot successfully connect to MGS and a lot
860 of the error messages complaining that MGS is not connected.
864 Frequency : on network error
865 Description: panic with double free request if network error
866 Details : mdc_finish_enqueue is finish request if any network error ocuring,
867 but it's true only for synchronus enqueue, for async enqueue
868 (via ptlrpcd) this incorrect and ptlrpcd want finish request
871 Severity : enhancement
873 Description: client-side metadata stat-ahead during readdir(directory readahead)
874 Details : perform client-side metadata stat-ahead when the client detects
875 readdir and sequential stat of dir entries therein
878 Frequency : on start mds
880 Description: Implement get_info(last_id) in obdfilter.
883 Frequency : occasional
885 Description: Correctly check stale fid, not start epoch if ost not support SOM
886 Details : open with flag O_CREATE need set old fid in op_fid3 because
887 op_fid2 was overwritten with new generated fid, but mds can answer
888 with one of these two fids and both is not stale. Setattr
889 incorrectly started an epoch and assume will be called
890 done_writing, but without SOM done_writing ever being called.
893 Frequency : rare, depends on device drivers and load
895 Description: MDS or OSS nodes crash due to stack overflow
896 Details : Code changes in 1.8.0 increased the stack usage of some functions.
897 In some cases, in conjunction with device drivers that use a lot
898 of stack the MDS (or possibly OSS) service threads could overflow
899 the stack. One change which was identified to consume additional
900 stack has been reworked to avoid the extra stack usage.
903 Frequency : occasional
905 Description: Do not fail import if osc_interpret_create gets -EAGAIN
906 Details : If osc_interpret_create got -EAGAIN it immediately exits and
907 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
908 oscc_has_objects and see OSC has no objests and call
909 oscc_internal_create to resend create request.
911 Severity : enhancement
913 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
915 Severity : enhancement
917 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
920 Frequency : very rare
922 Description: Fix replay if there is an un-replied request and open
923 Details : In some cases, older replay request will revert the
924 mcd->mcd_last_xid on MDS which is used to record the client's
927 Severity : enhancement
929 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
931 Severity : enhancement
933 Description: Add rhel5 support to HEAD.
935 Serverity : enhancement
937 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
942 Description: Don't allow skipping OSTs if index has been specified.
943 Details : Don't allow skipping OSTs if index has been specified, make
944 locking in internal create lots better.
948 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
949 Details : access to bitfield in structure is always rounded to long
950 and this produce problem with not atomic change any bit.
954 Description: Lustre make rpms failed.
955 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
956 builds Lustre rpms from tarball.
959 Frequency : rare on shutdown ost
961 Description: If llog cancel was not send before clean_exports phase, this can
962 produce deadlock in llog code.
963 Details : If llog thread has last reference to obd and call class_import_put
964 this produce deadlock because llog_cleanup_commit_master wait when
965 last llog_commit_thread exited, but this never success because was
966 called from llog_commit_thread.
970 Description: allow userland application know is lost one of stripes.
971 Details : fill lvb_blocks with error code on ost and return it to
972 application if error flag found.
976 Description: NULL lov_tgts causing MDS oops
977 Details : more safe checks for NULL lov_tgts for avoid oops.
979 Severity : enhancement
981 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
985 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
986 Details : If an OST is down the MDS will hang indefinitely in
987 obd_statfs_rqset() waiting for the statfs data. While for
988 MDS QOS usage of statfs, it should not stuck in waiting.
990 Severity : enhancement
992 Description: remote_acl support
993 Details : Support ACL-based permission check for remote user.
994 Support setfacl/getfacl for remote user with the utils
995 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
996 the system "{s,g}etfacl" utils.
998 Severity : enhancement
1000 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1002 Severity : enhancement
1004 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1008 Description: make mgs_setparam() handle fsname containing dash
1009 Details : fsname containing a dash does not work with lctl conf_param
1011 Severity : enhancement
1013 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1015 Severity : enhancement
1017 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1021 Frequency : rare, at shutdown
1022 Description: access already free / zero obd_namespace.
1023 Details : if client_disconnect_export was called without force flag set,
1024 and exist connect request in flight, this can produce access to
1025 NULL pointer (or already free pointer) when connect_interpret
1026 store ocd flags in obd_namespace.
1030 Frequency : only at startup
1031 Description: not alloc memory with spinlock held.
1032 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1033 if any spinlock held.
1035 Severity : enhancement
1037 Description: make lustre randomly fail allocating memory
1038 Details : Make lustre randomly failed allocating memory for testing purpose.
1040 Severity : enhancement
1042 Description: lost problems with lov objid file
1043 Details : Fixes some scability and access to not inited memory problems
1044 in work with lov objdid file.
1049 Description: lfs find does not continue on file error
1050 Details : Continue other files processing when a file/dir is absent.
1054 Description: Inconsistent usage of lustre_pack_reply()
1055 Details : Standardize the usage of lustre_pack_reply() such that it
1056 always generate a CERROR on failure.
1059 Frequency : occasional
1061 Description: MDS hang or stay in waiting lock
1062 Details : If client receive lock with CBPENDING flag ldlm need send lock
1063 cancel as separate rpc, to avoid situation when cancel request
1064 can't processed due all i/o threads stay in wait lock.
1068 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1072 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1076 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1080 Description: Update to RHEL4 latest kernel.
1082 Severity : enhancement
1084 Description: Build SLES10 patchless client fails
1085 Details : The configure was broken by run ./configure with
1086 --with-linux-obj=.... argument for patchless client. When the
1087 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1088 can't search header adequately. Use absolute path such as
1089 -I($LINUX)/include instead.
1093 Description: interrupt oig_wait produce painc on resend.
1094 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1095 private set, and this produce situation when rq_ptlrpcd_data not
1096 copyed to new allocated request and triggered LBUG on assert
1097 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1098 ptlrpcd set if request is changed and can be safety changed to use
1103 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1104 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1105 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1106 these macros and replace them with their difinitions instead.
1110 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1111 Details : Update which_patch & target file for SLES10 latest kernel.
1114 Description: Remove client patches from SLES 10 kernel.
1115 Details : This causes SLES 10 clients to behave as patchless clients
1116 even on a Lustre-patched (server) kernel.
1118 Severity : enhancement
1120 Description: self-adjustable client's lru lists
1121 Details : use adaptive algorithm for managing client cached locks lru
1122 lists according to current server load, other client's work
1123 pattern, memory activities, etc. Both, server and client
1124 side namespaces provide number of proc tunables for controlling
1127 Severity : enhancement
1129 Description: light-weight GSS support
1130 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1131 and reduce performance overhead.
1133 Severity : enhancement
1135 Description: Linux keyring support
1136 Details : Support using service of Linux keyring for Lustre GSS internal
1137 context refresh/cache mechanism.
1141 Description: Fix errors in lfs documentation
1142 Details : Fixes man pages
1146 Description: don't use GFP_* in generic Lustre code.
1147 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1151 Description: obdclass is limited by single OBD_ALLOC(idarray)
1152 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1156 Description: warnings with build patchless client with vanila 2.6.19 and up
1157 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1158 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1162 Frequency : Only for SLES
1163 Description: sanity_quota fail test_1
1164 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1165 file, which makes TSTID[2] inunique.
1169 Frequency : Only for SLES9
1170 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1171 Details : Update target/ChangeLog/which_patch .
1175 Frequency : Only for test_10 in sanity-quota.sh
1176 Description: a bug in quota test code
1177 Details : checking a test flag in a wrong place
1181 Frequency : Only for test_13 in sanity-quota.sh
1182 Description: a bug in quota test script
1183 Details : shouldn't check overfull allocated quota
1187 Description: jbd statistics
1188 Details : Port older jbd statistics patch for sles10
1190 Severity : enhancement
1192 Description: Add printing inode info into message about error in writepage.
1197 Description: reply_lock_interpret crash due to race with it and lock cancel.
1198 Details : Do not replay locks that are being cancelled. Do not reference
1199 locks by their address during replay, just by their handle.
1203 Frequency : When flocks are used.
1204 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1205 Details : Flock locks might destroy just granted lock if it could be merged
1206 with another existing flock, this is done in completion handler,
1207 so teach ldlm_cli_enquque_fini that this is a valid case for
1213 Description: Oops in read and write path when failing to allocate lock.
1214 Details : Check if lock allocation failed and return error back.
1218 Description: lstripe command fails for valid OST index
1219 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1220 instead of lov->desc.ld_active_tgt_count.
1224 Description: sanity.sh failed test 103
1225 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1230 Description: df doesn't work properly if diskfs blocksize != 4K
1231 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1235 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1236 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1238 Severity : enhancement
1240 Description: Adaptive timeouts
1241 Details : RPC timeouts adapt to changing server load and network
1242 conditions to reduce resend attempts and improve recovery time.
1246 Description: llapi_file_create() does not allow some changes
1247 Details : add llapi_file_open() that allows specifying the mode and
1248 open flags, and also returns an open file handle.
1252 Description: merge and cleanup kernel patches.
1253 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1257 Description: Add journal checksum support.(Kernel part)
1258 Details : The journal checksum feature adds two new flags i.e
1259 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1260 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1261 indicates that the commit block contains the checksum for
1262 the blocks described by the descriptor blocks. Now commit
1263 record can be sent to disk without waiting for descriptor
1264 blocks to be written to disk. This behavior is controlled
1265 using JBD2_FEATURE_ASYNC_COMMIT flag.
1269 Description: OSS needs mutliple precreate threads
1270 Details : Add ability to start more than one create thread per OSS.
1274 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1275 Details : execute lfs setstripe on client
1279 Description: mds_obd_create error creating tmp object
1280 Details : When the user sets quota on root, llog will be affected and can't
1281 create files and write files.
1284 Frequency : Always on ia64 patchless client, and possibly others.
1286 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1287 Details : This allows the patchless client to be loaded on architectures
1288 without this export.
1292 Description: RedHat Update kernel for RHEL5
1293 Details : Add the kernel config file for RHEL5.
1297 Description: RedHat Update kernel for RHEL5
1298 Details : Modify the kernel config file more closer RHEL5.
1302 Description: Build failure against Centos5 (RHEL5)
1303 Details : Define PAGE_SIZE when it isn't present.
1307 Description: test_7 on sanity-quota.sh can't be used on separate machines
1308 Details : add facet to handle it
1312 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1313 Details : with NFS, the anon dentry's parent was set to itself in
1314 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1315 corresponding dentry other than use rec->ur_name.
1317 Severity : enhancement
1319 Description: lfs setstripe enhancement
1320 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1324 Description: enable data checksumming by default
1325 Details : enable checksum by default, allow --disable-checksum
1326 configure option and "-o nochecksum" mount option. Checksums
1327 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1331 Description: lustre support for RHEL5
1332 Details : Add support for RHEL5.
1336 Description: Client eviction due to failover config
1337 Details : after a connection loss, the lustre client should attempt to
1338 reconnect to the last active server first before trying the
1339 other potential connections.
1343 Description: when mds and osts use different quota unit(32bit and 64bit),
1344 quota will be released repeatly.
1345 Details : void sending multiple quota reqs to mds, which will keep the
1346 status between the reqs.
1350 Description: osts not allocated evenly to files
1351 Details : change the condition to increase offset_idx
1354 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1356 Description: Data corruption for OSTs that are formatted larger than 2TB
1358 Details : When generating the bio request for lustre file writes the
1359 sector number would overflow a temporary variable before being
1360 used for the IO. The data reads correctly from Lustre (which
1361 will overflow in a similar manner) but other file data or
1362 filesystem metadata may be corrupted in some cases.
1366 Description: Tune the kernel for good SCSI performance.
1367 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1368 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1373 Description: rewrite ext2-derived code in obdclass/uuid.c
1374 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1375 from scratch preserving functionality.
1379 Description: "lfs find -obd UUID" prints directories
1380 Details : "lfs find -obd UUID" will return all directory names instead
1381 of just file names. It is incorrect because the directories
1382 do not reside on the OSTs.
1386 Description: use i_size_read and i_size_write in 2.6 port
1387 Details : replace inode->i_size access with i_size_read/write()
1390 Frequency : when removing large files
1392 Description: scheduling issue during removal of large Lustre files
1393 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1394 It causes scheduling issues when removing large files (17TB in the
1398 Frequency : only with liblustre clients on XT3
1400 Description: evictions taking too long
1401 Details : allow llrd to evict clients directly on OSTs
1404 Frequency : only on ppc
1406 Description: /proc/fs/lustre/devices broken on ppc
1407 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1408 We should make sure the type of 'index' is loff_t and then cast
1409 explicitly as needed below. Do not assign an explicitly cast
1413 Frequency : only for rhel5
1415 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1416 Details : Modify the target file & which_kernel.
1418 Severity : enhancement
1420 Description: omit set fsid for export NFS
1421 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1422 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1426 Frequency : Always, for kernels after 2.6.16
1427 Description: Fix warning idr_remove called for id=.. which is not allocated.
1428 Details : Last kernels save old s_dev before kill super and not allow
1429 to restore from callback - restore it before call kill_anon_super.
1433 Description: buffer overruns could theoretically occur
1434 Details : llapi_semantic_traverse() modifies the "path" argument by
1435 appending values to the end of the origin string, and a
1436 overrun may occur. Adding buffer overrun check in liblustreapi.
1440 Description: Fix error on 'ls .' at the top of the Lustre mount.
1441 Details : Don't revalidate dentry if it is a root dentry.
1445 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1446 Details : Modify vm-tunables-rhel4.patch.
1450 Description: Kernel config for 2.6.18-vanilla.
1451 Details : Modify targets/2.6-vanilla.target.in.
1452 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1453 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1454 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1455 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1459 Description: improve handling recoverable errors
1460 Details : if request processig with error which can be recoverable on server
1461 request should be resend, otherwise page released from cache and
1466 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1467 Details : Modify target file & which_patch.
1468 A flaw was found in the IA32 system call emulation provided
1469 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1470 value could be stored in the %RAX register, which could trigger an
1471 out-of-bounds system call table access. An untrusted local user
1472 could exploit this flaw to run code in the kernel
1473 (ie a root privilege escalation). (CVE-2007-4573).
1477 Description: O_DIRECT bypasses client statistics.
1478 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1479 read_ahead_stats were updated. copied the stats section from
1480 osc_send_oap_rpc() into async_internal().
1484 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1488 Description: change order of libsysio includes
1489 Details : '#include sysio.h' should always come before '#include xtio.h'
1493 Description: handle "serious error: objid * already exists" more gracefully
1494 Details : If LAST_ID value on disk is smaller than the objects existing in
1495 the O/0/d* directories, it indicates disk corruption and causes an
1496 LBUG(). If the object is 0-length, then we should use the existing
1497 object. This will help to avoid a full fsck in most cases.
1499 Severity : enhancement
1501 Description: adapt the lustre_config script to support the upgrade case
1502 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1503 server targets to 1.6 in parallel.
1507 Description: To avoid grant space > avaible space when the disk is almost
1508 full. Without this patch you might see the error "grant XXXX >
1509 available" or some LBUG about grant, when the disk is almost
1511 Details : In filter_check_grant, for non_grant cache write, we should
1512 check the left space by if (*left > ungranted + bytes), instead
1513 of (*left > ungranted), because only we are sure the left space
1514 is enough for another "bytes", then the ungrant space should be
1515 increase. In client, we should update cl_avail_grant only there
1516 is OBD_MD_FLGRANT in the reply.
1520 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1521 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1522 described in CVE-2007-4573. This problem could allow untrusted
1523 local users to gain root access.
1526 Frequency : when using O_DIRECT and quotas
1528 Description: Incorrect file ownership on O_DIRECT output files
1529 Details : block usage reported by 'lfs quota' does not take into account
1530 files that have been written with O_DIRECT.
1535 Description: touch file failed when fs is not full
1536 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1537 otherwise we can get ENOSP while fs is not full.
1541 Description: parallel lock callbacks
1542 Details : Instead of sending blocking and completion callbacks as separated
1543 requests, adding them to a set and sending in parallel.
1546 Frequency : only for Cray XT3
1547 Bugzilla : 12829/13455
1548 Description: Changing primary group doesn't change the group lustre assigns to
1550 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1551 overridden with the primary group provided by the group upcall,
1552 whereas the supplied fsgid can be trusted if it is in the list of
1553 supplementary groups returned by the group upcall.
1555 Severity : enhancement
1557 Description: Allow masking D_WARNING, D_ERROR messages from console
1558 Details : Console messages can now be disabled via lnet.printk.
1562 Description: User code with malformed file open parameter crashes client node
1563 Details : Before packing join_file req, all the related reference should be
1564 checked carefully in case some malformed flags cause fake
1565 join_file req on client.
1569 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1570 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1571 mode is equal to the granted mode. However, it should also check
1572 that LDLM_FL_CP_REQD is not set.
1576 Description: Heavy nfs access might result in deadlocks
1577 Details : After ELC code landed, it is now improper to enqueue any mds
1578 locks under och_sem, because enqueue might want to decide to
1579 cancel open locks for same inode we are holding och_sem for.
1583 Description: Client eviction while running blogbench
1584 Details : A lot of unlink operations with concurrent I/O can lead to a
1585 deadlock causing evictions. To address the problem, the number of
1586 oustanding OST_DESTROY requests is now throttled to
1587 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1588 callbacks are processed in priority.
1592 Description: enable ACLs on MDS by default
1593 Details : ACLs must be enabled on MDS by default.
1596 Frequency : PPC/PPC64 only
1598 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1599 Details : fix duplicated definitions between asm-ppc64/types.h and
1600 lustre_types.h on PPC.
1603 Frequency : PPC/PPC64 only
1605 Description: asm-ppc/segment.h does not exist
1606 Details : fix compile issue on PPC.
1610 Description: better handle error messages in extents code
1613 Frequency : RHEL4 only
1615 Description: mkfs is very slow on IA64/RHEL4
1616 Details : A performance regression has been discovered in the MPT Fusion
1617 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1618 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1619 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1621 Severity : enhancement
1623 Description: SNMP support enhancement
1624 Details : Adding total number of sampled request for an MDS node in snmp
1627 Severity : enhancement
1629 Description: Optimize ldlm waiting list processing for PR extent locks
1630 Details : When processing waiting list for read extent lock and meeting read
1631 lock that is same or wider to it that is not contended, skip
1632 processing rest of the list and immediatelly return current
1633 status of conflictness, since we are guaranteed there are no
1634 conflicting locks in the rest of the list.
1638 Description: Time out and refuse to reconnect
1639 Details : When the failover node is the primary node, it is possible
1640 to have two identical connections in imp_conn_list. We must
1641 compare not conn's pointers but NIDs, otherwise we can defeat
1642 connection throttling.
1646 Description: port llog fixes from b1_6 into HEAD
1647 Details : Port llog reference couting and some llog cleanups from b1_6
1648 (bug 10800) into HEAD, for protect from panic and access to already
1649 free llog structures.
1653 Description: Detect stride IO mode in read-ahead
1654 Details : When a client does stride read, read-ahead should detect that and
1655 read-ahead pages according to the detected stride pattern.
1659 Description: data checksumming impacts single node performance
1660 Details : add support for several checksum algorithm. Currently, only CRC32
1661 and Adler-32 are supported. The checksum type can be changed on
1662 the fly via /proc/fs/lustre/osc/*/checksum_type.
1666 Description: use adler32 for page checksums
1667 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1672 Description: build for x2 fails
1673 Details : fix compile issue on Cray systems.
1677 Description: Properly match for duplicate locks
1678 Details : Due to different lock order from skiplists code, we need to
1679 traverse entire list for now
1682 Frequency : only on PPC/SLES10
1684 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1685 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1686 be defined. Add a hack in mkfs_lustre.c to work around this compile
1691 Description: LASSERT on MDS when client holding flock lock dies
1692 Details : ldlm pool logic depends on number of granted locks equal to
1693 number of released locks which is not true for flock locks, so
1694 just exclude such locks from consideration.
1698 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1699 Details : Use fsfilt_sync() for both the whole filesystem sync and
1700 individual file sync to eliminate dangerous inode locking
1701 with I_LOCK that can lead to a deadlock.
1705 Description: performance in 1.6.3
1706 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1707 sure that 1MB requests are not fragmented by the block layer.
1709 Severity : enhancement
1711 Description: organize the server-side client stats on per-nid basis
1712 Details : Change the structure of stats under obdfilter and mds to
1720 The "uuid"s file would list the uuids of _active_ exports.
1721 And the clear entry is to clear all stats and stale nids.
1723 Severity : enhancement
1725 Description: eliminate client locks in face of contention
1726 Details : file contention detection and lockless i/o implementation
1727 for contended files.
1731 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1734 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1736 Description: MDS may lose file striping (and hence file data) in some cases
1737 Details : If there are additional extended attributes stored on the MDS,
1738 in particular ACLs, SELinux, or user attributes (if user_xattr
1739 is specified for the client mount options) then there is a risk
1740 of attribute loss. Additionally, the Lustre file striping
1741 needs to be larger than default (e.g. striped over all OSTs),
1742 and an additional attribute must be stored initially in the
1743 inode and then increase in size enough to be moved to the
1744 external attribute block (e.g. ACL growing in size) for file
1749 Description: skiplist implementation simplification
1750 Details : skiplists are used to group compatible locks on granted list
1751 that was implemented as tracking first and last lock of each
1752 lock group the patch changes that to using doubly linked lists
1756 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1757 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1758 because LASSERT on both the data supplied by a client, and the
1759 data on disk is dangerous and incorrect.
1761 Severity : enhancement
1763 Description: Slow truncate/writes to huge files at high offsets.
1764 Details : Directly associate cached pages to lock that protect those pages,
1765 this allows us to quickly find what pages to write and remove
1766 once lock callback is received.
1770 Description: more ldlm soft lockups
1771 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1772 starve other threads from the resource lock for a long time in
1773 case of long waiting queue, so change the debug level from
1774 D_OTHER to the less frequently used D_INFO.
1776 Severity : enhancement
1778 Description: add -gid, -group, -uid, -user options to lfs find
1782 Description: Hung threads in invalidate_inode_pages2_range
1783 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1784 once one is completed. As a result, some RPCs are stuck in the
1785 queue and are never sent.
1789 Description: filter threads hungs on waiting journal commit
1790 Details : Cleanup filter group llog code, then only filter group llog will
1791 be only created in the MDS/OST syncing process.
1795 Description: Procfs and llog threads access destoryed import sometimes.
1796 Details : Sync the import destoryed process with procfs and llog threads by
1797 the import refcount and semaphore.
1799 Severity : enhancement
1801 Description: openlock cache of b1_6 port to HEAD
1806 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1807 Details : If insertion of an extent fails, then discard the inode
1808 preallocation and free data blocks else it can lead to duplicate
1813 Description: don't always update ctime in ext3_xattr_set_handle()
1814 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1815 In some cases the ctime should not be updated, for example for
1816 2.0->1.8 compatibility it is necessary to delete an xattr and it
1817 should not update the ctime.
1821 Bugzilla : 15713/16362
1822 Description: Assertion in iopen_connect_dentry in 1.6.3
1823 Details : looking up an inode via iopen with the wrong generation number can
1824 populate the dcache with a disconneced dentry while the inode
1825 number is in the process of being reallocated. This causes an
1826 assertion failure in iopen since the inode's dentry list contains
1827 both a connected and disconnected dentry.
1831 Description: assertion failure in ldlm_handle2lock()
1832 Details : fix a race between class_handle_unhash() and class_handle2object()
1833 introduced in lustre 1.6.5 by bug 13622.
1838 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1839 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1844 Description: oops in sd_iostats_seq_show()
1845 Details : unloading/reloading the scsi low level driver triggers a kernel
1846 bug when trying to access the sd iostat file.
1851 Description: Kernel panics during QLogic driver reload
1852 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1853 patch, causing memory corruption.
1858 Description: journal_dev option does not work in b1_6
1859 Details : pass mount option during pre-mount.
1861 Severity : enhancement
1863 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1864 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1865 extent information of a file. It can be used to map logical blocks
1866 in a file to physical blocks in the block device.
1870 Description: LDLM soft lockups - improvement
1871 Details : It is be possible to send the lock handle along with each read
1872 or write request because the client is already doing a lock match
1873 itself so there isn't any reason the OST should have to re-do that
1877 Frequency : only on Cray X2
1879 Description: X2 build failures
1880 Details : fix build failures on Cray X2.
1884 Description: xid & resent requests
1885 Details : Initialize RPC XID from clock at startup (randomly if clock is
1888 Severity : enhancement
1890 Description: Add lustre_start utility to start or stop multiple Lustre servers
1895 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1896 Details : In case of memory pressure, list_del() can be called twice on
1897 req->rq_history_list, causing a kernel oops.
1901 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1902 Details : fix stack overflow in the distributed lock manager by defering
1903 export eviction after a failed AST to the elt thread instead of
1904 handling it in the dlm interpret routine.
1908 Description: Convert some comments to new format.
1909 Details : Update documenting comments to match doxygen conventions.
1913 Description: Grammar fixes.
1914 Details : A couple of trivial spelling fixes.
1918 Description: OSD_COUNTERS-mandatory
1919 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1920 to simplify the code.
1924 Description: simplify cmm_device freeing logic.
1925 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1929 Description: Add lockdep support to dt_object_operations locking interface.
1930 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1931 parameter indicating lock ordering. Update mdd code to use new
1936 Description: Introduce failloc constants for lockless IO tests.
1937 Details : Add two new failloc constants to test lockless IO. Only one of
1938 them in implemented---another is checked in yet to be landed
1943 Description: Add lockdep support for inode mutex.
1944 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1945 in the situations where Lustre has to lock more than one inode
1950 Description: Add optional invariants checking support.
1951 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1952 configure switch. This macro is to be used for consistency and
1953 sanity checks that are too expensive to be left in `production'
1958 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1959 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1964 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1965 Details : Use spin_lock_nested() in (the only) situation where more than
1966 one ldlm_lock is locked simultaneously. Also, fix possible
1967 dead-lock in ldlm_lock_change_resource() by enforcing particular
1972 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1973 Details : Instead of specifying each ldlm_lock call-back through separate
1974 parameter, wrap them into struct ldlm_callback_suite.
1978 Description: Kill join_lru obd method and its callers.
1979 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1980 in memory---a requirement implemented through ->o_join_lru() obd
1981 method in HEAD. Remove this method and its users.
1985 Description: Add asynchronous ldlm ENQUEUE completion handler.
1986 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1987 case that stock ldlm_completion_ast() cannot handle as it waits
1988 until lock is granted. Introduce new ldlm_completion_ast_async()
1989 for this. Also comment ldlm_completion_ast().
1993 Description: ldlm_error <-> errno conversion.
1994 Details : Add functions to map (rather arbitrary) between LDLM error codes
1995 and standard errno values. CLIO needs this to prevent LDLM specific
1996 constants from escaping ldlm and osc.
2000 Description: Kill unused ldlm_handle2lock_ns() function.
2001 Details : Kill unused ldlm_handle2lock_ns() function.
2005 Description: Add lu_ref support to ldlm_lock
2006 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2007 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2008 and ldlm_resource. LDLM interface has to be changed, because code that
2009 releases a reference on a lock, has to "know" what reference this is.
2010 In the most frequent case
2012 lock = ldlm_handle2lock(handle);
2014 LDLM_LOCK_PUT(lock);
2016 no changes are required. When any other reference (received _not_ from
2017 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2018 instead of LDLM_LOCK_PUT().
2020 Arguably, changes are pervasive, and interface requires some discipline
2021 for proper use. On the other hand, it was very instrumental in finding
2022 a few leaked lock references.
2026 Description: Add ldlm_lock_addref_try().
2027 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2028 attempts to addref a lock that might be being canceled
2033 Description: Add ldlm_weigh_callback().
2034 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2035 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2036 hard-coded `number of pages' logic.
2040 Description: Add lockdep annotations to llog code.
2041 Details : Use appropriately tagged _nested() locking calls in the places
2042 where llog takes more than one ->lgh_lock lock.
2046 Description: Add loi_kms_set().
2047 Details : Wrap kms updates into a helper function.
2051 Description: Constify instances of struct lsm_operations.
2052 Details : Constify instances of struct lsm_operations.
2056 Description: lu_conf support.
2057 Details : On a server, a file system object is uniquely identified
2058 by a fid, which is sufficient to locate and load all object
2059 state (inode). On a client, on the other hand, more data are
2060 necessary instantiate an object. Change lu_object_find() and
2061 friends to take additional `lu_conf' argument describing object.
2062 Typically this includes layout information.
2066 Description: lu_context fixes.
2067 Details : Introduce new lu_context functions that are needed on the client
2068 side, where some system threads (ptlrpcd) are shared by multiple
2069 modules, and so cannot be stopped during module shutdown.
2073 Description: Add start and stop methods to lu_device_type_operations.
2074 Details : Introduce two new methods in lu_device_type_operations, that are
2075 invoked when first instance of a given type is created and last one
2076 is destroyed respectively. This is need by CLIO.
2080 Description: Add lu_ref support to struct lu_device.
2081 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2082 track leaked references.
2086 Description: Introduce lu_kmem_descr.
2087 Details : lu_kmem_descr and its companion interface allow to create
2088 and destroy a number of kmem caches at once.
2092 Description: Fix lu_object finalization race.
2093 Details : Fix a race between lu_object_find() finding an object and its
2094 concurrent finalization. This race is (most likely) not possible
2095 on the server, but might happen on the client.
2099 Description: Introduce lu_ref interface.
2100 Details : lu_ref is a debugging module allowing to track references to
2101 a given object. It is quite cpu expensive, and has to be
2102 explicitly enabled with --enable-lu_ref. See usage description
2107 Description: Factor lu_site procfs stats into a separate function.
2108 Details : Separate lu_site stats printing code into a separate function
2109 that can be reused on a client.
2113 Description: Constify instances of struct {lu,dt,md}_device_operations.
2114 Details : Constify instances of struct {lu,dt,md}_device_operations.
2118 Description: Introduce struct md_site and move meta-data specific parts of
2119 struct lu_site here.
2120 Details : Move md-specific fields out of struct lu_site into special struct
2121 md_site, so that lu_site can be used on a client.
2125 Description: Kill mdd_lov_destroy().
2126 Details : Remove unused mdd code.
2130 Description: Add st_block checking to multistat.c.
2131 Details : Add st_block checking to multistat.c.
2135 Description: Add lu_ref support to struct obd_device.
2136 Details : Add lu_ref tracking to obd_device.
2140 Description: Kill obd_set_fail_loc().
2141 Details : Remove unused code.
2145 Description: Add special type for ptlrpc_request interpret functions.
2146 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2147 there. Actual usage will be in CLIO.
2151 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2152 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2153 doesn't work with lockdep.
2157 Description: Add rwv.c test program.
2158 Details : New testing program exercising readv(2) and writev(2) (Qian).
2162 Description: Add sendfile.c test program.
2163 Details : New testing program exercising sendfile(2) (Jay).
2167 Description: Ratelimit a message that can be very frequent.
2168 Details : Ratelimit a memory allocation failure message that can
2173 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2174 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2178 Description: Liblustre build fixes.
2179 Details : Liblustre build fixes.
2183 Description: libcfs: add cfs_{need,cond}_resched() interface.
2184 Details : libcfs: add cfs_{need,cond}_resched() definition and
2185 implementations for Linux, NT, and liblustre.
2187 Severity : enhancement
2189 Description: More exported tunables for mballoc
2190 Details : Add support for tunable preallocation window and new tunables for
2191 large/small requests
2195 Description: Detect corruption of block bitmap and checking for preallocations
2196 Details : Checks validity of on-disk block bitmap. Also it does better
2197 checking of number of applied preallocations. When corruption is
2198 found, it turns filesystem readonly to prevent further corruptions.
2202 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2203 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2204 window to 0. If it does hit cache miss, and read-pattern is still
2205 stride-io mode, does not reset the stride window, but also does not
2206 increase the stride window length in this case.
2210 Frequency : only for big-endian servers
2211 Description: Check if system is big-endian while mounting fs with extents feature
2212 Details : Mounting a filesystem with extents feature will fail on big-endian
2213 systems since ext3-based ldiskfs is not supported on big-endian
2214 systems. This can be over-riden with "bigendian_extents" mount option.
2216 Severity : enhancement
2218 Description: The root squash functionality
2219 Details : A security feature, which is to prevent users from being able
2220 to mount lustre on their desktop, run as root, and delete
2221 all of the files in the filesystem. The goal is accomplished by
2222 remapping user id (UID) and group id (GID) of the root user to
2223 a UID and GID specified by the system administartor via Lustre
2224 configuration management server (MGS). The functionality also
2225 allows to specify sets of clients for which the remapping does
2230 Description: Excessive recovery window
2231 Details : With AT enabled, the recovery window can be excessively long (6000+
2232 seconds). To address this problem, we no longer use
2233 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2234 timeout no longer depends on the service time, it is set to
2235 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2236 time via pb_service_time.
2240 Description: Watchdog triggered on MDS failover
2241 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2242 the OSTs know that the MDS "UUID" can be reused for the same export
2243 from a different NID, so we do not need to wait for the export to be
2247 Frequency : rare, only if using MMP with Linux RAID
2249 Description: MMP doesn't work with Linux RAID
2250 Details : While using HA for Lustre servers with Linux RAID, it is possible
2251 that MMP will not detect multiple mounts. To make this work we
2252 need to unplug the device queue in RAID when the MMP block is being
2253 written. Also while reading the MMP block, we should read it from
2254 disk and not the cached one.
2256 Severity : enhancement
2258 Description: open file using fid
2259 Details : A file can be opened using just its fid, like
2260 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2263 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2266 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2267 Details : add proper locking for creation of jbd_4k slab cache
2271 Description: MMP check in ext3_remount() fails without displaying any error
2272 Details : When multiple mount protection fails during remount, proper error
2275 Severity : enhancement
2277 Description: Allow stripe size to be up to 4G-64k
2278 Details : Fix math logic to allow large stripe sizes.
2282 Description: add check for >8TB ldiskfs filesystems
2283 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2284 Don't allow >8TB ldiskfs filesystems to be mounted without
2285 force_over_8tb mount option
2290 Descriptoin: OSS complains "too many blocks" and crashes
2291 Details : In filter_commitrw_write():
2292 1. lquota_chkquota();
2293 2. obdo_from_inode();
2294 3. lquota_pending_commit();
2295 "2." has changed uid and gid in this case, then caused
2296 lquota_chkquota() and lquota_pending_commit() to use
2297 different uid and gid. That is the root of the bug.
2301 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2302 Details : Detect and print out the 'stripe count = -1' case properly.
2306 Description: Out or order replies might be lost on replay
2307 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2308 request with tid smaller then one currently being added, add it
2309 to the start, not end of the list.
2311 --------------------------------------------------------------------------------
2313 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2315 * Support for kernels:
2316 2.6.5-7.283 (SLES 9),
2317 2.6.9-55.EL (RHEL 4),
2318 2.6.16.46-0.14 (SLES 10),
2319 2.6.18.8 vanilla (kernel.org)
2320 * Client support for unpatched kernels:
2321 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2322 2.6.16 - 2.6.22 vanilla (kernel.org)
2323 * Due to recently discovered recovery problems, we do not recommend
2324 using patchless RHEL 4 clients with this or any earlier release.
2325 * Recommended e2fsprogs version: 1.39.cfs8
2326 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2327 * Starting with this release, the ldiskfs backing filesystem required
2328 by Lustre is now in its own package, lustre-ldiskfs. This package
2329 should be installed. It is versioned separately from Lustre and
2330 may be released separately in future.
2334 Description: block reactivating mgc import until all deactivates complete
2335 Details : Fix race when failing back MDT/MGS to itself (testing)
2338 Frequency : at statup only
2340 Description: mds_lov_synchronize race leads to various problems
2341 Details : simultaneous MDT->OST connections at startup can cause the
2342 sync to abort, leaving the OSC in a bad state.
2344 Severity : enhancement
2346 Description: add optional extra BUILD_VERSION info
2347 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2348 to override the lustre version.
2351 Frequency : 2.6.18 servers only
2353 Description: ll_kern_mount() doesn't release the module reference
2354 Details : The ldiskfs module reference count never drops down to 0
2355 because ll_kern_mount() doesn't release the module reference.
2360 Description: server LBUG when using old ost_num_threads parameter
2361 Details : Accept the old ost_num_threads parameter but warn that it
2362 is deprecated, and fix an off-by-one error that caused an LBUG.
2367 Description: Transient SCSI error results in persistent IO issue
2368 Details : iobuf->dr_error is not reinitialized to 0 between two
2372 Frequency : sometimes when underlying device returns I/O errors
2374 Description: OSTs not going read-only during write failures
2375 Details : OSTs are not remounted read-only when the journal commit threads
2376 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2377 instead of the ext3 wrappers.
2381 Description: poor connect scaling with increasing client count
2382 Details : Don't run filter_grant_sanity_check for more than 100 exports
2383 to improve scaling for large numbers of clients.
2386 Frequency : SLES10 only
2388 Description: sanity-quota.sh quotacheck failed: rc = -22
2389 Details : Quotas cannot be enabled on SLES10.
2392 Frequency : liblustre clients only
2394 Description: getdirentries does not give error when run on compute nodes
2395 Details : getdirentries does not fail when the size specified as an argument
2396 is too small to contain at least one entry
2398 Severity : enhancement
2400 Description: Add LNET router traceability for debug purposes
2401 Details : If a checksum failure occurs with a router as part of the
2402 IO path, the NID of the last router that forwarded the bulk data
2403 is printed so it can be identified.
2408 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2409 Details : Due to a race condition, liblustre clients were occasionally
2410 evicted incorrectly.
2412 Severity : enhancement
2414 Description: lfs setstripe use optional parameters instead of postional
2417 Severity : enhancement
2419 Description: Nanosecond timestamp support for ldiskfs
2420 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2421 resolution timestamps. There is not yet support for this at
2422 the Lustre filesystem level.
2425 Frequency : during server recovery
2427 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2428 Details : request with rq_no_resend flag not awake l_wait_event if they get
2432 Frequency : nfs export on patchless client
2434 Description: connectathon hang when test nfs export over patchless client
2435 Details : Disconnected dentry cannot be found with lookup, so we do not need
2436 to unhash it or make it invalid
2439 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2440 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2441 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2446 Description: open req refcounting wrong on reconnect
2447 Details : If reconnect happened between getting open reply from server and
2448 call to mdc_set_replay_data in ll_file_open, we will schedule
2449 replay for unreferenced request that we are about to free.
2450 Subsequent close will crash in variety of ways.
2451 Check that request is still eligible for replay in
2452 mdc_set_replay_data().
2457 Description: disable writes to filesystem when reading health_check file
2458 Details : the default for reading the health_check proc file has changed
2459 to NOT do a journal transaction and write to disk, because this
2460 can cause reads of the /proc file to hang and block HA state
2461 checking on a healthy but otherwise heavily loaded system. It
2462 is possible to return to the previous behaviour during configure
2463 with --enable-health-write.
2465 Severity : enhancement
2467 Description: 64-bit inode version
2468 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2469 to the inode. This will be required for version-based recovery.
2474 Description: MDS fails to start if a duplicate client export is detected
2475 Details : in some rare cases it was possible for a client to connect to
2476 an MDS multiple times. Upon recovery the MDS would detect this
2477 and fail during startup. Handle this more gracefully.
2479 Severity : enhancement
2481 Description: Add -o localflock option to simulate old noflock
2483 Details : This will achieve local-only flock/fcntl locks
2489 Description: log_commit_thread vs filter_destroy race leads to crash
2490 Details : Take import reference before releasing llog record semaphore
2495 Description: Wrong request locking in request set processing
2496 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2497 imp_delayed_list, in this place should be used imp_lock.
2500 Frequency : when reconnection
2502 Description: Grant Leak when osc reconnect to OST
2503 Details : When osc reconnect ost, OST(filter) should check whether it
2504 should grant more space to client by comparing fed_grant and
2505 cl_avail_grant, and return the granted space to client instead
2506 of "new granted" space, because client will call osc_init_grant
2507 to update the client grant space info.
2510 Frequency : when client reconnect to OST
2512 Description: Grant Leak when osc do resend and replay bulk write
2513 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2514 bulk write request, because the grant info while be sync between
2515 OSC and OST when reconnect, and we should ignore the grant info
2516 these of resend/replay write req.
2521 Description: Grant space more than avaiable left space sometimes.
2522 Details : When then OST is about to be full, if two bulk writing from
2523 different clients came to OST. Accord the avaliable space of the
2524 OST, the first req should be permitted, and the second one
2525 should be denied by ENOSPC. But if the seconde arrived before
2526 the first one is commited. The OST might wrongly permit second
2527 writing, which will cause grant space > avaiable space.
2530 Frequency : when client is evicted
2532 Description: Grant might be wrongly erased when osc is evicted by OST
2533 Details : when the import is evicted by server, it will fork another
2534 thread ptlrpc_invalidate_import_thread to invalidate the
2535 import, where the grant will be set to 0. While the original
2536 thread will update the grant it got when connecting. So if
2537 the former happened latter, the grant will be wrongly errased
2538 because of this race.
2543 Description: Checking Stale with correct fid
2544 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2545 to check whether it is stale, because sometimes, we want the
2546 enqueue happened anyway, and op_data.fid2 will not be initialized.
2548 Severity : enhancement
2550 Description: update patchless client
2551 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2554 Frequency : only with 2.4 kernel
2556 Description: random memory corruption
2557 Details : size of struct ll_inode_info is to big for union inode.u and this
2558 can be cause of random memory corruption.
2563 Description: Memory leak in recovery
2564 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2565 It should also check obd_fail before fsfilt_start, otherwise if
2566 fsfilt_start return -EROFS,(failover mds during mds recovery).
2567 then the req will return with repmsg->transno = 0 and rc = EROFS.
2568 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2569 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2570 be freed no matter whether fsfilt_commit success or not.
2573 Frequency : only with huge count clients
2575 Description: Prevents from taking the superblock lock in llap_from_page for
2577 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2578 for llap_from_page call in ll_removepage prevents from taking the
2579 superblock lock for a soon died page.
2584 Description: Not check open intent error before release open handle
2585 Details : in some rare cases, the open intent error is not checked before
2586 release open handle, which may cause
2587 ASSERTION(open_req->rq_transno != 0), because it tries to release
2588 the failed open handle.
2593 Description: Set cat log bitmap only after create log success.
2594 Details : in some rare cases, the cat log bitmap is set too early. and it
2595 should be set only after create log success.
2599 Description: Accessing a block bevice can re-enable I/O when Lustre is
2600 tearing down a device.
2601 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2606 Description: service threads may hog cpus when there are a lot of requests
2607 Details : Insert cond_resched to give other threads a chance to use some CPU
2612 Description: the cat log was not initialized in recovery
2613 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2614 unlink log on mds will not be initialized until mds post
2615 recovery. And also in mds post recovery, the unlink log will
2616 initialization will be done asynchronausly, so there will be race
2617 between add unlink log and unlink log initialization.
2621 Description: brw_stats were being printed incorrectly
2622 Details : brw_stats were being printed as log2 but all of them were not
2623 recorded as log2. Also remove some code duplication arising from
2624 filter_tally_{read,write}.
2628 Frequency : rare, only in recovery.
2629 Description: ASSERTION(req->rq_type != LI_POISON) failed
2630 Details : imp_lock should be held while iterating over imp_sending_list for
2631 prevent destroy request after get timeout in ptlrpc_queue_wait.
2635 Description: replay-single.sh test 52 fails
2636 Details : A lock's skiplist need to be cleanup when it being unlinked
2637 from its resource list.
2641 Description: Short directio read returns full requested size rather than
2643 Details : Direct I/O operations should return actual amount of bytes
2644 transferred rather than requested size.
2646 Severity : enhancement
2648 Description: metadata RPC reduction (e.g. for rm performance)
2649 Details : decrease the amount of synchronous RPC between clients and servers
2650 by canceling conflicing lock before the operation on the client
2651 and packing thier handles into the main operation RPC to server.
2653 Severity : enhancement
2655 Description: Async OSC create to avoid the blocking unnecessarily.
2656 Details : If a OST has no remain object, system will block on the creating
2657 when need to create a new object on this OST. Now, ways use
2658 pre-created objects when available, instead of blocking on an
2659 empty osc while others are not empty. If we must block, we block
2660 for the shortest possible period of time.
2664 Description: improve handling recoverable errors
2665 Details : if request processig with error which can be recoverable on server
2666 request should be resend, otherwise page released from cache and
2669 Severity : enhancement
2671 Description: refine locking for avoid write wrong info into lov_objid file
2672 Details : fix possible races with add new target and write/update data in
2675 --------------------------------------------------------------------------------
2677 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2682 Frequency : on some architectures
2684 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2685 Details : Uninitialized flags sometimes cause configuration commands to
2689 Frequency : patchless clients only
2691 Description: missing __iget() symbol export
2692 Details : The __iget() symbol export is missing. To avoid the need for
2693 this on patchless clients the deathrow inode reaper is turned
2694 off, and we depend on the VM to clean up old inodes. This
2695 dependency was during via the fix for bug 12181.
2697 --------------------------------------------------------------------------------
2699 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2701 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2702 INTEROPERATE with older versions automatically. In many cases a
2703 special upgrade step is needed. Please read the
2704 user documentation before upgrading any part of a live system.
2705 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2706 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2707 * WARNING: Lustre configuration and startup changes are required with
2708 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2712 Severity : enhancement
2714 Description: MountConf
2715 Details : Lustre configuration is now managed via mkfs and mount
2716 commands instead of lmc and lconf. New obd types (MGS, MGC)
2717 are added for dynamic configuration management. See
2718 https://mail.clusterfs.com/wikis/lustre/MountConf for
2721 Severity : enhancement
2723 Description: dynamic OST addition
2724 Details : OSTs can now be added to a live filesystem
2726 Severity : enhancement
2728 Description: startup order invariance
2729 Details : MDTs and OSTs can be started in any order. Clients only
2730 require the MDT to complete startup.
2732 Severity : enhancement
2734 Description: parallel, asynchronous orphan cleanup
2735 Details : orphan cleanup is now performed in separate threads for each
2736 OST, allowing parallel non-blocking operation.
2738 Severity : enhancement
2740 Description: optimized stripe assignment
2741 Details : stripe assignments are now made based on ost space available,
2742 ost previous usage, and OSS previous usage, in order to try
2743 to optimize storage space and networking resources.
2745 Severity : enhancement
2747 Description: Permanently set tunables
2748 Details : All writable /proc/fs/lustre tunables can now be permanently
2749 set on a per-server basis, at mkfs time or on a live system.
2751 Severity : enhancement
2753 Description: Lustre message v2
2754 Details : Add lustre message format v2.
2756 Severity : enhancement
2758 Description: client OST exclusion list
2759 Details : Clients can be started with a list of OSTs that should be
2760 declared "inactive" for known non-responsive OSTs.
2764 Description: SPEC SFS validation failure on NFS v2 over lustre.
2765 Details : Changes the blocksize for regular files to be 2x RPC size,
2766 and not depend on stripe size.
2768 Severity : enhancement
2770 Description: Multiple MD RPCs in flight.
2771 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2772 To avoid overly-overloading MDS, introduce a limit on number of
2773 MDS RPCs in flight for a single client and add /proc controls
2774 to adjust this limit.
2776 Severity : enhancement
2778 Description: client read/write statistics
2779 Details : Add client read/write call usage stats for performance
2780 analysis of user processes.
2781 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2782 file access. extents_stats shows chunk size distribution.
2783 extents_stats_per_process show chunk size distribution per
2786 Severity : enhancement
2788 Description: mds statistics
2789 Details : Add detailed mds operations statistics in
2790 /proc/fs/lustre/mds/*/stats.
2794 Description: Failure of copying files with lustre special EAs.
2795 Details : Client side always return success for setxattr call for lustre
2796 special xattr (currently only "trusted.lov").
2801 Description: Refcount LNET uuids
2802 Details : The global LNET uuid list grew linearly with every startup;
2803 refcount repeated list entries instead of always adding to
2806 Severity : enhancement
2808 Description: Dynamic service threads
2809 Details : Within a small range, start extra service threads
2810 automatically when the request queue builds up.
2813 Frequency : mixed-endian client/server environments
2815 Description: mixed-endian crashes
2816 Details : The new msg_v2 system had some failures in mixed-endian
2819 Severity : enhancement
2821 Description: Easy OST removal
2822 Details : OSTs can be permanently deactivated with e.g. 'lctl
2823 conf_param lustre-OST0001.osc.active=0'
2825 Severity : enhancement
2827 Description: MGS proc entries
2828 Details : Added basic proc entries for the MGS showing what filesystems
2831 Severity : enhancement
2833 Description: provide MGS failover
2834 Details : Added config lock reacquisition after MGS server failover.
2836 Severity : enhancement
2838 Description: add Linux 2.4 support
2839 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2843 Description: a large application tries to do I/O to the same resource and dies
2844 in the middle of it.
2845 Details : Check the req->rq_arrival time after the call to
2846 ost_brw_lock_get(), but before we do anything about
2847 processing it & sending the BULK transfer request. This
2848 should help move old stale pending locks off the queue as
2849 quickly as obd_timeout.
2852 Frequency : when an incorrect nid is specified during startup
2854 Description: ptlrpc connect to non-existant node causes kernel crash
2855 Details : LNET can't be re-entered from an event callback, which
2856 happened when we expire a message after the export has been
2857 cleaned up. Instead, hand the zombie cleanup off to another
2860 Severity : enhancement
2862 Description: plain/inodebits lock performance improvement
2863 Details : Grouping plain/inodebits in granted list by their request modes
2864 and bits policy, thus improving the performance of search through
2868 Frequency : only if OST filesystem is corrupted
2870 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2871 Details : for a short time RPCs with bulk IO are in the replay list,
2872 but replay of bulk IOs is unimplemented. If the OST filesystem
2873 is corrupted due to disk cache incoherency and then replay is
2874 started it is possible to trip an assertion. Avoid putting
2875 committed RPCs into the replay list at all to avoid this issue.
2878 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2880 Description: System hang on startup
2881 Details : This bug allowed the liblustre (e.g. catamount) client to
2882 return to the app before handling all startup RPCs. This
2883 could leave the node unresponsive to lustre network traffic
2884 and manifested as a server ptllnd timeout.
2886 Severity : enhancement
2888 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2889 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2890 boolean to control whether to print peer debug info when a
2891 client's RPC times out.
2894 Frequency : only for kernels with patches from Lustre below 1.4.3
2896 Description: Remove old rdonly API
2897 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2900 Frequency : only for devices with external journals
2902 Description: Set external device read-only also
2903 Details : During a commanded failover stop, we set the disk device
2904 read-only while the server shuts down. We now also set any
2905 external journal device read-only at the same time.
2908 Frequency : when upgrading from 1.4 while trying to change parameters
2910 Description: The wrong (new) MDC name was used when setting parameters for
2911 upgraded MDT's. Also allows changing of OSC (and MDC)
2912 parameters if --writeconf is specified at tunefs upgrade time.
2915 Frequency : when setting specific ost indicies
2917 Description: QOS code breaks on skipped indicies
2918 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2919 created with --index need not be sequential.
2921 Severity : enhancement
2923 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2924 Details : The uninit_groups feature works in conjunction with the kernel
2925 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2926 up the pass1 processing of e2fsck. This is a read-only feature
2927 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2928 mount filesystems that have had this feature enabled.
2930 Severity : enhancement
2932 Description: Improve multi-block allocation algorithm to avoid fragmentation
2933 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2934 allocation locality and avoid filesystem fragmentation.
2936 ------------------------------------------------------------------------------
2938 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2940 * Support for kernels:
2941 2.4.21-47.0.1.EL (RHEL 3)
2942 2.6.5-7.283 (SLES 9)
2943 2.6.9-42.0.10.EL (RHEL 4)
2944 2.6.12.6 vanilla (kernel.org)
2945 2.6.16.27-0.9 (SLES 10)
2946 * Recommended e2fsprogs version: 1.39.cfs5
2948 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2952 Frequency : occasional, depends on client load and configuration
2953 Bugzilla : 12181, 12203
2954 Description: data loss for recently-modified files
2956 Details : In some cases it is possible that recently written or created
2957 files may not be written to disk in a timely manner (this should
2958 normally be within 30s unless client IO load is very high).
2959 The problem appears as zero-length files or files that are a
2960 multiple of 1MB in size after a client crash or client eviction
2961 that are missing data at the end of the file.
2963 This problem is more likely to be hit on clients where files are
2964 repeatedly created and unlinked in the same directory, clients
2965 have a large amount of RAM, have many CPUs, the filesystem has
2966 many OSTs, the clients are rebooted frequently, and/or the files
2967 are not accessed by other nodes after being written.
2969 The presence of the problem can be detected by looking at
2970 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2971 smaller than the second (nr_unused) then dirty files will not
2972 be flushed automatically to disk. "sync; sleep 10" should be
2973 run several times on the node before unmounting it to update
2974 Lustre (this is also safe to run on nodes without this problem).
2976 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2977 that can cause this same problem, so customers using that kernel
2978 also need to update the kernel in addition to Lustre. In order
2979 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2981 It is normal that files written just before a client crash (less
2982 than 30s) may not yet have been flushed to disk, even for local
2986 Frequency : frequent on thin XT3 nodes
2988 Description: UUID collision on thin XT3 Linux nodes
2989 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2990 randomly, since we relied on an insufficiently-seeded PRNG.
2995 Description: OSS hangs after "All ost request buffers busy"
2996 Details : A deadlock between quota and journal operations caused OSS
2997 hangs after printing "All ost request buffers busy."
3000 Frequency : always on liblustre builds
3002 Description: Cleanup compiler warnings on liblustre
3005 Frequency : always on liblustre builds on XT3
3007 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3011 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3012 > 32000 subdirectories
3013 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3014 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3015 subdirectory count crosses 32000. This will aid e2fsck to
3016 correctly handle more than 32000 subdirectories.
3019 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3021 Description: System hang on startup
3022 Details : This bug allowed the liblustre (e.g. catamount) client to
3023 return to the app before handling all startup RPCs. This
3024 could leave the node unresponsive to lustre network traffic
3025 and manifested as a server ptllnd timeout.
3027 Severity : enhancement
3029 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3030 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3031 boolean to control whether to print peer debug info when a
3032 client's RPC times out.
3037 Description: make O_SYNC working on 2.6 kernels
3038 Details : 2.6 kernels use different method for mark pages for write,
3039 so need add a code to lustre for O_SYNC work.
3044 Description: Failure to close file and release space on NFS
3045 Details : Put inode details into lock acquired in ll_intent_file_open.
3046 Use mdc_intent_lock in ll_intent_open to properly
3047 detect all kind of errors unhandled by mdc_enqueue.
3052 Description: proc file read during shutdown sometimes raced obd removal,
3054 Details : Add lock to prevent obd access after proc file removal.
3057 Frequency : Only for files larger than 4GB on 32-bit clients.
3059 Description: improperly doing page alignment of locks
3060 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3061 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3066 Description: RPCs being resent when they shouldn't be
3067 Details : Some RPCs that should not be resent are being resent. This
3068 can cause inconsistencies in the RPC state machine. Do not
3069 resend such requests.
3072 Frequency : rare, only with NFS export
3074 Description: Crash on NFS re-export node
3075 Details : under very unusual load conditions an assertion is hit in
3076 ll_intent_file_open()
3079 Frequency : only if OST filesystem is corrupted
3081 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3082 Details : for a short time RPCs with bulk IO are in the replay list,
3083 but replay of bulk IOs is unimplemented. If the OST filesystem
3084 is corrupted due to disk cache incoherency and then replay is
3085 started it is possible to trip an assertion. Avoid putting
3086 committed RPCs into the replay list at all to avoid this issue.
3091 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3092 Details : Large single O_DIRECT read and write calls can fail to allocate
3093 a sufficiently large buffer to process the request. In case of
3094 allocation failure the allocation is retried with a smaller
3095 buffer and broken into smaller requests.
3097 Severity : enhancement
3099 Description: Add -o localflock option to simulate old noflock behaviour.
3100 Details : This will achieve local-only flock/fcntl locks coherentness.
3105 Description: versioning check is incomplete
3106 Details : Checking the version difference of client vs. server, report
3107 error if the gap is too big.
3112 Description: add support PG_writeback bit
3113 Details : add support for PG_writeback bit for Lustre, for more carefull
3114 work with page cache in 2.6 kernel. This also fix some deadlocks
3115 and remove hack for work O_SYNC with 2.6 kernel.
3117 Severity : enhancement
3119 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3120 Details : The uninit_groups feature works in conjunction with the kernel
3121 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3122 up the pass1 processing of e2fsck. This is a read-only feature
3123 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3124 mount filesystems that have had this feature enabled.
3126 Severity : enhancement
3128 Description: Improve multi-block allocation algorithm to avoid fragmentation
3129 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3130 allocation locality and avoid filesystem fragmentation.
3132 ------------------------------------------------------------------------------
3134 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3136 * Support for kernels:
3137 2.6.9-42.0.3.EL (RHEL 4)
3138 2.6.5-7.276 (SLES 9)
3139 2.4.21-47.0.1.EL (RHEL 3)
3140 2.6.12.6 vanilla (kernel.org)
3141 2.6.16.21-0.8 (SLES10)
3142 * Recommended e2fsprogs version: 1.39.cfs2-0
3144 * The backwards-compatible /proc/sys/portals symlink has been removed
3145 in this release. Before upgrading, please ensure that you change
3146 any configuration scripts or /etc/sysctl.conf files that access
3147 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3148 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3149 in advance of the upgrade on any system running Lustre 1.4.6 or
3150 newer, since /proc/sys/lnet was added in that version.
3151 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3155 Frequency : only when quota is used
3157 Description: avoid scanning export list for quota master
3158 Details : Change the algorithms to avoid scanning export list in order
3159 to improve the efficiency.
3162 Frequency : MDS failover only, very rarely
3164 Description: "went back in time" messages on mds failover
3165 Details : The greatest transno may be lost when the current operation
3166 finishes with an error (transno==0) and the client's last_rcvd
3167 record is over-written. Save the greatest transno in the
3168 mds_last_transno for this case.
3171 Frequency : always for specific kernels and striping counts
3173 Description: client may get "Matching packet too big" without ACL support
3174 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3175 when trying to access files in certain configurations. The
3176 clients should in fact be denied when mounting because they do
3177 not understand ACLs.
3180 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3182 Description: many clients connecting with IO in progress causes connect timeouts
3183 Details : Avoid synchronous journal commits to avoid delays caused by many
3184 clients connecting/disconnecting when bulk IO is in progress.
3185 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3186 OST_IO_PORTAL to avoid delays behind potentially many pending
3190 Frequency : occasionally with multiple writers to a single file
3192 Description: shared writes to file may result in wrong size reported by stat()
3193 Details : Allow growing of kms when extent lock is cancelled
3196 Frequency : always with random mmap IO to multi-striped file
3198 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3199 Details : Only if the hole is at the end of OST object so that kms is too
3200 small. Fix is to increase kms accordingly in ll_nopage.
3203 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3205 Description: writes to a missing object would leak memory on the OST
3206 Details : If there is an inconsistency between the MDS and OST filesystems,
3207 such that the MDS references an object that doesn't exist, writes
3208 to that object will leak memory due to incorrect cleanup in the
3209 error handling path, eventually running out of memory on the OST.
3214 Description: Creating too long symlink causes lustre errors
3215 Details : Check symlink and name lengths before sending requests to MDS.
3218 Frequency : only if flock is enabled (not on by default)
3220 Description: posix locks not released on fd closure on 2.6.9+
3221 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3222 kernels, this caused such locks not to be released on fd close and
3223 then assertions on fs unmount about still used locks.
3226 Frequency : MDS failover only, very rarely
3228 Description: clients may get ASSERTION(granted_lock != NULL)
3229 Details : When request was taking a long time, and a client was resending
3230 a getattr by name lock request. The were multiple lock requests
3231 with the same client lock handle and
3232 mds_getattr_name->fixup_handle_for_resent_request found one of the
3233 lock handles but later failed with ASSERTION(granted_lock != NULL).
3238 Description: handle->h_buffer_credits > 0, assertion failure
3239 Details : h_buffer_credits is zero after truncate, causing assertion
3240 failure. This patch extends the transaction or creates a new
3244 Frequency : NFS re-export or patchless client
3245 Bugzilla : 11179, 10796
3246 Description: Crash on NFS re-export node (__d_move)
3247 Details : We do not want to hash the dentry if we don't have a lock.
3248 But if this dentry is later used in d_move, we'd hit uninitialised
3249 list head d_hash, so we just do this to init d_hash field but
3250 leave dentry unhashed.
3253 Frequency : NFS re-export or patchless client
3255 Description: NFS exports has problem with symbolic link
3256 Details : lustre client didn't properly install dentry when re-exported
3257 to NFS or running patchless client.
3260 Frequency : NFS re-export or patchless client
3262 Description: Various nfs/patchless fixes.
3263 Details : fixes reuse disconected alias for lookup process - this fixes
3264 warning "find_exported_dentry: npd != pd",
3265 fix permission error with open files at nfs.
3266 fix apply umask when do revalidate.
3269 Frequency : occasional
3271 Description: Crash on NFS re-export node
3272 Details : calling clear_page() on the wrong pointer triggered oops in
3273 generic_mapping_read().
3276 Frequency : rarely, using O_DIRECT IO
3278 Description: unaligned directio crashes client with LASSERT
3279 Details : check for unaligned buffers before trying any requests.
3282 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3284 Description: stale data returned from RAID cache
3285 Details : If only a small amount of IO is done to the RAID device before
3286 reading it again it is possible to get stale data from the RAID
3287 cache instead of reading it from disk.
3290 Frequency : always for sles10 kernel
3292 Description: sles10 support
3293 Details : ll_follow_link: compile fixes and using of nd_set_link
3294 under newer kernels.
3297 Frequency : depends on arch, kernel and compiler version, always on sles10
3300 Description: recursive or deep enough symlinks cause stack overflow
3301 Details : getting rid of large stack-allocated variable in
3305 Frequency : depends on hardware
3307 Description: lustre write performance loss in the SLES10 kernel
3308 Details : the performance loss is caused by using of write barriers in the
3309 ext3 code. The SLES10 kernel turns barrier support on by
3310 default. The fix is to undo that change for ldiskfs.
3312 ------------------------------------------------------------------------------
3314 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3316 * Support for kernels:
3317 2.6.9-42.0.3EL (RHEL 4)
3318 2.6.5-7.276 (SLES 9)
3319 2.4.21-47.0.1.EL (RHEL 3)
3320 2.6.12.6 vanilla (kernel.org)
3324 Frequency : quota enabled and large files being deleted
3326 Description: releasing more than 4GB of quota at once hangs OST
3327 Details : If a user deletes more than 4GB of files on a single OST it
3328 will cause the OST to spin in an infinite loop. Release
3329 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3334 Description: statfs data retrieved from /proc may be stale or zero
3335 Details : When reading per-device statfs data from /proc, in the
3336 {kbytes,files}_{total,free,avail} files, it may appear
3337 as zero or be out of date.
3340 Frequency : systems with MD RAID1 external journal devices
3342 Description: lconf's call to blkid is confused by RAID1 journal devices
3343 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3344 of returning all block devices that match the journal UUID.
3347 Frequency : always, for aggregate stripe size over 4GB
3349 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3350 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3351 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3354 Frequency : always if "lfs find" used on a local file/directory
3356 Description: "lfs find" segfaults if used on a local file/directory
3357 Details : The case where a directory component was not specified wasn't
3358 handled correctly. Handle this properly.
3361 Frequency : always on ppc64
3363 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3364 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3365 perform endianness conversion.
3368 Frequency : rarely (truncate to non-zero file size after write under load)
3369 Bugzilla : 10730, 10687
3370 Description: Files padded with zeros to next 4K multiple
3371 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3372 it is possible that files that are truncated to a non-zero size
3373 immediately after being written are filled with zero bytes beyond
3374 the truncated size. No file data is lost.
3376 Severity : enhancement
3378 Description: Allow recovery/failover for liblustre clients.
3379 Details : liblustre clients were unaware of failover configurations until
3382 Severity : enhancement
3384 Description: user file locks should fail when not mounting with flock option
3385 Details : Set up an error-returning stub in ll_file_operations.lock field
3386 to prevent incorrect behaviour when client is mounted without
3387 flock option. Also, set up properly f_op->flock field for
3391 Frequency : always on ia64
3393 Description: "lfs df" loops on printing out MDS statfs information
3394 Details : The obd_ioctl_data was not initialized and in some systems
3395 this caused a failure during the ioctl that did not return
3396 an error. Initialize the struct and return an error on failure.
3399 Frequency : SLES 9 only
3401 Description: Error of copying files with lustre special EAs as root
3402 Details : Client side always return success for setxattr call for lustre
3403 special xattr (currently only "trusted.lov").
3406 Frequency : rarely on clusters with both ia64+i386 clients
3408 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3409 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3410 clients are doing concurrent writes to the same file it is
3411 possible that the ia64 clients may LASSERT because the OST
3412 extent locks are not PAGE_SIZE aligned. Ensure that grown
3413 locks are always aligned on the request boundary.
3416 Frequency : specific use, occasional
3418 Description: Overwriting in use executable truncates on-disk binary image
3419 Details : If one node attempts to overwrite an executable in use by
3420 another node, we now correctly return ETXTBSY instead of
3421 truncating the file.
3423 Severity : enhancement
3425 Description: Async OSC create to avoid the blocking unnecessarily.
3426 Details : If a OST has no remain object, system will block on the creating
3427 when need to create a new object on this OST. Now, ways use
3428 pre-created objects when available, instead of blocking on an
3429 empty osc while others are not empty. If we must block, we block
3430 for the shortest possible period of time.
3435 Description: chmod on Lustre root is propagated to other clients
3436 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3437 invalid by the follow_mount time.
3440 Frequency : liblustre clients only
3442 Description: Race in 'instant cancel' lock handling could lead to such locks
3443 never to be granted in case of SMP MDS
3444 Details : Do not destroy not yet granted but cbpending locks in
3448 Frequency : replay/resend of open
3450 Description: non null lock assetion failure in mds_intent_policy
3451 Details : Trying to replay/resend lockless open requests resulted in
3452 mds_open() returning 0 with no lock. Now it sets a flag if
3453 a lock is going to be returned.
3455 Severity : enhancement
3457 Description: Checksum enhancements
3458 Details : New checksum enhancements allow for resending RPCs that failed
3461 Severity : enhancement
3463 Description: Tunables on number of dirty pages in cacche
3464 Details : Allow to set limit on number of dirty pages cached.
3469 Description: client crash on unmount - lock still has references
3470 Details : In some error handling cases it was possible to leak a lock
3471 reference on a client while accessing a file. This was not
3472 harmful to the client during operation, but would cause the
3473 client to crash when the filesystem is unmounted.
3476 Frequency : specific case, rare
3478 Description: ETXTBSY on mds though file not in use by client
3479 Details : ETXTBSY is no longer incorrectly returned when attempting to
3480 chmod or chown a directory that the user previously tried to
3481 execute or a currently-executing binary.
3484 Frequency : extremely rare except on liblustre-based clients
3486 Description: Lustre space not freed when files are deleted
3487 Details : Clean up open-unlinked files after client eviction. Previously
3488 the unlink was skipped and the files remained as orphans.
3493 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3494 Details : In some cases it was possible to send a blocking callback to
3495 a client doing a glimpse, even though that client didn't get
3496 a lock granted. When the glimpse lock is cancelled on the OST
3497 the freed lock is left on the waiting list and corrupted the list.
3500 Frequency : all core dumps
3502 Description: Broke core dumps to lustre
3503 Details : Negative dentry may be unhashed if parent does not have UPDATE
3504 lock, but some callers, e.g. do_coredump, expect dentry to be
3505 hashed after successful create, hash it in ll_create_it.
3507 ------------------------------------------------------------------------------
3509 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3511 * Support for kernels:
3512 2.6.9-42.0.2.EL (RHEL 4)
3513 2.6.5-7.276 (SLES 9)
3514 2.4.21-40.EL (RHEL 3)
3515 2.6.12.6 vanilla (kernel.org)
3519 Frequency : always on RHEL 3
3521 Description: Number of open files grows over time
3522 Details : The number of open files grows over time, whether or not
3523 Lustre is started. This was due to a filp leak introduced
3524 by one of our kernel patches.
3526 ------------------------------------------------------------------------------
3528 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3530 * Support for kernels:
3531 2.6.9-42.EL (RHEL 4)
3532 2.6.5-7.276 (SLES 9)
3533 2.4.21-40.EL (RHEL 3)
3534 2.6.12.6 vanilla (kernel.org)
3539 Bugzilla : 5719, 9635, 9792, 9684
3540 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3541 Details : If a server is under heavy load and cannot reply to new
3542 connection requests before the client resends the (re)connect,
3543 the connection handling code can behave badly if two service
3544 threads are concurrently handing separate (re)connections from
3545 the same client. Add better locking to the connection handling
3546 code, and ensure that only a single connection will be processed
3547 for a given client UUID, even if the lock is dropped.
3549 Severity : enhancement
3551 Description: add TCP zero-copy support to kernel
3552 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3553 sends if the hardware supports scatter-gather and checksumming.
3554 This allows socklnd to do client-write and server-read more
3555 efficiently and reduce CPU utilization from skbuf copying.
3558 Frequency : only if NFS exporting from client
3560 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3561 Details : If setting attributes on a file created under NFS that had
3562 never been opened it would be possible to oops the client
3563 if the file had no objects.
3566 Frequency : always for liblustre
3568 Description: liblustre client does MDS+OSTs setattr RPC for each write
3569 Details : When doing a write from a liblustre client, the client
3570 incorrectly issued an RPC to the MDS and each OST the file was
3571 striped over in order to update the timestamps. When writing
3572 with small chunks and many clients this could overwhelm the MDS
3573 with RPCs. In all cases it would slow down the write because
3574 these RPCs are unnecessary.
3576 Severity : enhancement
3578 Description: allow number of MDS service threads to be changed at module load
3579 Details : It is now possible to change the number of MDS service threads
3580 running. Adding "options mds mds_num_threads={N}" to the MDS's
3581 /etc/modprobe.conf will set the number of threads for the next
3582 time Lustre is restarted (assuming the "mds" module is also
3583 reloaded at that time). The default number of threads will
3584 stay the same, 32 for most systems.
3589 Description: OST crash if filesystem is unformatted or corrupt
3590 Details : If an OST is started on a device that has never been formatted
3591 or if the filesystem is corrupt and cannot even mount then the
3592 error handling cleanup routines would dereference a NULL pointer.
3597 Description: NULL pointer deref in llap_from_page.
3598 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3599 incorrect) mapping if the page was truncated/reclaimed while it was
3600 searched for. Check for this condition and skip such pages when
3601 doing readahead. Introduce extra check to llap_from_page() to
3602 verify page->mapping->host is non-NULL (so page is not anonymous).
3605 Frequency : Sometimes when using sys_sendfile
3607 Description: "page not covered by a lock" warnings from ll_readpage
3608 Details : sendfile called ll_readpage without right page locks present.
3609 Now we introduced ll_file_sendfile that does necessary locking
3610 around call to generic_file_sendfile() much like we do in
3614 Frequency : with certain MDS communication failures at client mount time
3616 Description: NULL pointer deref after failed client mount
3617 Details : a client connection request may delayed by the network layer
3618 and not be sent until after the PTLRPC layer has timed out the
3619 request. If the client fails the mount immediately it will try
3620 to clean up before the network times out the request. Add a
3621 reference from the request import to the obd device and delay
3622 the cleanup until the network drops the request.
3625 Frequency : occasionally during client (re)connect
3627 Description: assertion failure during client (re)connect
3628 Details : processing a client connection request may be delayed by the
3629 client or server longer than the client connect timeout. This
3630 causes the client to resend the connection request. If the
3631 original connection request is replied in this interval, the
3632 client may trip an assertion failure in ptlrpc_connect_interpret()
3633 which thought it would be the only running connect process.
3636 Frequency : only with obd_echo servers and clients that are rebooted
3638 Description: kernel BUG accessing uninitialized data structure
3639 Details : When running an obd_echo server it did not start the ping_evictor
3640 thread, and when a client was evicted an uninitialized data
3641 structure was accessed. Start the ping_evictor in the RPC
3642 service startup instead of the OBD startup.
3644 Severity : enhancement
3645 Bugzilla : 10193 (patchless)
3646 Description: Remove dependency on various unexported kernel interfaces.
3647 Details : No longer need reparent_to_init, exit_mm, exit_files,
3648 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3651 Frequency : rare (only users of deprecated and unsupported LDAP config)
3653 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3654 Details : LDAP apparently contains 'lustreName' attributes instead of
3655 'name'. A simple remapping of the name is sufficient.
3658 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3660 Description: waiting_locks_callback trips kernel BUG if client is evicted
3661 Details : Running with the dump_on_timeout debug flag turned on makes
3662 it possible that the waiting_locks_callback() can try to dump
3663 the Lustre kernel debug logs from an interrupt handler. Defer
3664 this log dumping to the expired_lock_main() thread.
3666 Severity : enhancement
3668 Description: Support NFS exporting on 2.6 kernels.
3669 Details : Implement non-rawops metadata methods for NFS server to use without
3670 changing NFS server code.
3673 Frequency : very rare (synthetic metadata workload only)
3675 Description: two racing renames might cause an MDS thread to deadlock
3676 Details : Running the "racer" program may cause one MDS thread to rename
3677 a file from being the source of a rename to being the target of
3678 a rename at exactly the same time that another thread is doing
3679 so, and the second thread has already enqueued these locks after
3680 doing a lookup of the target and is trying to relock them in
3681 order. Ensure that we don't try to re-lock the same resource.
3684 Frequency : only very large systems with liblustre clients
3686 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3687 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3688 all OSTs in parallel. This allows the eviction of stale liblustre
3689 clients to proceed much faster than if they were done in series,
3690 and also offers similar improvements for other set_info RPCs.
3695 Description: excessive CPU usage during initial read phase on client
3696 Details : During the initial read phase on a client, it would agressively
3697 retry readahead on the file, consuming too much CPU and impacting
3698 performance (since 1.4.5.8). Improve the readahead algorithm
3699 to avoid this, and also improve some other common cases (read
3700 of small files in particular, where "small" is files smaller than
3701 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3706 Description: MDS crash when receiving packet with unknown intent.
3707 Details : Do not LBUG in unknown intent case, just return -EFAULT
3709 Severity : enhancement
3710 Bugzilla : 9293, 9385
3711 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3712 Details : Do not serialize getattr (non-intent version) and statfs.
3715 Frequency : occasional, when OST network is overloaded/intermittent
3717 Description: client evicted by OST after bulk IO timeout
3718 Details : If a client sends a bulk IO request (read or write) the OST
3719 may evict the client if it is unresposive to its data GET/PUT
3720 request. This is incorrect if the network is overloaded (takes
3721 too long to transfer the RPC data) or dropped the OST GET/PUT
3722 request. There is no need to evict the client at all, since
3723 the pinger and/or lock callbacks will handle this, and the
3724 client can restart the bulk request.
3727 Frequency : Always when mmapping file with no objects
3729 Description: client crashes when mmapping file with no objects
3730 Details : Check that we actually have objects in a file before doing any
3731 operations on objects in ll_vm_open, ll_vm_close and
3737 Description: Request leak when working with deleted CWD
3738 Details : Introduce advanced request refcount tracking for requests
3739 referenced from lustre intent.
3741 Severity : Enhancement
3743 Description: Cache open file handles on client.
3744 Details : MDS now will return special lock along with openhandle, if
3745 requested and client is allowed to hold openhandle, even if unused,
3746 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3747 closing files for every read/write openration.
3749 Severity : Enhancement
3751 Description: Cache open negative dentries on client when possible.
3752 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3753 negative dentries on lock revocation.
3758 Description: Remounting a client read-only wasn't possible with a zconf mount
3759 Details : It wasn't possible to remount a client read-only with llmount.
3761 Severity : enhancement
3762 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3763 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3764 MPICH is included to add Lustre-specific ADIO interfaces.
3765 This is based closely on the UFS ADIO layer and only differs
3766 in file creation, in order to allow the OST striping to be set.
3767 This is user-contributed code and not supported by CFS.
3772 Description: extended inode attributes (immutable, append-only) work improperly
3773 when 2.4 and 2.6 kernels are used on client/server or vice versa
3774 Details : Introduce kernel-independent values for these flags.
3776 Severity : enhancement
3779 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3780 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3781 granularity. Now, Fractional MB granularity is be supported,
3782 this is very useful for low memory system.
3784 Severity : enhancement
3786 Description: Getattr by fid
3787 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3788 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3789 have locks and avoids getattr rpc requests for it on every stat.
3792 Frequency : Always, for filesystems larger than 2TB
3794 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3795 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3796 filesystems larger than 2TB to be mounted with mballoc enabled.
3799 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3801 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3802 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3803 filesystems larger than 2TB is mounted then the kernel will
3804 silently corrupt the start of the filesystem. CONFIG_LBD is
3805 enabled for all CFS-supported kernels, but the possibility of
3806 this happening with a modified kernel config exists.
3808 Severity : enhancement
3810 Description: add client O_DIRECT support for 2.6 kernels
3811 Details : It is now possible to do O_DIRECT reads and writes to files
3812 in the Lustre client mountpoint on 2.6 kernel clients.
3814 Severity : enhancement
3816 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3817 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3818 parallel, not waiting for response from every OST before sending
3819 a rpc to the next OST.
3824 Description: setattr vs write race when updating file timestamps
3825 Details : Client processes that update a file timestamp into the past
3826 right after writing to the file (e.g. tar) it is possible that
3827 the updated file modification time can be reset to the current
3828 time due to a race between processing the setattr and write RPC.
3830 Severity : enhancement
3832 Description: Bring 'lfs find' closer in line with regular Linux find.
3833 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3834 -print0 options and obtains all the needed info through the lustre
3837 Severity : enhancement
3839 Description: support up to 1024 configured devices on one node
3840 Details : change obd_dev array from statically allocated to dynamically
3841 allocated structs as they are first used to reduce memory usage
3846 Description: Flush dirty partially truncated pages during truncate
3847 Details : Immediatelly flush partially truncated pages in filter_setattr,
3848 this way we completely avoid having any pages in page cache on OST
3849 and can retire ugly workarounds during writes to flush such pages.
3854 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3855 Details : protect inode from truncation within vfs_unlink() context
3856 just take a reference before calling vfs_unlink() and release it
3857 when parent's i_sem is free.
3862 Description: last_id value checked outside lock on OST caused LASSERT failure
3863 Details : If there were multiple MDS->OST object precreate requests in
3864 flight, it was possible that the OST's last object id was checked
3865 outside a lock and incorrectly tripped an assertion. Move checks
3866 inside locks, and discard old precreate requests.
3869 Frequency : always, if extents are used on OSTs
3871 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3872 Details : OSTs using the extents format would not zero the high 16 bits of
3873 the index physical block number. This is not a problem for any
3874 OST filesystems smaller than 16TB, and no kernels support ext3
3875 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3876 new/modified files) and can be fixed for existing filesystems
3877 with e2fsprogs-1.39-cfs1.
3882 Description: import connection selection may be incorrect if timer wraps
3883 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3884 import connections to be ignored if the 32-bit jiffies counter
3885 wraps. Use a 64-bit jiffies counter.
3888 Frequency : very large clusters immediately after boot
3890 Description: LNET request buffers exhausted under heavy short-term load
3891 Details : If a large number of client requests are generated on a service
3892 that has previously never seen so many requests it is possible
3893 that the request buffer growth cannot keep up with the spike in
3894 demand. Instead of dropping incoming requests, they are held in
3895 the LND until the RPC service can accept more requests.
3898 Frequency : Sometimes during replay
3900 Description: Assertion failure in ll_local_open after replay.
3901 Details : If replay happened on an open request reply before we were able
3902 to set replay handler, reply will become not swabbed tripping the
3903 assertion in ll_local_open. Now we set the handler right after
3904 recognising of open request
3907 Frequency : very rare
3909 Description: kernel reports "badness in vsnprintf"
3910 Details : Reading from the "recovery_status" /proc file in small chunks
3911 may cause a negative length in lprocfs_obd_rd_recovery_status()
3912 call to vsnprintf() (which is otherwise harmless). Exit early
3913 if there is no more space in the output buffer.
3915 Severity : enhancement
3917 Description: clear OBD RPC statistics by writing to them
3918 Details : It is now possible to clear the OBD RPC statistics by writing
3919 to the "stats" file.
3924 Description: Client mtime is not the same on different clients after utimes
3925 Details : In some cases, the client was using the utimes() syscall on
3926 a file cached on another node. The clients now validate the
3927 ctime from the MDS + OSTs to determine which one is right.
3932 Description: Inability to activate failout mode
3933 Details : lconf script incorrectly assumed that in pythong string's numeric
3934 value is used in comparisons.
3937 Frequency : always with multiple stripes per file
3939 Description: Inefficient object allocation for mutli-stripe files
3940 Details : When selecting which OSTs to stripe files over, for files with
3941 a stripe count that divides evenly into the number of OSTs,
3942 the MDS is always picking the same starting OST for each file.
3943 Return the OST selection heuristic to the original design.
3948 Description: mount failures may take full timeout to return an error
3949 Details : Under some heavy load conditions it is possible that a
3950 failed mount can wait for the full obd_timeout interval,
3951 possibly several minutes, before reporting an error.
3952 Instead return an error as soon as the status is known.
3954 Frequency : quota enabled and large files being deleted
3956 Description: releasing more than 4GB of quota at once hangs OST
3957 Details : If a user deletes more than 4GB of files on a single OST it
3958 will cause the OST to spin in an infinite loop. Release
3959 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3964 Description: statfs data retrieved from /proc may be stale or zero
3965 Details : When reading per-device statfs data from /proc, in the
3966 {kbytes,files}_{total,free,avail} files, it may appear
3967 as zero or be out of date.
3970 Frequency : systems with MD RAID1 external journal devices
3972 Description: lconf's call to blkid is confused by RAID1 journal devices
3973 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3974 of returning all block devices that match the journal UUID.
3977 Frequency : always, for aggregate stripe size over 4GB
3979 Description: assertion fails when trying to use 4GB stripe size
3980 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3981 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3984 Frequency : always on ppc64
3986 Description: the first write on an ext3 filesystem with mballoc got stuck
3987 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3988 perform endianness conversion.
3990 ------------------------------------------------------------------------------
3992 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3994 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3995 INTEROPERATE with older versions automatically. Please read the
3996 user documentation before upgrading any part of a live system.
3997 * WARNING: Lustre networking configuration changes are required with
3998 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4001 * Support for kernels:
4002 2.6.9-22.0.2.EL (RHEL 4)
4003 2.6.5-7.244 (SLES 9)
4004 2.6.12.6 vanilla (kernel.org)
4007 Severity : enhancement
4008 Bugzilla : 7981/8208
4009 Description: Introduced Lustre Networking (LNET)
4010 Details : LNET is new networking infrastructure for Lustre, it includes
4011 a reorganized network configuration mode (see the user
4012 documentation for full details) as well as support for routing
4013 between different network fabrics. Lustre Networking Devices
4014 (LNDs) for the supported network fabrics have also been
4015 created for this new infrastructure.
4017 Severity : enhancement
4018 Description: Introduced Access control lists
4019 Details : clients can set ACLs on files and directories in order to have
4020 more fine-grained permissions than the standard Unix UGO+RWX.
4021 The MDS must be started with the "-o acl" mount option.
4023 Severity : enhancement
4024 Description: Introduced filesystem quotas
4025 Details : Administrators may now establish per-user quotas on the
4028 Severity : enhancement
4030 Description: Configuration change for the XT3
4031 The PTLLND is now used to run Lustre over Portals on the XT3
4032 The configure option(s) --with-cray-portals are no longer used.
4033 Rather --with-portals=<path-to-portals-includes> is used to
4034 enable building on the XT3. In addition to enable XT3 specific
4035 features the option --enable-cray-xt3 must be used.
4040 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4041 Details : A race between allocating a new llcd and re-getting the llcd_lock
4042 allowed another thread to grab newly-allocated llcd.
4044 Severity : enhancement
4046 Description: 2.6 OST async journal commit and locking fix to improve performance
4047 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4048 2.6 kernels are now async as they already were in 2.4 kernels so
4049 that they can commit concurrently with the network bulk transfer.
4050 For block-allocated files the filter allocation semaphore is held
4051 to avoid filesystem fragmentation during allocation. BKL lock
4052 removed for 2.6 xattr operations where it is no longer needed.
4057 Description: lconf incorrectly determined whether two IP networks could talk
4058 Details : In some more complicated routing and multiple-network
4059 configurations, lconf will avoid trying to make a network
4060 connection to a disjoint part of the IP space. It was doing the
4061 math incorrectly for one set of cases.
4066 Description: Fix for potential infinite loop processing records in an llog.
4067 Details : If an llog record is corrupted/zeroed, it is possible to loop
4068 forever in llog_process(). Validate the llog record length
4069 and skip the remainder of the block on error.
4072 Frequency : occasional (liblustre only)
4074 Description: liblustre could not open files whose last component is a symlink
4075 Details : sysio_path_walk() would incorrectly pass the open intent to
4076 intermediate path components.
4079 Frequency : rare (liblustre only with non-standard tuning)
4080 Bugzilla : 7201 (7350)
4081 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4082 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4083 after it has been released, only do this for referenced locks
4085 Severity : enhancement
4087 Description: specifying an (invalid) directory default stripe_size of -1
4088 would reset the directory default striping
4089 Details : stripe_size -1 was used internally to signal directory stripe
4090 removal, now use "all default" to signal dir stripe removal
4091 as a directory striping of "all default" is not useful
4094 Frequency : common for large clusters running liblustre clients
4096 Description: doing an ls when liblustre clients are running is slow
4097 Details : sending a glimpse AST to a liblustre client waits for every AST
4098 to time out, as liblustre clients will not respond. Since they
4099 cannot cache data we refresh the OST lock LVB from disk instead.
4101 Severity : enhancement
4103 Description: doing an ls at the same time as file IO can be slow
4104 Details : enqueue and other "small" requests can be blocked behind many
4105 large IO requests. Create a new OST IO portal for non-IO
4106 requests so they can be processed faster.
4109 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4111 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4112 created file with mode 000
4113 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4114 allow an ftruncate() as a non-root user to a file with mode 000.
4115 Lustre now respects this flag to disable mode checks when
4116 truncating a file owned by the user
4119 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4122 Description: Revoking locks from clients that went dead or catatonic might take
4124 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4125 cancellation of such locks instant on servers without waiting for
4126 any reply from clients. Clients drops these locks when cancel
4127 notification from server is received without replying.
4130 Frequency : liblustre-only, when liblustre client dies or becomes busy
4132 Description: Doing ls on Linux clients can take a long time with active
4134 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4135 granting such locks to it in the first place if possible. Locks
4136 are taken by proxy on the OST during the read or write and
4137 dropped immediately afterward. Add connect flags handling, do
4138 not grant locks to liblustre clients for glimpse ASTs.
4140 Severity : enhancement
4142 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4143 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4144 sequential reads and behaves badly with applications doing random
4145 reads. Improve it by reading ahead at least read region, and
4146 avoiding excessive large RPC for small reads.
4148 Severity : enhancement
4150 Description: Creating more than 1000 files for a single job may cause a load
4151 imbalance on the OSTs if there are also a large number of OSTs.
4152 Details : qos_prep_create() uses an OST index reseed value that is an
4153 even multiple of the number of available OSTs so that if the
4154 reseed happens in the middle of the object allocation it will
4155 still utilize the OSTs as uniformly as possible.
4160 Description: OST or MDS may oops in ping_evictor_main()
4161 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4162 but doesn't restart at beginning of obd_exports_timed list
4165 Severity : enhancement
4167 Description: improve by-nid export eviction on the MDS and OST
4168 Details : allow multiple exports with the same NID to be evicted at one
4169 time without re-searching the exports list.
4172 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4174 Description: MDS may oops in groups_free()
4175 Details : in rare race conditions a newly allocated group_info struct is
4176 freed again, and this can be NULL. The 2.4 compatibility code
4177 for groups_free() checked for a NULL pointer, but 2.6 did not.
4180 Frequency : common for liblustre clients doing little filesystem IO
4181 Bugzilla : 9352, 7313
4182 Description: server may evict liblustre clients accessing contended locks
4183 Details : if a client is granted a lock or receives a completion AST
4184 with a blocking AST already set it would not reply to the AST
4185 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4188 Frequency : lfs setstripe, only systems with more than 160 OSTs
4190 Description: unable to set striping with a starting offset beyond OST 160
4191 Details : llapi_create_file() incorrectly limited the starting stripe
4192 index to the maximum single-file stripe count.
4195 Frequency : LDAP users only
4197 Description: lconf did not handle in-kernel recovery with LDAP properly
4198 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4200 Severity : enhancement
4202 Description: bind OST threads to NUMA nodes to improve performance
4203 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4204 node and do their allocations there to localize memory access
4206 Severity : enhancement
4208 Description: llmount can determine client NID directly from Myrinet (GM)
4209 Details : the client NID code from gmnalnid was moved directly into
4210 llmount, removing the need to use this or specifying the
4211 client NID explicitly when mounting GM clients with zeroconf
4214 Frequency : if client is started with down MDS
4216 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4217 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4218 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4219 ptlrpc_import_delayed_req() because we didn't check if the
4220 request was interrupted, and we also didn't break out of the
4221 event loop if there was no timeout
4226 Description: data loss during non-page-aligned writes to a single file from
4227 both multiple nodes and multiple threads on one node at same time
4228 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4229 inconsistency led to false short-reads, that were cached and later
4230 used by ->prepare_write() to fill in partially written page,
4231 leading to data loss.
4234 Frequency : always, if lconf --abort_recovery used
4236 Description: lconf --abort_recovery fails with 'Operation not supported'
4237 Details : lconf was attempting to abort recovery on the MDT device and not
4240 Severity : enhancement
4242 Description: remove cleanup logs
4243 Details : replace lconf-generated cleanup logs with lustre internal
4244 cleanup routines. Eliminates the need for client-cleanup and
4247 Severity : enhancement
4249 Description: add support for EAs (user and system) on lustre filesystems
4250 Details : it is now possible to store extended attributes in the Lustre
4251 client filesystem, and with the user_xattr mount option it
4252 is possible to allow users to store EAs on their files also
4254 Severity : enhancement
4256 Description: Add possibility (config option) to show minimal available OST free
4258 Details : When compiled with --enable-mindf configure option, statfs(2)
4259 (and so, df) will return least minimal free space available from
4260 all OSTs as amount of free space on FS, instead of summary of
4261 free spaces of all OSTs.
4263 Severity : enhancement
4265 Description: do not expand extent locks acquired on OST-side
4266 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4267 by server: they are not cached anyway.
4270 Frequency : when mmap is used/binaries executed from Lustre
4272 Description: Unmmap pages before throwing them away from read cache.
4273 Details : llap_shrink cache now attempts to unmap pages before discarding
4274 them (if unmapping failed - do not discard). SLES9 kernel has
4275 extra checks that trigger if this unmapping is not done first.
4280 Description: lconf didn't resolve symlinks before checking to see whether a
4281 given mountpoint was already in use
4284 Frequency : when migrating failover services
4285 Bugzilla : 6395, 9514
4286 Description: When migrating a subset of services from a node (e.g. failback
4287 from a failover service node) the remaining services would
4288 time out and evict clients.
4289 Details : lconf --force (implied by --failover) sets the global obd_timeout
4290 to 5 seconds in order to quickly disconnect, but this caused
4291 other RPCs to time out too quickly. Do not change the global
4292 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4294 Severity : enhancement
4295 Frequency : if MDS is started with down OST
4296 Bugzilla : 9439,5706
4297 Description: Allow startup/shutdown of an MDS without depending on the
4298 availability of the OSTs.
4299 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4300 Add appropriate locking and lov-osc refcounts for safe
4301 cleaning. Add osc abort_inflight calls in case the
4302 synchronize never started.
4305 Frequency : occasional (Cray XT3 only)
4307 Description: root not authorized to access files in CRAY_PORTALS environment
4308 Details : The client process capabilities were not honoured on the MDS in
4309 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4310 been accessed by an authorized user then root was able to access
4311 the file on the local client also. The root user capabilities
4312 are now allowed on the MDS, as this environment has secure UID.
4315 Frequency : occasional
4317 Description: ldiskfs "too long searching" message happens too often
4318 Details : A debugging message (otherwise harmless) prints too often on
4319 the OST console. This has been reduced to only happen when
4320 there are fragmentation problems on the filesystem.
4325 Description: Division by zero in statfs when all OSCs are inactive
4326 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4327 lov_statfs divides by value returned by lov_get_stripecnt().
4331 Bugzilla : 9489, 3273
4332 Description: First write from each client to each OST was only 4kB in size,
4333 to initialize client writeback cache, which caused sub-optimal
4334 RPCs and poor layout on disk for the first writen file.
4335 Details : Clients now request an initial cache grant at (re)connect time
4336 and so that they can start streaming writes to the cache right
4337 away and always do full-sized RPCs if there is enough data.
4338 If the OST is rebooted the client also re-establishes its grant
4339 so that client cached writes will be honoured under the grant.
4344 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4345 Details : Now I/O RPCs go to different portal number and (presumably) fast
4346 lock requests (and glimses) and other RPCs get their own service
4347 threads pool that should be able to service those RPCs
4350 Severity : enhancement
4352 Description: Ability to exchange lustre version between client and servers and
4353 issue warnings at client side if client is too old. Also for
4354 liblustre clients there is ability to refuse connection of too old
4356 Details : New 'version' field is added to connect data structure that is
4357 filled with version info. That info is later checked by server and
4361 Frequency : rare, liblustre only.
4362 Bugzilla : 9296, 9581
4363 Description: Two simultaneous writes from liblustre at offset within same page
4364 might proceed at the same time overwriting eachother with stale
4366 Details : I/O lock withing llu_file_prwv was released too early, before data
4367 actually was hitting the wire. Extended lock-holding time until
4368 server acknowledges receiving data.
4371 Frequency : extremely rare. Never observed in practice.
4373 Description: avoid generating lustre_handle cookie of 0.
4374 Details : class_handle_hash() generates handle cookies by incrementing
4375 global counter, and can hit 0 occasionaly (this is unlikely, but
4376 not impossible, because initial value of cookie counter is
4377 selected randonly). Value of 0 is used as a sentinel meaning
4378 "unassigned handle" --- avoid it. Also coalesce two critical
4379 sections in this function into one.
4381 Severity : enhancement
4383 Description: allow liblustre clients to delegate truncate locking to OST
4384 Details : To avoid overhead of locking, liblustre client instructs OST to
4385 take extent lock in ost_punch() on client's behalf. New connection
4386 flag is added to handle backward compatibility.
4388 Severity : enhancement
4389 Bugzilla : 4928, 7341, 9758
4390 Description: allow number of OST service threads to be specified
4391 Details : a module parameter allows the number of OST service threads
4392 to be specified via "options ost ost_num_threads={N}" in the
4393 OSS's /etc/modules.conf or /etc/modprobe.conf.
4397 Bugzilla : 6146, 9635, 9895
4398 Description: servers crash with bad pointer in target_handle_connect()
4399 Details : In rare cases when a client is reconnecting it was possible that
4400 the connection request was the last reference for that export.
4401 We would temporarily drop the export reference and get a new
4402 one, but this may have been the last reference and the export
4403 was just destroyed. Get new reference before dropping old one.
4405 Severity : enhancement
4406 Frequency : if client is started with failover MDS
4408 Description: Allow multiple MDS hostnames in the mount command
4409 Details : Try to read the configuration from all specified MDS
4410 hostnames during a client mount in case the "primary"
4413 Severity : enhancement
4415 Description: Stop sending data to evicted clients as soon as possible.
4416 Details : Check if the client we are about to send or are sending data to
4417 was evicted already. (Check is done every second of waiting,
4418 for which l_wait_event interface was extended to allow checking
4419 of exit condition at specified intervals).
4422 Frequency : rare, normally only when NFS exporting is done from client
4424 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4426 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4429 Frequency : always, if srand() is called before liblustre initialization
4431 Description: Liblustre uses system PRNG disturbing its usage by user application
4432 Details : Introduce internal to lustre fast and high-quality PRNG for
4433 lustre usage and make liblustre and some other places in generic
4434 lustre code to use it.
4436 Severity : enhancement
4437 Bugzilla : 9477, 9557, 9870
4438 Description: Verify that the MDS configuration logs are updated when xml is
4439 Details : Check if the .xml configuration logs are newer than the config
4440 logs stored on the MDS and report an error if this is the case.
4441 Request --write-conf, or allow starting with --old_conf.
4443 Severity : enhancement
4445 Description: Handle symlinks in the path when checking if Lustre is mounted.
4446 Details : Resolve intermediate symlinks when checking if a client has
4447 mounted a filesystem to avoid duplicate client mounts.
4452 Description: lconf can hit an error exception but still return success.
4453 Details : The lconf command catches the Command error exception at the top
4454 level script context and will exit with the associated exit
4455 status, but doesn't ensure that this exit status is non-zero.
4460 Description: failure of ptlrpc thread startup can cause oops
4461 Details : Starting a ptlrpc service thread can fail if there are a large
4462 number of threads or the server memory is very fragmented.
4463 Handle this without oopsing.
4466 Frequency : always, only if liblustre and non-default acceptor port was used
4468 Description: liblustre cannot connect to servers with non-default acceptor port
4469 Details : tcpnal_set_default_params() was not called and was therefore
4470 ignoring the environment varaible TCPNAL_PORT, as well as other
4471 TCPNAL_ environment variables
4476 Description: two objects could be created on the same OST for a single file
4477 Details : If an OST is down, in some cases it was possible to create two
4478 objects on a single OST for a single file. No problems other
4479 than potential performance impact and spurious error messages.
4483 Bugzilla : 5681, 9562
4484 Description: Client may oops in ll_unhash_aliases
4485 Details : Client dcache may become inconsistent in race condition.
4486 In some cases "getcwd" can fail if the current directory is
4492 Description: Inode refcounting problems in NFS export code
4493 Details : link_raw functions used to call d_instantiate without obtaining
4494 extra inode reference first.
4498 Bugzilla : 9942, 9903
4499 Description: Referencing freed requests leading to crash, memleaks with NFS.
4500 Details : We used to require that call to ll_revalidate_it was always
4501 followed by ll_lookup_it. Also with revalidate_special() it is
4502 possible to call ll_revalidate_it() twice for the same dentry
4503 even if first occurence returned success. This fix changes semantic
4504 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4505 reference on a request referred from the intent.
4506 ll_intent_release() then releases such a request.
4509 Frequency : rare, normally benchmark loads only
4511 Description: unlinked inodes were kept in memory on the client
4512 Details : If a client is repeatedly creating and unlinking files it
4513 can accumulate a lot of stale inodes in the inode slab cache.
4514 If there is no other client load running this can cause the
4515 client node to run out of memory. Instead flush old inodes
4516 from client cache that have the same inode number as a new inode.
4519 Frequency : SLES9 2.6.5 kernel and long filenames only
4520 Bugzilla : 9969, 10379
4521 Description: utime reports stale NFS file handle
4522 Details : SLES9 uses out-of-dentry names in some cases, which confused
4523 the lustre dentry revalidation. Change it to always use the
4527 Frequency : rare, unless heavy write-truncate concurrency is continuous
4528 Bugzilla : 4180, 6984, 7171, 9963, 9331
4529 Description: OST becomes very slow and/or deadlocked during object unlink
4530 Details : filter_destroy() was holding onto the parent directory lock
4531 while truncating+unlinking objects. For very large objects this
4532 may block other threads for a long time and slow overall OST
4533 responsiveness. It may also be possible to get a lock ordering
4534 deadlock in this case, or run out of journal credits because of
4535 the combined truncate+unlink. Solution is to do object truncate
4536 first in one transaction without parent lock, and then do the
4537 final unlink in a new transaction with the parent lock. This
4538 reduces the lock hold time dramatically.
4541 Frequency : rare, 2.4 kernels only
4543 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4544 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4545 before unmounting their filesystem in order to avoid deadlock.
4546 The kernel_locked() function in 2.4 kernels only checks whether
4547 the kernel lock is held, not whether it is this process that is
4548 holding it as 2.6 kernels do.
4553 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4554 Details : The client ptlrpc code may be trying to reconnect to a down
4555 server before a previous connection attempt has timed out.
4556 Increase the reconnect interval to be longer than the connection
4557 timeout interval to avoid sending duplicate connections to
4561 Frequency : echo_client brw_test command
4563 Description: fix echo_client to work with OST preallocated code
4564 Details : OST preallocation code (5137) didn't take echo_client IO path
4565 into account: echo_client calls filter methods outside of any
4566 OST thread and, hence, there is no per-thread preallocated
4567 pages and buffers to use. Solution: hijack pga pages for IO. As
4568 a byproduct, this avoids unnecessary data copying.
4572 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4573 Description: Client can oops in mdc_commit_close() after open replay
4574 Details : It was possible for the MDS to return an open request with no
4575 transaction number in mds_finish_transno() if the client was
4576 evicted, but without actually returning an error. Clients
4577 would later try to replay that open and may trip an assertion
4578 Simplify the client close codepath, and always return an error
4579 from the MDS in case the open is not successful.
4582 Frequency : rare, 2.6 OSTs only
4584 Description: OST may deadlock under high load on fragmented files
4585 Details : If there was a heavy load and highly-fragmented OST filesystems
4586 it was possible to have all the OST threads deadlock waiting on
4587 allocation of biovecs, because the biovecs were not released
4588 until the entire RPC IO was completed. Instead, release biovecs
4589 as soon as they are complete to ensure forward IO progress.
4591 Severity : enhancement
4593 Description: Support for specifying external journal device at mount
4594 Details : If an OST or MDS device is formatted with an external journal
4595 device, this device major/minor is stored in the ext3 superblock
4596 and may not be valid for failover. Allow detecting and
4597 specifying the external journal at mount time.
4602 Description: Mounting an MDS with pending unlinked files may cause oops
4603 Details : target_finish_recovery() calls mds_postrecov() which returned
4604 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4605 considers this an error and immediately begins cleaning up the
4606 lov, just after starting the mds_lov process
4608 Severity : enhancement
4610 Description: Implement 'lfs df' to report actual free space on per-OST basis
4611 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4612 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4613 usage of objects; '-h' to report in human readable format.
4615 ------------------------------------------------------------------------------
4617 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4624 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4625 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4626 at startup during orphan unlinking. Instead, skip buddy update
4627 before setup, as it will be regenerated after recovery is complete.
4632 Description: If an OST is inactive, its locks might reference stale inodes.
4633 Details : lov_change_cbdata() must iterate over all namespaces, even if
4634 they are inactive to clear inode references from the lock.
4636 Severity : enhancement
4637 Frequency : occasional, if non-standard max_dirty_mb used
4639 Description: Client will block write RPCs if not enough grant
4640 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4641 then the client will block writes while waiting for another RPC
4642 to complete instead of consuming its dirty limit. With change
4643 we get improved performance when max_dirty_mb is small.
4645 Severity : enhancement
4646 Bugzilla : 3389, 6253
4647 Description: Add support for supplementary groups on the MDS.
4648 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4649 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4650 lookups for user supplementary groups into a cache.
4654 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4655 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4656 for liblustre O_EXCL is enforced by MDS.
4660 Description: readdir never returns NULL in liblustre.
4661 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4662 dentry in struct dirent.
4666 Frequency : liblustre only, depends on application IO pattern
4667 Description: liblustre clients evicted if not contacting servers
4668 Details : Don't put liblustre clients into the ping_evictor list, so
4669 they will not be evicted by the pinger ever.
4671 Severity : enhancement
4673 Description: Add ability to evict clients by NID from MDS.
4674 Details : By echoing "nid:$NID" string into
4675 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4676 $NID would be instantly evicted from this MDS and from all active
4677 OSTs connected to it.
4681 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4682 Details : lookup_it_finish() used to query file size from OSTs that was not
4687 Description: service threads change working directory to that of init
4688 Details : Starting lustre service threads may pin the working directory
4689 of the parent thread, making that filesystem busy. Threads
4690 now change to the working directory of init to avoid this.
4694 Frequency : during shutdown only
4695 Description: shutdown with a failed MDS or OST can cause unmount to hang
4696 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4701 Frequency : 2.6 only
4702 Description: chmod/chown may include an extra supplementary group
4703 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4704 supplementary group and if none is specified this is used.
4707 Bugzilla : 5479 (6816)
4709 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4710 Details : If lookup is in progress on a file that is unlinked we might try
4711 to revalidate the inode and fail in revalidate after lookup is
4712 complete and ll_file_open() enqueues the open again but
4713 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4716 Frequency : always, if lconf --abort_recovery used
4718 Description: lconf --abort_recovery fails with 'Operation not supported'
4719 Details : lconf was attempting to abort recovery on the MDT device and not
4722 ------------------------------------------------------------------------------
4724 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4729 Frequency : rare (only unsupported configurations with a node running as an
4731 Bugzilla : 6514, 5137
4732 Description: Mounting a Lustre file system on a node running as an OST could
4734 Details : OSTs now preallocates memory needed to write out data at
4735 startup, instead of when needed, to avoid having to
4736 allocate memory in possibly low memory situations.
4737 Specifically, if the file system is mounted on on OST,
4738 memory pressure could force it to try to write out data,
4739 which it needed to allocate memory to do. Due to the low
4740 memory, it would be unable to do so and the node would
4741 become unresponsive.
4743 Severity : enhancement
4745 Description: Addition of lconf --service command line option
4746 Details : lconf now accepts a '--service <arg>' option, which is
4747 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4749 Severity : enhancement
4751 Description: Failover mode is now the default for OSTs.
4752 Details : By default, OSTs will now run in failover mode. To return to
4753 the old behaviour, add '--failout' to the lmc line for OSTs.
4755 Severity : enhancement
4757 Description: Health checks are now provided for MDS and OSTs
4758 Details : Additional detailed health check information on MSD and OSTs
4759 is now provided through the procfs health_check value.
4762 Frequency : occasional, depends on IO load
4764 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4765 after numerous create/delete cycles
4766 Details : The ext3 inode allocation policy would not allocate new inodes
4767 very well on the OSTs because there are no new directories
4768 being created. Instead we look for groups with free space if
4769 the parent directories are nearly full.
4774 Description: Network or server problems during mount may cause partially
4775 mounted clients instead of returning an error.
4776 Details : The config llog parsing code may overwrite the error return
4777 code during mount error handling, returning success instead
4783 Description: MDS can fail to allocate large reply buffers
4784 Details : After long uptimes the MDS can fail to allocate large reply
4785 buffers (e.g. zconf client mount config records) due to memory
4786 fragmentation or consumption by the buffer cache. Preallocate
4787 some large reply buffers so that these replies can be sent even
4788 under memory pressure.
4792 Frequency : rare (liblustre)
4793 Description: fsx running with liblustre complained that using truncate() to
4794 extend the file doesn't work. This patch corrects that issue.
4795 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4796 ATTR_SIZE and lsm use in llu_setattr_raw.
4800 Frequency : rare, only 2.6 kernels
4801 Description: Unusual file access patterns on the MDS may result in inode
4802 data being lost in very rare circumstances.
4803 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4804 code in the 2.6 kernel caused the inode and/or EA data not to
4805 be read from disk, causing single-file corruption.
4809 Frequency : rare, only 2.6 filesystems using extents
4810 Description: Heavy concurrent write and delete load may cause data corruption.
4811 Details : It was possible under high-load situations to have an extent
4812 metadata block in the block device cache from a just-unlinked
4813 file overwrite a newly-allocated data block. We now unmap any
4814 metadata buffers that alias just-allocated data blocks.
4818 Frequency : filesystems with default stripe_count larger than 77
4819 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4820 Details : lconf specifies an inode size of 4096 bytes when the default
4821 stripe_count is larger than 77. This conflicts with the default
4822 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4823 this case to avoid pinning too much memory for large EAs.
4825 ------------------------------------------------------------------------------
4827 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4832 Frequency : rare (extremely heavy IO load with hundreds of clients)
4834 Description: Client is evicted, gets IO error writing to file
4835 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4836 caused clients to be evicted for AST timeouts. The fixes in
4837 bug 5192 mean we no longer need to have such short AST timeouts
4838 so ldlm_timeout has been increased.
4841 Frequency : occasional during --force or --failover shutdown under load
4842 Bugzilla : 5949, 4834
4843 Description: Server oops/LBUG if stopped with --force or --failover under load
4844 Details : a collection of import/export refcount and cleanup ordering
4845 issues fixed for safer force cleanup
4848 Frequency : only filesystems larger than 120 OSTs
4849 Bugzilla : 5990, 6223
4850 Description: lfs getstripe would oops on a very large filesystem
4851 Details : lov_getconfig used kfree on vmalloc'd memory
4854 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4855 Bugzilla : 6242, 6243
4856 Description: reading from files that had been truncated to a non-zero size
4857 but never opened returned no data
4858 Details : ll_file_read() reads zeros from no-object files to EOF
4863 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4864 Details : There's little harm in aborting MDS/OSS recovery and letting it
4865 try again, so I removed the LASSERT and return an error instead.
4867 Severity : enhancement
4869 Description: New debugging infrastructure for tracking down data corruption
4870 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4871 (b) cover more of the client-side code path, and (c) try to narrow
4872 down where problems occurred
4876 Bugzilla : 3819, 4364, 4397, 6313
4877 Description: Racing close and eviction MDS could cause assertion in mds_close
4878 Details : It was possible to get multiple mfd references during close and
4879 client eviction, leading to one thread referencing a freed mfd.
4881 Severity: : enhancement
4882 Bugzilla : 3262, 6359
4883 Description: Attempts to reconnect to servers are now more aggressive.
4884 Details : This builds on the enhanced upcall-less recovery that was added
4885 in 1.4.2. When trying to reconnect to servers, clients will
4886 now try each server in the failover group every 10 seconds. By
4887 default, clients would previously try one server every 25 seconds.
4892 Description: After recovery, certain operations trigger a failed
4893 assertion on a client.
4894 Details : Failing over an mds, using lconf -d --failover, while a
4895 client was doing a readdir() call would cause the client to
4896 LBUG after recovery completed and the readdir() was resent.
4898 Severity : enhancement
4900 Description: Default groups are now added by lconf
4901 Details : You can now run lconf --group <servicename> without having to
4902 manually add groups with lmc.
4905 Frequency : occasional
4907 Description: Nodes with an elan id of 0 trigger a failed assertion
4910 Frequency : always when accessing e.g. tty/console device nodes
4912 Description: tty and some other devices nodes cannot be used on lustre
4913 Details : file's private_data field is used by device data and lustre
4914 values in there got lost. New field was added to struct file to
4915 store fs-specific private data.
4918 Frequency : when exporting Lustre via NFS
4920 Description: NFSD failed occasionally when looking up a path component
4921 Details : NFSD is looking up ".." which was broken in ext3 directories
4922 that had grown large enough to become hashed.
4925 Frequency : Clusters with multiple interfaces not on the same subnet
4927 Description: Nodes will repeatedly try to reconnect to an interface which it
4928 cannot reach and report an error to the log.
4929 Details : Extra peer list entries will be created by lconf with some peers
4930 unreachable. lconf now validates the peer before adding it.
4933 Frequency : Only if a default stripe is set on the filesystem root.
4935 Description: Setting a default stripe on the filesystem root prevented the
4936 filesystem from being remounted.
4937 Details : The client was sending extra request flags in the root getattr
4938 request and did not allocate a reply buffer for the dir EA.
4941 Frequency : occasional, higher if lots of files are accessed by one client
4942 Bugzilla : 6159, 6097
4943 Description: Client trips assertion regarding lsm mismatch/magic
4944 Details : While revalidating inodes the VFS looks up inodes with ifind()
4945 and in rare cases can find an inode that is being freed.
4946 The ll_test_inode() code will free the lsm during ifind()
4947 when it finds an existing inode and then the VFS later attaches
4948 this free lsm to a new inode.
4952 Bugzilla : 6422, 7030
4953 Description: MDS deadlock between mkdir and client eviction
4954 Details : Creating a new file via mkdir or mknod (starting a transaction
4955 and getting the ns lock) can deadlock with client eviction
4956 (gets ns lock and trying to finish a synchronous transaction).
4959 Frequency : occasional
4960 Description: While starting a server, the fsfilt_ext3 module could not be
4962 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4963 kernels. Previously, lconf would still use the ext3 name
4964 when trying to load modules. Now, it will correctly use
4965 ext3 on 2.4 and ldiskfs on 2.6.
4967 Severity : enhancement
4968 Description: The default stripe count has been changed to 1
4969 Details : The interpretation of the default stripe count (0, to lfs
4970 or lmc) has been changed to mean striping across a single
4971 OST, rather than all available. For general usage we have
4972 found a stripe count of 1 or 2 works best.
4974 Severity : enhancement
4975 Description: Add support for compiling against Cray portals.
4976 Details : Conditional compiling for some areas that are different
4980 Frequency : occasional
4981 Bugzilla : 6409, 6834
4982 Description: Creating files with an explicit stripe count may lead to
4983 a failed assertion on the MDS
4984 Details : If some OSTs are full or unavailable, creating files may
4985 trigger a failed assertion on the MDS. Now, Lustre will
4986 try to use other servers or return an error to the
4990 Frequency : occasional
4992 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4993 caused lock timeout during readahead (since 1.4.2).
4994 Details : Processes doing readahead might match a lock that hasn't been
4995 granted yet if there are overlapping and conflicting lock
4996 requests. The readahead process waits on ungranted lock
4997 (original lock is CBPENDING), while OST waits for that process
4998 to cancel CBPENDING read lock and eventually evicts client.
5000 Severity : enhancement
5002 Description: Initial enabling of flock support for clients
5003 Details : Implements fcntl advisory locking and file status functions.
5004 This feature is provided as an optional mount flag (default
5005 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5006 locking are implemented yet, and those that are are not guaranteed
5007 to be completely correct in production environments.
5008 mount -t lustre -o [flock|noflock] ...
5011 Frequency : occasional
5013 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5014 assertion in the ext3 JBD (journaling) layer.
5015 Details : The b_committed_data struct is protected by the big kernel lock
5016 in 2.4 kernels, serializing journal_commit_transaction() and
5017 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5018 access to this struct. In 2.6 kernels there is finer grained
5019 locking to improve SMP performance of the JBD layer.
5023 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5028 Description: merge and cleanup kernel patches.
5031 -----------------------------------------------------------------------------
5033 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5035 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5036 versions of Lustre. Please update all servers and clients to
5037 version 1.4.2 or later at the same time. You must also run
5038 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5039 to update the configuration logs.
5041 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5042 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5043 - swab llog records of type '0' so we get proper header size/idx (5861)
5044 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5045 - fix rename of one directory over another leaking an inode (5953)
5046 - avoid SetPageDirty on 2.6 (5981)
5047 - don't re-add just-being-destroyed locks to the waiting list (5653)
5048 - when creating new directories, inherit the parent's custom
5049 striping settings if present parent (3048)
5050 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5051 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5052 - don't hold client locks on temporary worklist from l_lru (5666)
5053 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5054 - automatically evict dead clients (5921)
5055 - Update file size properly in create+truncate+fstat case (6196)
5056 - Do not unhash mountpoint dentries, do not allow removal of
5058 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5059 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5060 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5061 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5062 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5063 - Correctly handle reads of files with no objects (6243)
5064 - lctl recover will also mark a device active if deactivate used (5933)
5066 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5067 - allow --write-conf on an MDS with different nettype than client (5619)
5068 - don't write config llogs to MDS for mounts not from that MDS (5617)
5069 - lconf should create multiple TCP connections from a client (5201)
5070 - init scripts are now turned off by default; run chkconfig --on
5071 lustre and chkconfig --on lustrefs to use them
5072 - upcalls are no longer needed for clients to recover to failover
5074 - add --abort-recovery option to lconf to abort recovery on device
5076 - add support for an arbitrary number of OSTs (3026)
5077 - Quota support protocol changes.
5078 - forward compatibility changes to wire structs (6007)
5079 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5080 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5081 - improved "lustre status" script
5082 - initialize blocksize for non-regular files (6062)
5083 - added --disable-server and --disable-client configure options (5782)
5084 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5085 - Vanilla 2.4.29 support
5086 - increase maximum number of obd devices to 520 (6242)
5087 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5088 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5089 SLES 9 2.6.5 kernels
5090 - limit stripes per file to 160 (the maximum EA size) (6093)
5092 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5095 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5096 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5097 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5098 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5099 - small ext3 extents cleanups and fixes (5733)
5100 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5101 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5102 - avoid needless client->OST connect, fix handle mismatch (5317)
5103 - fix DLM error path that led to out-of-sync client, long delays (5779)
5104 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5105 - fix several locking issues related to i_size (5492,5624,5654,5672)
5106 - don't move pending lock onto export if it is already evicted (5683)
5107 - fix kernel oops when creating .foo in unlinked directory (5548)
5108 - fix deadlock in obdfilter statistics vs. object create (5811)
5109 - use time_{before,after} to avoid timer jiffies wrap (5882)
5110 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5111 - Do not leak request if server was not able to process it (5154)
5112 - If mds_open unable to find parent dir, make that negative lookup(5154)
5113 - don't create new directories with extent-mapping (5909, 5936)
5115 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5116 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5117 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5118 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5119 - support for RHEL4 kernel on i686 (5773)
5120 - provide error messages when incompatible logs are encountered (5898)
5122 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5123 * version 1.4.0.10 (1.4.1 release candidate 1)
5125 - don't keep a lock reference when lock is not granted (4238)
5126 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5127 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5128 - drop import inflight refcount on signal_completed_replay error (5255)
5129 - unlock page after async write error during send (3677)
5130 - handle missing objects in filter_preprw_read properly (5265)
5131 - no transno return for symlink open, don't save no-trasno open (3440)
5132 - don't try to complete elan receive that already failed (4012)
5133 - free RPC server reply state on error (5406)
5134 - clean up thread from ptlrpc_start_thread() on error (5160)
5135 - readahead could read extra page into cache that wasn't ejected (5388)
5136 - prevent races in class_attach/setup/cleanup/detach (5260)
5137 - don't dereference de->d_inode after l_dput of de (5458)
5138 - use "int" for stripe value returned from lock_to_stripe (5544)
5139 - mballoc allocation and error-checking fixes in 2.6 (5504)
5140 - block device patches to fix I/O request sizes in 2.6 (5482)
5141 - look up hostnames for IB nals (5602)
5142 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5143 - don't start multiple acceptors for the same port (5277)
5144 - fix incorrect LASSERT in mds_getattr_name (5635)
5145 - export a proc file for general "ping" checking (5628)
5146 - fix "lfs check" to not block when the MDS is down (5628)
5148 - service request history (4965)
5149 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5150 - create an "evict_client" /proc entry on OSTs, like the MDS has
5151 - fix mount usage message, return errors per mount(8) (5168)
5152 - change grep [] to grep "[]" in tests so they work in more UMLs
5153 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5154 - remove ancient LOV_MAGIC_V0 EA support (5047)
5155 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5156 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5157 - allow readahead allocations to fail when low on memory (5383)
5158 - mmap locking landed again, after considerable improvement (2828)
5159 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5161 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5164 - send OST transaction number in read/write reply to free req (4966)
5165 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5166 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5167 - the watchdog thread now runs as interruptible (5246)
5168 - flock/lockf fixes (but it's still disabled, pending 5135)
5169 - don't use EXT3 constants in llite code (5094)
5170 - memory shortage at startup could cause assertion (5176)
5172 - reorganization of lov code
5173 - single portals codebase
5175 - add extents/mballoc support (5025)
5176 - direct I/O reads in the obdfilter (4048)
5177 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5179 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5182 - send OST transaction number in read/write reply to free req (4966)
5183 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5184 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5185 - the watchdog thread now runs as interruptible (5246)
5186 - handle missing objects in filter_preprw_read properly (5265)
5187 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5188 - drop import inflight refcount on signal_completed_replay error (5255)
5189 - unlock page after async write error during send (3677)
5190 - return original error code on reconstructed replies (3761)
5191 - no transno return for symlink open, don't save no-trasno open (3440)
5193 - add pid to ldlm debugging output (4922)
5194 - bump the watchdog timeouts -- we can't handle 30sec yet
5195 - extra debugging for orphan dentry/inode bug (5259)
5197 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5200 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5201 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5202 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5203 - fix NULL dereference in /proc/sys/portals/routes (4827)
5204 - allow failed mdc_close() operations to be interrupted (4561)
5205 - stop precreate on OST before MDS would time out on it (4778)
5206 - don't send partial-page writes before EOF from client (4410)
5207 - discard client grant for sub-page writes on large-page clients (4520)
5208 - don't free dentries not owned by NFS code, check generation (4806)
5209 - fix lsm leak if mds_create_objects() fails (4801)
5210 - limit debug_daemon file size, always print CERROR messages (4789)
5211 - use transno after validating reply (3892)
5212 - process timed out requests if import state changes (3754)
5213 - update mtime on OST during writes, return in glimpse (4829)
5214 - add mkfsoptions to LDAP (4679)
5215 - use ->max_readahead method instead of zapping global ra (5039)
5216 - don't interrupt __l_wait_event() during strace
5218 - add software watchdogs to catch hung threads quickly (4941)
5219 - make lustrefs init script start after nfs is mounted
5220 - fix CWARN/ERROR duplication (4930)
5221 - return async write errors to application if possible (2248)
5222 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5223 - print NAL number in %x format (4645)
5224 - update barely-supported suse-2.4.21-171 series (4842)
5225 - support for sles 9 %post scripts
5226 - support for building 2.6 kernel-source packages
5227 - support for sles km_* packages
5229 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5232 - ignore -ENOENT errors in osc_destroy (3639)
5233 - notify osc create thread that OSC is being cleaned up (4600)
5234 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5235 - reconstruct ost_handle() like mds_handle() (4657)
5236 - create a new thread to do import eviction to avoid deadlock (3969)
5237 - let lconf resolve symlinked-to devices (4629)
5238 - don't unlink "objects" from directory with default EA (4554)
5239 - hold socknal file ref over connect in case target is down (4394)
5240 - allow more than 32000 subdirectories in a single directory (3244)
5241 - fix blocks count for O_DIRECT writes (3751)
5242 - OST returns ENOSPC from object create when no space left (4539)
5243 - don't send truncate RPC if file size isn't changing (4410)
5244 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5245 - bind to privileged port in socknal and tcpnal (3689)
5247 - rate limit CERROR/CWARN console message to avoid overload (4519)
5248 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5249 - basic mmap support (3918)
5250 - kernel patch series update from b1_4 (4711)
5252 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5255 - avoid crash during MDS cleanup with OST shut down (2775)
5256 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5257 - don't use bad inodes on the MDS (3744)
5258 - dynamic object preallocation to improve recovery speed (4236)
5259 - don't hold spinlock over lock dumping or change debug flags (4401)
5260 - don't zero obd_dev when it is force cleaned (3651)
5261 - print grants to console if they go negative (4431)
5262 - "lctl deactivate" will stop automatic recovery attempts (3406)
5263 - look for existing locks in ldlm_handle_enqueue() (3764)
5264 - don't resolve lock handle twice in recovery avoiding race (4401)
5265 - revalidate should check working dir is a directory (4134)
5267 - don't always mark "slow" obdfilter messages as errors (4418)
5269 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5272 - don't close LustreDB during write_conf until it is done (3860)
5273 - fix typo in lconf for_each_profile (3821)
5274 - allow dumping logs from multiple threads at one time (3820)
5275 - don't allow multiple threads in OSC recovery (3812)
5276 - fix debug_size parameters (3864)
5277 - fix mds_postrecov to initialize import for llog ctxt (3121)
5278 - replace config semaphore with spinlock (3306)
5279 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5280 - don't allow enqueue to complete on a destroyed export (3822)
5281 - down write_lock before checking llog header bitmap (3825)
5282 - recover from lock replay timeout (3764)
5283 - up llog sem before sending rpc (3652)
5284 - reduce ns lock hold times when setting kms (3267)
5285 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5286 - fix NULL deref and obd_dev leak on setup error (3312)
5287 - replace some LBUG about llog ops with error handling (3841)
5288 - don't match INVALID dentries from d_lookup and spin (3784)
5289 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5290 - fix invalid assertion in ptlrpc_set_wait (3880)
5292 - add libwrap support for the TCP acceptor (3996)
5293 - add /proc/sys/portals/routes for non-root route listing (3994)
5294 - allow setting MDS UUID in .xml (2580)
5295 - print the stack of a process that LBUGs (4228)
5297 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5300 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5301 - make sure to unset replay flag from failed open requests (3440)
5302 - if default stripe count is 0, use OST count for inode size (3636)
5303 - update parent mtime/ctime on client for create/unlink (2611)
5304 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5305 - free recovery state on server during a forced cleanup (3571)
5306 - unregister_reply for resent reqs (3063)
5307 - loop back devices mounting and status check on 2.6 (3563)
5308 - fix resource-creation race that can provoke i_size == 0 (3513)
5309 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5310 - more debugging for page-accounting assertion (3746)
5311 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5312 - avoid deadlock after precreation failure (3758)
5313 - fix race and lock order deadlock in orphan handling (3450, 3750)
5314 - add validity checks when grabbing inodes from l_ast_data (3599)
5316 - add /proc/.../recovery_status to obdfilter (3428)
5317 - lightweight CDEBUG infrastructure, debug daemon (3668)
5318 - change default OSC RPC parameters to be better on small clusters
5319 - turn off OST read cache for files smaller than 32MB
5320 - install man pages and include them in rpms (3100)
5321 - add new init script for (un)mounting lustre filesystems (2593)
5322 - run chkconfig in %post for init scripts (3701)
5323 - drop scimac NAL (unmaintained)
5325 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5328 - clean kiobufs before and after use (3485)
5329 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5330 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5331 - change init script to fail more clearly if not run as root (1528)
5332 - allow clients to reconnect during replay (1742)
5333 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5334 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5335 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5336 - fix cond_resched (3554)
5337 - only evict unfinished clients after recovery (3515)
5338 - allow bulk resend, prevent data loss (3570)
5339 - dynamic ptlrpc request buffer allocation (2102)
5340 - don't allow unlinking open directory if it isn't empty (2904)
5341 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5342 - remove extraneous obd dereference causing LASSERT failure (3334)
5343 - don't use get_cycles() when creating temp. files on the mds (3156)
5344 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5345 - handle EEXIST for set-stripe, set proper directory name (3336)
5347 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5348 - fix ksocknal_fmb_callback() error messages (2918)
5350 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5353 - don't copy lvb into (possibly NULL) reply on error (2983)
5354 - don't deref dentry after dput, don't free lvb on error (2922)
5355 - use the kms to determine writeback rpc length (2947)
5356 - increment oti_logcookies when osc is inactive (2948)
5357 - update client's i_blocks count via lvb messages (2543)
5358 - handle intent open/close of special files properly (1557)
5359 - mount MDS with errors=remount-ro, like obdfilter (2009)
5360 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5361 - don't use cancelling-locks' kms values (2947)
5362 - use highest lock extent for kms, not last one (2925)
5363 - don't dereference ERR_PTR() dentry in error handling path (3107)
5364 - fix thread race in portals_debug_dumplog() (3122)
5365 - create lprocfs device entries at setup instead of at attach (1519)
5366 - common AST error handler, don't evict client on completion race (3145)
5367 - zero nameidata in detach_mnt in 2.6 (3118)
5368 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5369 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5370 - initialize RPC timeout timer earlier for 2.6 (3219)
5371 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5372 - print nal/nid for unknown nid (3258)
5373 - additional checks for oscc recovery before doing precreate (3284)
5374 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5375 - don't crash in mdc_close for bad permissions on open (3285)
5376 - zero i_rdev for non-device files (3147)
5377 - clear page->private before handing to FS, better assertion (3119)
5378 - tune the read pipeline (3236)
5379 - fix incorrect decref of invalidated dentry (2350)
5380 - provide read-ahead stats and refine rpc in flight stats (3328)
5381 - don't hold journal transaction open across create RPC (3313)
5382 - update atime on MDS at close time (3265)
5383 - close LDAP connection when recovering to avoid server load (3315)
5384 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5385 - don't leak open file on MDS after open resend (3325)
5386 - serialize filter_precreate and filter_destroy_precreated (3329)
5387 - loop device shouldn't call sync_dev() for nul device (3092)
5388 - clear page cache after eviction (2766)
5389 - resynchronize MDS->OST in background (2824)
5390 - refuse to mount the same filesystem twice on same mountpoint (3394)
5391 - allow llmount to create routes for mounting behind routers (3320)
5392 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5393 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5394 - fix rare problem with rename on htree directories (3417)
5396 - allow default OST striping configuration per directory (1414)
5397 - fix compilation for qswnal for 2.6 kernels (3125)
5398 - increase maximum number of MDS request buffers for large systems
5399 - change liblustreapi to be useful for external progs like lfsck (3098)
5400 - increase local configuration timeout for slow disks (3353)
5401 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5403 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5406 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5407 - don't overwrite extent policy data in reply if lock was blocked (2901)
5408 - drop filter export grants atomically with removal from device (2663)
5409 - del obd_self_export from work_list in class_disconnect_exports (2908)
5410 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5411 - swab reply message in mdc_close, other PPC fixes (2464)
5412 - fix destroying of named logs (2325)
5413 - overwrite old logs when running lconf --write_conf (2264)
5414 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5415 - fix race in target_handle_connect (2898)
5416 - mds_reint_create() should take same inode create lock (2926)
5417 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5418 - don't close files for self_export to avoid uninitialized obd (2936)
5419 - allow MDS with the same name as client node (2939)
5420 - hold dentry reference for closed log files for unlink (2325)
5421 - reserve space for all logs during transactions (2059)
5422 - don't evict page beyond end of stripe extent (2925)
5423 - don't oops on a deleted current working directory (2399)
5424 - handle hard links to targets without a parent properly (2517)
5425 - don't dereference NULL lock when racing during eviction (2867)
5426 - don't grow lock extents when lots of conflicting locks (2919)
5428 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5431 - account for cache space usage on clients to avoid data loss (974)
5432 - lfsck support in lustre kernel code (2349)
5433 - reduce journal credits needed for BRW writes (2370)
5434 - orphan handling to avoid losing space on client/server crashes
5435 - ptlrpcd can be blocked, stopping ALL progress (2477)
5436 - use lock value blocks to assist in proper KMS, faster stat (1021)
5437 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5438 - recovery for initial connections (2355)
5439 - fixes for mds_cleanup_orphans (1934)
5440 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5441 - block all file creations until orphan recovery completes (1901)
5442 - client remove rq_connection from request struct (2423)
5443 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5444 - recovery timer race (2670)
5445 - mdc_close recovey bug (2532)
5446 - ptlrpc cleanup bug (2710)
5447 - mds timeout on local locks (2588)
5448 - namespace lock held during RPCs (2431)
5449 - handle interrupted sync write properly (2503)
5450 - don't try to handle a message that hasn't been replied to (2699)
5451 - client assert failure during cleanup after abort recovery (2701)
5452 - leak mdc device after failed mount (2712)
5453 - ptlrpc_check_set allows timedout requests to complete (2714)
5454 - wait for inflight reqs when ptlrpcd finishes (2710)
5455 - make sure unregistered services are removed from the srv_list
5456 - reset bulk XID's when resending them (caught by 1138 test)
5457 - unregister_bulk after timeout
5458 - fix lconf error (2694)
5459 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5460 - readahead locks pages, leaves pending causing memory pressure (2673)
5461 - increase OST request buffers to 4096 on large machines (2729)
5462 - fix up permission of existing directories in simple_mkdir (2661)
5463 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5464 - don't assign transno to errored transactions (2742)
5465 - don't delete objects on OST if given a bogus objid from MDS (2751)
5466 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5467 - if rq_no_resend, then timeout request after recovery (2432)
5468 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5469 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5470 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5471 - MDS recovery shouldn't delete objects that it has given out (2730)
5472 - if enqueue arrives after completion, don't clobber LVB (2819)
5473 - don't unlock pages twice when trigger_group_io returns error (2814)
5474 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5475 - don't write pages to disk if there was an error (1450)
5476 - don't ping imports that have recovery disabled (2676)
5477 - take buffered bytes into account when balancing socknal conn (2817)
5478 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5479 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5480 - remove little-endian swabbing of llog records (1987)
5481 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5482 - retry reposting request buffers if they fail (1191)
5483 - grow extent at grant time to avoid granting a revoked lock (2809)
5484 - lock revoke doesn't evict page if covered by a second lock (2765)
5485 - disable VM readahead to avoid reading outside lock extents (2805)
5487 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5488 - updated kernel patches for hp-2.4.20 kernel (2681)
5490 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5493 - fix truncated write corruption (2366)
5494 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5496 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5497 - ensure len(uuid) < 37 in lmc (1171)
5498 - fix ia64 OOPS in llog_test (2255)
5499 - zero end of page at obdfilter for partial page writes (2648)
5500 - don't leave stale dentries around after renames (bug 2428)
5501 - fix timeouts when evicting a client with a single lock held (2642)
5502 - set deadline for the initial HELLO message to drain (2634)
5503 - print out dotted-quad IP addresses in the socknal (2302)
5505 - additional debugging for MDS client eviction problem (2443)
5506 - fix mkfsoptions support for osts (2603, 2604)
5508 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5511 - add series for the vanilla 2.6.0 kernel
5512 - add series for the vanilla 2.4.24 kernel
5513 - add series for a cray x86/64 UL kernel drop
5514 - fix xattr patches for the vanilla 2.4.19 series
5516 - generate true UUIDs in lmc (1171)
5517 - have portals stack dumping break in UML (2466)
5518 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5519 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5520 - wake callback waiting threads on client eviction (2460)
5521 - Add --ptldebug and --subsystem to lmc (1719)
5522 - update assertion to allow safe interrupt allocation
5523 - set rq_no_resend for cancel requests (2432)
5524 - recalculate ptlrpcd timeout after resend (2494)
5525 - call vfs_rmdir when removing pending directories (2368)
5526 - fix renaming a file to itself (2429)
5527 - lmc creates a default one-stripe lov (2454)
5528 - expand procfs space to handle large clusters (2326)
5529 - increase UML stack to avoid overflow
5530 - update lconf's list of debug and subsystem masks
5531 - fix lfs find --obd (2510)
5532 - /proc tunable for disabling filter read caching (2591)
5533 - stop rpm packages from altering slapd.conf (2301)
5534 - disable nagle in the socknal under 0conf (2578)
5535 - choose mds inode size based on stripe count (2572)
5536 - fix kernel-source rpm problems (2516)
5538 - add --disable-doc to avoid pdf generation (2421)
5539 - update documentation, tests, type-os, comments
5540 - avoid format warnings on ia64
5541 - remove the TOE NAL
5542 - tiny code cleanups by removing unused fields
5544 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5547 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5548 - remove the most blatant lies from BUILDING file (2371)
5549 - change default debug level to reasonable production setting
5550 - reduce client side cache size to reduce cache flush time
5551 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5552 - make TCP zerocopy and pinger support enabled by default (2476)
5553 - sync writes completed after process exits caused crashes (2319)
5554 - maintain correct mount count on the MDS (2356)
5555 - backout 1557, because 2316 wasn't really fixed
5556 - better file I/O statistics gathering in /proc
5557 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5558 - another kernel patch to fix zero-copy TCP function export
5559 - don't take duplicate lock when processing re-sent getattr (2420)
5560 - lctl uses obd_self_export instead of creating new conn (2353)
5561 - MDS/OST recovery case which requires object creation asserted (2425)
5562 - move lfs from /usr/sbin to /usr/bin in packages
5563 - fix race between mds_client_add and mds_client_free (2417)
5564 - use kmalloc instead of slabs in portals (2430)
5565 - don't create duplicate records when a failover MDS is present (2442)
5566 - remove unnecessary mount age check (2332)
5567 - don't remove directory inodes from locks prematurely (2451)
5568 - don't break if MDS service name is the same as hostname (2103)
5569 - fix races in client write RPC generation when cache full (2482)
5571 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5574 - remove now-unused request->rq_obd (278)
5575 - if an allocation fails, print out how much memory we've used (1933)
5576 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5577 - add the "configurable stack size" patch to most series files (1256)
5578 - ability to write large log records, for 100+ OST configs (2306)
5579 - fix NULL deref when filter_prep fails (2314)
5580 - fix operator precedence error in filter_sync
5581 - dynamic allocation of socknal TX descriptors (2315)
5582 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5583 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5584 - ensure that CWARN messages in Portals always get to the syslog
5585 - __init/__exit are not for prototype decls (ldlm_init/exit)
5586 - x86-64 compile warning fixes
5587 - fix gateway LMC keyword conflict (2318)
5588 - fix MDS lock inversions in getattr/reint paths (1844)
5589 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5590 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5591 - DLM race condition prevented some lock evictions (2328)
5592 - ENOMEM detection and retry on socknal sends (2230)
5593 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5594 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5595 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5596 - better error messages when a client is rejected during recovery (1505)
5597 - avoid cancelling locks which were never granted, after failure (2330)
5598 - fix i_sem/journal inversion in mds_client_add (2333)
5599 - fix truncate/getattr lock cycle deadlock (2334)
5600 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5601 - fix two rare exit paths which could leak an l_lock() ref (2321)
5602 - fencepost error in MDS/OST orphan recovery (2226)
5603 - make log record alignment 8 bytes (1988)
5604 - lstripe now fails when requested offset > ost_count (2237)
5605 - ensure that all kernel series have a complete list.h (1607)
5606 - fix crashes in special-file operations (2316)
5607 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5609 - allow configurable automake binary, for testing new versions
5610 - small update to the lfs documentation
5612 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5614 * fix negative export reference count in fsfilt_sync (2312)
5616 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5617 * release candidate 0.9.1
5619 - orphans are moved into the PENDING directory for possible recovery
5620 - replayed opens now open by fid for orphan/rename safety (1042)
5621 - last close of an orphan inode generates a transno (683)
5622 - chdir() and mount() now pin the directory entry (1020)
5623 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5624 - discard very old requests without processing them (1502)
5625 - remove some common, well-understood CERRORs (1505)
5626 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5627 - clear "grant" flags in OST replies until OST grant code lands (1644)
5628 - fix read performance by not clobbering i_blksize on client (1598)
5629 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5630 - make LRU size a /proc tunable, clears locks when reduced (707)
5631 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5632 - server locks take references on exports now (1558)
5633 - build fixes for 2.4.20-rh trees (1663)
5634 - return an error from lov_create if all OSCs are inactive (1751)
5635 - fix import levels when a reconnect happens without a timeout (1597)
5636 - exit early from mds_open if we get a lookup error (1749)
5637 - partial page read at EOF wouldn't wait for disk before sending (1642)
5638 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5639 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5640 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5641 - write/truncate lock inversion (1639)
5642 - Don't auto-load obdclass, portals modules during cleanup (1495)
5643 - fix timestamps from jumping to "now" (1763)
5644 - extra journal assertions (1648)
5645 - add an extra multiunlink test (1771)
5646 - fix read_record/write_record API (1776)
5647 - fix leak of offset_extent, possible incorrect i_size later (1772)
5648 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5649 - Debugging for the kqswnal_get_idle_tx problems (1820)
5650 - Allow recovery to be attempted multiple times (1536)
5651 - Write out MDS last_rcvd file after it is first created (1600)
5652 - Fix tx_descriptor leak in failed transmit situations (1827)
5653 - ext3 journaling fixes for assertion failure after IO error (1871)
5654 - class_export_put() on freed export after completion AST error (1896)
5655 - Fix revalidate looping in VFS (1322)
5656 - Don't access a freed export during MDS_REINT timeout (1521)
5657 - Add open-unlink recovery support on the MDS (1673,1764)
5658 - Return an error if no MDS data was read from last_rcvd (1946)
5659 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5660 - Don't setup a disk device that doesn't match exported UUID (317)
5661 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5662 - avoid committing NULL handle in force close
5663 - local.sh is now a one-stripe LOV configuration
5664 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5665 - fix ext3 htree duplicate directory entry corruption (1516)
5666 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5667 - update to new LOV EA format (2097)
5668 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5670 2003-06-15 Phil Schwan <phil@clusterfs.com>
5673 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5674 - per-import/export recovery handling (958, 931, 959)
5675 - multiple last-rcvd slots, for serving multiple FSes (949)
5676 - connections are again shared between multiple imp/exports (963, 964)
5677 - "umount -f" would hang if any requests needed to be sent (393, 978)
5678 - avoid pinning large req buffer by copying for queued messages (989)
5679 - add "uuid" to "lctl device" command to help upcalls (991)
5680 - "open" RPCs with transnos would confuse recovery counters (1037)
5681 - do proper endian conversion of all wire messages (288, 340, 891)
5682 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5683 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5684 - fix ptlrpc_connection leak in target_handle_connect (1174)
5685 - fix import refcounting bug in OST and MDS cleanup (1134)
5686 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5687 - fix very unlikely obd_types race condition (501)
5688 - remove osc_open hack for echo_client (1187)
5689 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5690 - a failure in read_inode2 leads to deadlock (1139)
5691 - cancel ack-locks as soon as transaction is committed (1072)
5692 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5693 - make sure to commitrw after any preprw to avoid deadlock (1162)
5694 - failing to execute a file in a lustre FS would lock inode (1203)
5695 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5696 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5697 - obd_uuid2tgt would walk off the end of the list (1255)
5698 - on IA64 the osc would give portals incorrect bulk size (1258)
5699 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5700 - fix lock inversion caused by new llite matching code (1282)
5701 - limit the number of dirty pages on a client to 10MB (1286)
5702 - timed out locks were not being corrected cancelled (1289)
5703 - fix O_DIRECT above 4GB on IA-32 (1292)
5704 * major user-visible changes
5705 - fail out/fail over policy now controlled by the upcall (993)
5707 - add OBD_PING to check server availability and failure (954)
5708 - lustre messages are now sent in sending host order (288, 340, 891)
5709 - add eadatalen to MDS getattr reply (340)
5710 - OST read replies may contain second buffer, with per-page status (593)
5712 2003-03-11 Phil Schwan <phil@clusterfs.com>
5715 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5716 - failed open()s could cause deadlock; fixed (867, 869)
5717 - stop cancelling OST locks when files are closed (481)
5718 - overlapping XID spaces caused network corruption (851, 853)
5719 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5720 - setattr_raw added, to do single-RPC, server-side setattrs
5721 - lmc/lconf syntax change for OST UUIDs
5722 - fix crashy race condition between ptlrpc_free_req and osc_close
5723 - don't use request in mdc_enqueue if we hit a timeout (889)
5724 - don't set the inode i_size for regular files from the MDS (896)
5725 - handle out of order completion AST (842)
5726 - don't LBUG if a lock request times out after receiving AST (913)
5727 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5728 - if a bad lock AST arrives, send an error instead of dropping entirely
5729 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5730 - fix leak in bulk IO when only partially completed (899, 900, 926)
5731 - fix O_DIRECT for ia64 (55)
5732 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5733 - C-z after timeout could hang a process forever; fixed (977)
5735 - client-side I/O cache (678, 924, 929, 941, 970)
5737 - READPAGE and SETATTRs which don't take server-side locks get
5740 2003-02-11 Phil Schwan <phil@clusterfs.com>
5743 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5744 - Fix fsfilt_extN_readpage() to read a full page of directory
5745 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5746 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5747 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5748 - Most intent operations were removed, in favour of a new RPC mode
5749 that does a single RPC to the server and bypasses most of the VFS
5750 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5751 - Aggressively cancel local locks on DLM servers
5752 - mds_reint_unlink sends EA to the client if it's the last nlink.
5753 client uses that EA to unlink OST objects.
5754 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5755 - recursive symlinks were fixed (439)
5756 - fixed NULL deref in DEBUG_REQ
5757 - filter_update_lastobjid no longer calls sync, which annoyed extN
5758 - fixed multi-client small-writes to a single file problem (445)
5759 - fixed mtime updates during file writes (607)
5760 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5761 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5762 - fixed Linux OST size reporting problem (444, 656)
5763 - OST now updates object mtime with writes or setattr (607, 619)
5764 - client verifies file size before zeroing page past EOF (445)
5765 - OST now writes last allocated objid to disk with allocation (108)
5766 - LOV on echo now works (409)
5768 - mds_reint_unlink sends a new buffer, with the EA included. this
5769 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5770 set if a regular file was being unlinked, and it was the last link
5771 - use PtlGet from the target for bulk writes (315)
5772 - OST now updates object mtime with writes or setattr (607, 619)
5773 - LDLM now has a grant-time callback to revalidate locked items, if
5775 - Many MDS operations were reorganized to combat race conditions
5777 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5778 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5780 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5783 - Fully reactivate OST imports after reconnection (512, others)
5784 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5785 - More graceful error handling for truncating on dead OST (515)
5786 - Don't error out unless we're actually accessing dead stripes (474)
5787 - Fix garbage sizes when stripes are missing (410)
5788 - LRU counters were broken, causing constant lock purge (433, 432)
5789 - garbage on read from stripes with failed OSTs (441)
5790 - mark OSCs as active before reconnecting during recovery (438)
5791 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5792 - lfind did not preserve OST order in output (443)
5793 - symlinks cause hung clients, incorrect data (439)
5794 - stop dereferencing request after dropping refcount (457)
5795 - don't LASSERT(spin_is_locked) on non-SMP (455)
5796 - fixes for many rename() bugs
5797 - fstat didn't correctly synchronize attributes (399)
5798 - server must handle lock cancellation during blocking AST prep (487)
5799 - bulk descriptors were free()d too soon (511)
5800 - fix paths in lconf, which would load incorrect modules (451, 507)
5801 - fix confusing lconf 'host not found' error message (386)
5802 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5803 - fix race condition in mdc_blocking_ast() for inode access (526)
5804 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5805 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5806 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5807 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5808 - fix lstripe in several ways (kernel side) (527)
5809 - fix request leak in ldlm_cli_enqueue (262)
5810 - incorrect OSC was marked inactive after OST failure
5811 - call mds_fs_cleanup before unmounting filesystem (524)
5812 - fix races between taking ns_lock and ldlm_lock_change_resource
5813 - fix races updating LOV export open file list
5814 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5815 - fix recovery NULL deref in ldlm_cli_cancel_unused
5816 - fix some DLM races by using new hash table for lock handles (419)
5817 - permit the client to specify desired inodes, at replay
5818 - duplicate requests when we queue them for replay reintegration
5819 - fix last_rcvd offset calculation
5820 - sync after each recovered transaction, so we always make progress
5821 - never, not always, ERESTART requests without transnos
5822 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5824 - skip replay if the MDS says that the client is already connected
5825 - don't check for a recovery-enabled export to match lctl's UUID
5826 - don't INC_USE_COUNT for phantom exports
5827 - don't crash when cleaning up phantom exports (567)
5828 - don't double-finish or set replay data for errored mdc_open requests
5829 - abort requests when they time out, so we don't get old replies
5830 - send/receive replies for AST messages again
5831 - if the client says that it doesn't have the lock, cancel it on the
5833 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5834 mark it as destroyed, it will all work out when decref is called
5835 - fix module use counts (22, 581)
5837 - ASTs now expect a reply (server cancels lock on error reply)
5839 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5842 - fix many simultaneous client startup (392)
5843 - fix dentry->d_it clobbering
5844 - credentials weren't being shipped for readdir/getattr operations
5845 - remove invalid assertions triggered during some concurrent MD
5847 - proper Lustre versions added (336, 389)
5848 - fix memory leak for create error case (398)
5849 - fix LOV locking bug that would get cli/srv out of sync
5850 - fix echo client over LOV (409)
5851 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5852 - fix double-O_EXCL intent crash (424)
5853 - avoid sending multiple lock CANCELs (352)
5855 - MDS can do multi-client recovery (modulo bugs in new code)
5857 - many updates, edits, cleanups
5859 2002-11-18 Phil Schwan <phil@clusterfs.com>
5862 - fix null d_it dereference (346)
5863 - fix full OST/dbench hang (333)
5864 - fix permission problem with file removal (286)
5865 - fix removal of OSCs from LOV when they fail
5866 - fix NULL deref during bulk timeout (214)
5867 - fix problems related to multiple filesystems on one MDS (241)
5868 - fixed serious subtle metadata locking bugs
5869 - free locks on clients when inodes are removed due to memory
5871 - fix inode pointer in lock data (285)
5872 - partial support for multiple MDS on a single host (241)
5873 - data locks weren't cancelled at clear_inode time (290, 311)
5874 - intent locks could lead to unbounded lock growth (205)
5875 - added a maximum lock count, an LRU list, and a flusher
5876 - fix multiple rename (365)
5877 - properly abstracted the echo client
5878 - OSC locked 1 byte too many; fixed
5879 - rewrote brw callback code:
5880 - fixed recovery bugs related to LOVs (306)
5881 - fixed too-many-pages-in-one-write crash (191)
5882 - fixed (again) crash in sync_io_timeout (214)
5883 - probably fixed callback-related race (385)
5885 - Add capability to MDS protocol
5886 - LDLM cancellations and callbacks on different portals
5888 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5891 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5892 - defer open object destruction to close time (601981, 138)
5893 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5894 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5895 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5896 - don't call obd_setattr in ll_file_release for destroyed objects
5897 * protocol change to lustre_msg: move |version| and add |flags|
5898 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5899 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5900 * added replay of create, unlink, link and rename operations during
5901 MDS failover; recovery should be much more robust now
5902 * remove failed OSCs from LOVs (only lov_create uses this so far)
5903 * the lustre-HOWTO was brought (more) up to date (582544)
5905 2002-10-23 Phil Schwan <phil@clusterfs.com>
5908 - in-use dentries weren't being reused properly (617851)
5909 - prevent multiple LDLM setup (599178)
5910 - fix LOV size calculations for truncate (617853)
5911 - fix client handling of MDS intent errors (POSIX)
5912 - fix permission bug in lovstripe.c test (624321)
5913 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5914 - truncate past end of file could corrupt data
5915 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5916 - a race in recovery could return ETIMEDOUT to apps (623947)
5917 - building outside the source directory was fixed
5918 * the lustre-HOWTO was brought (more) up to date (582544)
5919 * major progress was made on recovery functionality
5921 2002-10-10 Phil Schwan <phil@clusterfs.com>
5924 - recovery deadlock fix
5925 - rm -rf causes LBUG fix (617817)
5926 - file open by multiple tasks fix (618962)
5927 - directory permissions bugs (602707 and 620007)
5928 - journal_stop fixed with locking (611313)
5929 - O_APPEND failures resolved (618273, perhaps 614459)
5930 - lconf PATH fix (619770)
5931 - IA64 build fix (621450)
5932 - RPC buffer sizes scale with amount of memory
5934 2002-10-01 Phil Schwan <phil@clusterfs.com>
5937 - locks would be cancelled without throwing away data pages,
5938 resulting in inconsistent data (605627)
5939 - inode attributes were not always being refreshed (605627, 612449)
5940 - lconf now continues to cleanup after lctl reports an error
5941 - MDS now enforces user permissions (602707)
5942 - lprocfs cleanup fixed, but not yet enabled (614157)
5943 - fixed infinite server hang, should a client not respond to an AST
5944 - avoid going into recovery if user calls readlink() with a buffer
5945 that's too small (613941)
5946 - AST RPCs no longer require replies (614867) -- this may be changed
5947 - don't crash server if client sends an IOV that's too big (611336)
5948 - fixed lock conversion deadlock (611892)
5949 - fixed the following of symlinks (614622)
5950 * recovery: the server can remove locks from a client that dies, other
5951 clients can make progress
5952 * more extN patch fixes
5953 * compile-time configurable ptlrpc buffer allocations
5955 - collaborative read cache document
5956 - Lustre Lite Performance CDR document-in-progress
5958 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5961 - fix typo in patch-2.4.18
5963 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5966 - clear ptlrpc request each time in handle_incoming_request()
5967 - unlink of files now destroys the object on the OST
5969 2002-09-19 Peter Braam <braam@clusterfs.com>
5971 * add hard link support
5972 * change obdfile creation method
5973 * kernel patch changed
5975 2002-09-19 Peter Braam <braam@clusterfs.com>
5978 - stack overflow bug in extN fixed
5980 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5982 * documentation updates
5983 - add man pages for config tools
5984 - update tests/README to describe testing with new config tools
5985 - finish metadata API descriptions
5986 * bug fixes and cleanups
5987 - statfs workaround for 16TB limit
5988 - LOV stripe allocation improved, can stripe on subset of OSTs
5989 - LOV file size/IO offset was wrong for files > 4GB in size
5990 - object EA data was being dropped, caused files to be unreadable
5991 - memory overflow with non-LOV OST caused memory corruption
5992 - fixed regression tests to work with new config tools, obdfilter
5993 - fixed bug when directory size became larger than 1 block
5994 - fixed bug (for single client case) when PWD was deleted
5995 - invalidate local directory pages when doing intent-based ops
5996 - avoid LDLM oops when lock callback contained bad data
5998 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6000 * documentation updates
6001 * bug fixes and cleanups
6002 - configuration tools
6005 - 64-bit compile warnings
6006 - 64-bit internal statfs data
6008 * test_brw on persistent OST devices
6010 * lprocfs (disabled)
6012 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6014 * documentation updates
6015 * bug fixes and cleanups
6016 * configuration tools
6018 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6021 * many small fixes to 0_5_4
6022 * io/network handling
6023 * thinkos in MDS operations
6025 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6028 * crucial basic fixes to 0.5.3
6029 * IOR, Iozone work over Elan
6032 2002-08-07 Phil Schwan <phil@clusterfs.com>
6033 * version 0_5_3, our first alpha
6034 * we use the new Portals iovs
6035 * documentation updates
6036 * bug fixes and cleanups
6037 * small changes in the DLM wire protocol
6039 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6040 * version 0_5_1 with some initial stability,
6041 * locking on MD and file I/O.
6042 * documentation updates
6043 * several bug fixes since 0.5.0
6044 * small changes in wire protocol
6046 2002-07-18 Phil Schwan <phil@clusterfs.com>
6048 * delivered as Lustre Light Alpha
6049 * fixed a crash after handling invalid MDS requests
6050 * fixed directory pages for architectures with non-4k pages sizes
6052 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6053 * release version v0_4_4
6054 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6056 * Fixes a number of interruption problems with OST operations.
6057 * Update documentation for portals header changes
6058 * Move all wire protocol structs/defines to lustre_idl.h
6059 * Fixes symlink length bug.
6060 * Add tcpdump to repository.
6062 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6063 * release version v0_4_3
6064 * Fixes statfs for inodes on extN.
6065 * Fixes bug in runtests which would delete /etc/hosts.
6066 * Use 64-bit object IDs wherever possible (not into VFS though)
6067 Remove ost_get_info, which is unused by lustre, and out of date.
6069 2002-07-03 Peter Braam <braam@clusterfs.com>
6070 * release version v0_4_2 Fixes a lookup error (type not passed)
6071 * move forward to head of Portals
6072 * move forward to latest Lustre kernel
6074 2002-06-25 Peter Braam <braam@clusterfs.com>
6075 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob