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).
19 Description: Unify req format on client/servers
20 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
21 buffers and swabbers approach..
26 Description: drop join file stuff.
28 Severity : enhancement
30 Description: Add 2.6.27 and SLES11 (patchless client) support
32 Severity : enhancement
34 Description: Adjust locks' extents on their first enqueue, so that at the time
35 they get granted, there is no need for another pass through the
36 queues since they are already shaped into the proper forms.
40 Description: Fix in ptlrpc_expire_one_request() to print the signed time
43 Severity : enhancement
45 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
48 Severity : enhancement
50 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
52 Severity : enhancement
54 Description: correctly handle big reply message.
55 Details : send LNet event if reply is bigger then buffer and adjust this buffer
60 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
62 Severity : enhancement
64 Description: Add support for OEL5.
68 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
72 Description: client could not reconnect to OST because of an active request.
73 Details : The client now retries to reconnect to the same server, if a
74 connect request failed with EBUSY or -EAGAIN. Test case 35b
75 conf-sanity.sh has been created.
79 Description: Don't allow make backward step on assignin osc next id.
80 Details : race between allocation next id and ll_sync thread can be cause
81 of set wrong osc next id and can be kill valid ost objects.
85 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
89 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
93 Description: Error handling in osc_statfs_interpret() has been improved.
94 Details : Check in osc_statfs_interpret() for EBADR.
98 Description: Do not send statfs() requests to OSTs disabled by administrator.
99 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
101 Severity : enhancement
103 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
105 Severity : enhancement
107 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
109 Severity : enhancement
111 Description: Enhancement for lfs(1) command to use numeric uid/gid.
115 Description: Improved cleanup procedure for conf-sanity test_46a.
119 Description: Do not start bulk transfer, if client tried to reconnect.
120 Details : Do not start bulk transfer, if client tried to reconnect,
121 but failed due to active rpcs. Check for exp_abort_active_req
122 has been implemented.
126 Description: Prevent inconsistences between linux and lustre mount structures.
127 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
128 Make the sleep interruptible.
130 Severity : enhancement
132 Description: provide server to client comms path
133 Details : server to client communications path via new LDLM_SET_INFO rpc
137 Description: truncate starts GFP_FS allocation under transaction causing deadlock
138 Details : ldiskfs_truncate calls grab_cache_page which may start page
139 allocation under an open transaction. This may lead to
140 calling prune_icache with consequent lustre reentrance.
144 Description: send correctly lsm on open replay
145 Details : MDS is trust to LSM size on replay open, but client can set wrong size
150 Description: enable client interface failover
151 Details : When a child reconnects from another NID, properly update export
152 nid hash position and ldlm reverse import.
154 Severity : enhancement
156 Description: Communicate OST degraded/readonly state via statfs to MDS
157 Details : Flags in the statfs returned from OSTs indicate whether the
158 OST is in a degraded RAID state, or if the filesystem has
159 turned read-only after a filesystem error is detected.
164 Description: don't panic if EPROTO was hit when reading symlink
165 Details : correctly handling request reference in error cases.
167 Severity : enhancement
169 Description: Add LustreNetLink, a kernel-userspace communcation path.
171 Severity : enhancement
173 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
178 Description: access to llog context before init.
179 Details : move handling CATALOGS file at osc layer and forbid access to llog
183 Frequency : rare, only when client is writing to OST on local node
185 Description: Avoid deadlock for local client writes
186 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
187 memory freeing context. This allows OST threads to set the
188 PF_MEMALLOC flag on task structures in order to allocate memory
189 from reserved pools and complete IO.
190 Use GFP_HIGHUSER for OST allocations for non-local client writes,
191 so that the OST threads generate memory pressure and allow
192 inactive pages to be reclaimed.
194 Severity : enhancement
196 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
201 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
202 Details : move ldlm namespace creation in setup phase to avoid grab
203 _lprocfs_lock with cli_sem held.
207 Description: Temporarily disable grant shrink.
208 Details : Disable the feature for debugging.
211 Frequency : only when formatting test filesystems
213 Description: Unable to run several mkfs.lustre on loop devices at the same time
214 Details : mkfs.lustre returns error 256 on the concurrent loop devices
215 formatting. The solution is to proper handle the error.
217 Severity : enhancement
219 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
221 Severity : enhancement
223 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
228 Description: resolve race between obd_disconnect and class_disconnect_exports
229 Details : if obd_disconnect will be called to already disconnected export he
230 forget release one reference and osc module can't unloaded.
232 Severity : enhancement
234 Description: Allow tuning service thread via /proc
235 Details : For each service a new
236 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
237 created that can be used to set min/max thread counts, and get the
238 current number of running threads.
243 Descriptoin: don't return error if have partially created objects for file.
244 Details : lov_update_create_set uses set->set_success as index for created
245 objects, so if some requests will be failed, they will have hole
246 at end of array and qos_shrink_lsm for allocate correct lsm.
248 Severity : enhancement
250 Description: Update OFED support to 1.4.1
255 Description: Reduce small size read RPC
256 Details : Set read-ahead limit for every file and only do read-ahead when
257 available read-ahead pages are bigger than 1M to avoid small size
260 Severity : enhancement
262 Description: move AT tunable parameters for more consistent usage
263 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
265 Severity : enhancement
267 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
268 Details : allow skip disconnected ost for send statfs request and hide error
274 Description: fix racy locking of mballoc block bitmaps causing BUG
275 Details : The locking of the mballoc buddy bitmap and the in-memory
276 block bitmap was using two different spin locks in some
277 cases. This made it possible to incorrectly access the
278 mballoc bitmap while another process was modifying it,
279 causing a sanity assertion to fail. While no on-disk corruption
280 was reported, there was some risk of this happening.
283 Frequency : rare, on llog test 6
285 Descriptoin: don't allow connect to already connected import
286 Details : allowing connect to already connected import is hide connecting problem.
291 Descriptoin: allow kill process which wait statahead result
292 Details : for some reasons 'ls' can stick in waiting result from statahead,
293 in this case need way for kill this process.
295 Severity : enhancement
297 Description: Add state history info file, enhance import info file
298 Details : Track import connection state changes in a new osc/mdc proc file;
299 add overview-type data to the osc/mdc import proc file.
301 Severity : enhancement
303 Description: MDS create should not wait for statfs RPC while holding DLM lock.
305 Severity : enhancement
307 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
312 Descriptoin: OST grant too much space to client even there are not enough space.
313 Details : Client will shrink its grant cache to OST if there are no write
314 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
315 this grant cache if there are already not enough avaible space
316 (left_space < total_clients * 32M).
319 Frequency : start MDS on uncleanly shutdowned MDS device
321 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
322 Details : stay in waiting mds<>ost recovery finished produce random bugs
323 due race between two ll_sync thread for one lov target. send
324 ACTIVATE event only if connect realy finished and import have
328 Frequency : rare, connect and disconnect target at same time
330 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
331 Details : don't call obd_disconnect under lov_lock. this long time
332 operation and can block ptlrpcd which answer to connect request.
337 Descriptoin: don't lose wakeup for imp_recovery_waitq
338 Details : recover_import_no_retry or invalidate_import and import_close can
339 both sleep on imp_recovery_waitq, but we was send only one wakeup
343 Frequency : always with long access acl
345 Descriptoin: mds can't pack reply with long acl.
346 Details : mds don't control size of acl but they limited by reint/getattr
349 Severity : enhancement
351 Description: Update to SLES10 kernel-2.6.16.60-0.33.
353 Severity : enhancement
355 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
358 Frequency : start MDS on uncleanly shutdowned MDS device
360 Descriptoin: aborting recovery hang on MDS
361 Details : don't throttle destroy RPCs for the MDT.
364 Frequency : on remount
366 Description: external journal device not working after the remount
367 Details : clear dev_rdonly flag for external journal devices in
373 Description: shutdown vs evict race
374 Details : client_disconnect_export vs connect request race.
375 if client will evicted at this time - we start invalidate
376 thread without referece to import and import can be freed
380 Frequency : rare, need acl's on inode.
382 Description: client can't handle ost additional correctly
383 Details : if ost was added after client connected to mds client can have
384 hit lnet_try_match_md ... to big messages to wide striped files.
385 in this case need teach client to handle config events about add
386 lov target and update client max ea size at that event.
388 Severity : enhancement
390 Description: Changelogs
391 Details : Changelogs are a lightweight mechanism to track filesystem
392 metadata and namespace changes. The changelog is recorded
393 permanently on the MDTs, and is periodically "consumed" / purged
394 when records are no longer needed.
396 Severity : enhancement
398 Description: compact fld format with extents
399 Details : Store range of seq rather than every seq in FLD. Seq
400 controller update FLD rather than clients. In Case of CMD, mdt0
401 has FLD, all other metadata server act as non persistent proxy
402 for FLD queries and cache fld entries in fld cache.
407 Description: don't skip ost target if they assigned to file
408 Details : Drop slow OSCs if we can, but not for requested start idx.
409 This means "if OSC is slow and it is not the requested
410 start OST, then it can be skipped, otherwise skip it only
411 if it is inactive/recovering/out-of-space.
415 Description: more cleanup in mds_lov
416 Details : not send LOV EA under replay, we can't know about they size at this
417 time. Don't allow client connect to mds before any ost connected,
418 for avoid problems with LOV EA size and returning EIO to client.
420 Severity : enhancement
422 Description: Interoperability at server side (Disk interoperability)
424 Severity : enhancement
426 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
428 Severity : enhancement
430 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
432 Severity : enhancement
434 Description: New client IO stack (CLIO).
436 Severity : enhancement
438 Description: Commit on sharing. Eliminate inter-client dependencies between
439 uncommitted transactions by doing transaction commits.
440 Thereby clients may recovery independently.
443 Frequency : Create a symlink file with a very long name
445 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
446 Details : If there is no extra space in the request for early cancels,
447 ldlm_req_handles_avail() returns 0 instead of a negative value.
449 Severity : enhancement
451 Description: Add /proc entry for import status
452 Details : The mdc, osc, and mgc import directories now have
453 an import directory that contains useful import data for debugging
456 Severity : enhancement
458 Description: Re-disable certain /proc logging
459 Details : Enable and disable client's offset_stats, extents_stats and
460 extents_stats_per_process stats logging on the fly.
463 Frequency : Only on FC kernels 2.6.22+
465 Description: oops in statahead
466 Details : Do not drop reference count for the dentry from VFS when lookup,
467 VFS will do that by itself.
469 Severity : enhancement
471 Description: Generic /proc file permissions
472 Details : Set /Proc file permissions in a more generic way to enable non-
473 root users operate on some /proc files.
477 Description: Hitting mdc_commit_close() ASSERTION
478 Details : Properly handle request reference release in
479 ll_release_openhandle().
483 Description: quota recovery deadlock during mds failover
484 Details : This patch includes att18982, att18236, att18237 in bz14840.
486 1. fix osts hang when mds does failover with quotaon
487 2. prevent watchdog storm when osts threads wait for the
492 Frequency : only patchless client
493 Description: add workaround for race between add/remove dentry from hash
495 Severity : enhancement
497 Description: Allow OST glimpses to return PW locks
501 Description: LBUG when llog conf file is full
502 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
506 Description: Prevent import from entering FULL state when server in recovery
510 Description: service mount cannot take device name with ":"
511 Details : Only when device name contains ":/" will mount treat it as
517 Description: replace ptlrpcd with the statahead thread to interpret the async
518 statahead RPC callback
522 Frequency : on recovery
523 Description: I/O failures after umount during fail back
524 Details : if client reconnected to restarted server we need join to recovery
525 instead of find server handler is changed and process self
526 eviction with cancel all locks.
528 Severity : enhancement
530 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
532 Severity : enhancement
534 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
536 Severity : enhancement
538 Description: Upcall on Lustre log has been dumped
539 Details : Allow for a user mode script to be called once a Lustre log has
540 been dumped. It passes the filename of the dumped log to the
541 script, the location of the script can be specified via
542 /proc/sys/lnet/debug_log_upcall.
547 Description: avoid idr_remove called for id which is not allocated.
548 Details : Move assigment s_dev for clustered nfs to end of initialization,
549 to avoid problem with error handling.
554 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
555 Details : When connection is reused this not moved from CONN_UNUSED_HASH
556 into CONN_USED_HASH and this prodice warning when put connection
557 again in unused hash.
560 Severity : enhancement
562 Description: File striping can now be set to use an arbitrary pool of OSTs.
564 Severity : enhancement
566 Description: Export bytes_read/bytes_write count on OSC/OST.
570 Description: Early reply size mismatch, MGC loses connection
571 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
572 the connect flags are properly negotiated.
577 Description: quotas are not honored with O_DIRECT
578 Details : all writes with the flag O_DIRECT will use grants which leads to
579 this problem. Now using OBD_BRW_SYNC to guard this.
583 Description: add quota statistics
584 Details : 1. sort out quota proc entries and proc code.
585 2. add quota statistics
587 Severity : enhancement
589 Description: enable quota support for HEAD.
593 Description: Properly propagate oinfo flags from lov to osc for statfs
594 Details : restore missing copy oi_flags to lov requests.
596 Severity : enhancement
598 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
599 : l_getgroups(8), lst(8), routerstat(8)
600 Details : included man pages for llobdstat(8), llstat(8),
601 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
603 Severity : enhancement
605 Description: configure's --enable-quota should check the
606 : kernel .config for CONFIG_QUOTA
607 Details : configure is terminated if --enable-quota is passed but
608 : no quota support is in kernel
612 Description: Remove portals compatibility
613 Details : Remove portals compatibility, not interoperable with releases
618 Description: Resolve device initialization race
619 Details : Prevent proc handler from accessing devices added to the
620 obd_devs array but yet be intialized.
622 Severity : enhancement
624 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
626 Severity : enhancement
628 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
633 Description: Using wrong pointer in osc_brw_prep_request
634 Details : Access to array[-1] can produce panic if kernel compiled with
635 CONFIG_PAGE_ALLOC enabled
639 Description: Client runs out of low memory
640 Details : Consider only lowmem when counting initial number of llap pages
644 Description: *optional* service tags registration
645 Details : if the "service tags" package is installed on a Lustre node
646 When the filesystem is mounted, a local-node service tag will
647 be created. See http://inventory.sun.com/ for more information
648 about the Service Tags asset management system.
652 Description: Kernel BUG tries to release flock
653 Details : Lustre does not destroy flock lock before last reference goes
654 away. So always drop flock locks when client is evicted and
655 perform unlock regardless of successfulness of speaking to MDS.
659 Description: add refcount protection for osc callbacks, avoid panic on shutdown
663 Description: sanity test 65a fails if stripecount of -1 is set
664 Details : handle -1 striping on filesystem in ll_dirstripe_verify
669 Description: ASSERTION(CheckWriteback(page,cmd)) failed
670 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
673 Severity : enhancement
675 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
679 Description: do not process already freed flock
680 Details : flock can possibly be freed by another thread before it reaches
681 to ldlm_flock_completion_ast.
685 Description: LBUG during stress test
686 Details : Need properly lock accesses the flock deadlock detection list.
690 Description: oops in page fault handler
691 Details : kernel page fault handler can return two special 'pages' in error
692 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
696 Description: timeout with invalidate import.
697 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
698 should be handled by ptlrpcd. This produce long age waiting and
699 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
701 Severity : enhancement
703 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
707 Description: enable MGS and MDT services start separately
708 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
709 a co-located MGS without starting the MGS, which is a complement
710 to 'nosvc' mount option.
714 Frequency : after recovery
715 Description: precreate to many object's after del orphan.
716 Details : del orphan st in oscc last_id == next_id and this triger growing
717 count of precreated objects. Set flag LOW to skip increase count
718 of precreated objects.
722 Frequency : rare, on clear nid stats
723 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
724 Details : when clean nid stats sometimes try destroy live entry,
725 and this produce panic in free.
729 Description: Stack overflow during MDS log replay
730 ease stack pressure by using a thread dealing llog_process.
734 Description: wait until IO finished before start new when do lock cancel.
735 Details : VM protocol want old IO finished before start new, in this case
736 need wait until PG_writeback is cleared until check dirty flag and
737 call writepages in lock cancel callback.
739 Severity : enhancement
741 Description: using special macro for print time and cleanup in includes.
745 Description: mds_mfd_close() ASSERTION(rc == 0)
746 Details : In mds_mfd_close(), we need protect inode's writecount change
747 within its orphan write semaphore to prevent possible races.
751 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
755 Frequency : rare, on shutdown ost
756 Description: don't hit live lock with umount ost.
757 Details : shrink_dcache_parent can be in long loop with destroy dentries,
758 use shrink_dcache_sb instead.
762 Description: don't panic with use echo client
763 Details : echo client pass NULL as client nid pointer and this produce null
768 Description: fix build on ppc32
769 Details : compile code with -m64 flag produce wrong object file for ppc32.
773 Description: add message levels for liblustreapi
777 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
778 Details : Move the 'good_osts' check before the 'total_bavail' check. This
779 will result in an -EAGAIN and in the exit call path we call
780 alloc_rr() which will with increasing aggressiveness attempt to
781 aquire precreated objects on the minimum number of required OSCs.
785 Description: Use old size assignment to avoid deadlock
786 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
787 the scheduling while holding a spinlock. We do not need locking
788 for size in ll_update_inode() because size is only updated from
789 the MDS for directories or files without objects, so there is no
790 other place to do the update, and concurrent access to such inodes
791 are protected by the inode lock.
795 Description: resolve "_IOWR redefined" build error on SLES10
799 Description: dump the memory debugging after all modules are unloaded to
800 suppress false negative in conf_sanity test 39
802 Severity : enhancement
804 Description: build kernel-ib packages for OFED 1.3 in our release cycle
809 Description: fix SLES kernel versioning
810 Details : the kernel version for our SLES 10 kernel did not include a "-"
811 before the "smp" at the end. while this was not a problem in
812 general, it did mean that software trying to use the kernel
813 version to try to detect a vendor specific kernel would fail.
814 this was most evident by the OFED build scripts.
818 Description: Don't update lov_desc members until making sure they are valid
819 Details : When updating lov_desc members via proc fs, need fix their
820 validities before doing the real update.
824 Description: don't put request into delay list while invalidate in flight.
825 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
826 import in flight. this produce timeout for invalidate and sometimes
827 can cause stale data.
831 Frequency : on ppc only
832 Description: not convert ost objects for directory because it's not exist.
833 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
837 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
839 Severity : enhancement
841 Description: Update to support 2.6.22.14 vanilla kernel.
845 Frequency : rare, on recovery
846 Description: read procfs can produce deadlock in some situation
847 Details : Holding lprocfs lock which send rpc can produce block for destroy
848 obd objects and this also block reconnect with -EALREADY. This
849 isn't fix all lprocfs bugs - but make it rare.
851 Severity : enhancement
853 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
856 Frequency : frequent on X2 node
858 Description: mdc_set_open_replay_data LBUG
859 Details : Set replay data for requests that are eligible for replay.
863 Description: lustre_mgs: operation 101 on unconnected MGS
864 Details : When MGC is disconnected from MGS long enough, MGS will evict the
865 MGC, and late on MGC cannot successfully connect to MGS and a lot
866 of the error messages complaining that MGS is not connected.
870 Frequency : on network error
871 Description: panic with double free request if network error
872 Details : mdc_finish_enqueue is finish request if any network error ocuring,
873 but it's true only for synchronus enqueue, for async enqueue
874 (via ptlrpcd) this incorrect and ptlrpcd want finish request
877 Severity : enhancement
879 Description: client-side metadata stat-ahead during readdir(directory readahead)
880 Details : perform client-side metadata stat-ahead when the client detects
881 readdir and sequential stat of dir entries therein
884 Frequency : on start mds
886 Description: Implement get_info(last_id) in obdfilter.
889 Frequency : occasional
891 Description: Correctly check stale fid, not start epoch if ost not support SOM
892 Details : open with flag O_CREATE need set old fid in op_fid3 because
893 op_fid2 was overwritten with new generated fid, but mds can answer
894 with one of these two fids and both is not stale. Setattr
895 incorrectly started an epoch and assume will be called
896 done_writing, but without SOM done_writing ever being called.
899 Frequency : rare, depends on device drivers and load
901 Description: MDS or OSS nodes crash due to stack overflow
902 Details : Code changes in 1.8.0 increased the stack usage of some functions.
903 In some cases, in conjunction with device drivers that use a lot
904 of stack the MDS (or possibly OSS) service threads could overflow
905 the stack. One change which was identified to consume additional
906 stack has been reworked to avoid the extra stack usage.
909 Frequency : occasional
911 Description: Do not fail import if osc_interpret_create gets -EAGAIN
912 Details : If osc_interpret_create got -EAGAIN it immediately exits and
913 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
914 oscc_has_objects and see OSC has no objests and call
915 oscc_internal_create to resend create request.
917 Severity : enhancement
919 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
921 Severity : enhancement
923 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
926 Frequency : very rare
928 Description: Fix replay if there is an un-replied request and open
929 Details : In some cases, older replay request will revert the
930 mcd->mcd_last_xid on MDS which is used to record the client's
933 Severity : enhancement
935 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
937 Severity : enhancement
939 Description: Add rhel5 support to HEAD.
941 Serverity : enhancement
943 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
948 Description: Don't allow skipping OSTs if index has been specified.
949 Details : Don't allow skipping OSTs if index has been specified, make
950 locking in internal create lots better.
954 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
955 Details : access to bitfield in structure is always rounded to long
956 and this produce problem with not atomic change any bit.
960 Description: Lustre make rpms failed.
961 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
962 builds Lustre rpms from tarball.
965 Frequency : rare on shutdown ost
967 Description: If llog cancel was not send before clean_exports phase, this can
968 produce deadlock in llog code.
969 Details : If llog thread has last reference to obd and call class_import_put
970 this produce deadlock because llog_cleanup_commit_master wait when
971 last llog_commit_thread exited, but this never success because was
972 called from llog_commit_thread.
976 Description: allow userland application know is lost one of stripes.
977 Details : fill lvb_blocks with error code on ost and return it to
978 application if error flag found.
982 Description: NULL lov_tgts causing MDS oops
983 Details : more safe checks for NULL lov_tgts for avoid oops.
985 Severity : enhancement
987 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
991 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
992 Details : If an OST is down the MDS will hang indefinitely in
993 obd_statfs_rqset() waiting for the statfs data. While for
994 MDS QOS usage of statfs, it should not stuck in waiting.
996 Severity : enhancement
998 Description: remote_acl support
999 Details : Support ACL-based permission check for remote user.
1000 Support setfacl/getfacl for remote user with the utils
1001 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1002 the system "{s,g}etfacl" utils.
1004 Severity : enhancement
1006 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1008 Severity : enhancement
1010 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1014 Description: make mgs_setparam() handle fsname containing dash
1015 Details : fsname containing a dash does not work with lctl conf_param
1017 Severity : enhancement
1019 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1021 Severity : enhancement
1023 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1027 Frequency : rare, at shutdown
1028 Description: access already free / zero obd_namespace.
1029 Details : if client_disconnect_export was called without force flag set,
1030 and exist connect request in flight, this can produce access to
1031 NULL pointer (or already free pointer) when connect_interpret
1032 store ocd flags in obd_namespace.
1036 Frequency : only at startup
1037 Description: not alloc memory with spinlock held.
1038 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1039 if any spinlock held.
1041 Severity : enhancement
1043 Description: make lustre randomly fail allocating memory
1044 Details : Make lustre randomly failed allocating memory for testing purpose.
1046 Severity : enhancement
1048 Description: lost problems with lov objid file
1049 Details : Fixes some scability and access to not inited memory problems
1050 in work with lov objdid file.
1055 Description: lfs find does not continue on file error
1056 Details : Continue other files processing when a file/dir is absent.
1060 Description: Inconsistent usage of lustre_pack_reply()
1061 Details : Standardize the usage of lustre_pack_reply() such that it
1062 always generate a CERROR on failure.
1065 Frequency : occasional
1067 Description: MDS hang or stay in waiting lock
1068 Details : If client receive lock with CBPENDING flag ldlm need send lock
1069 cancel as separate rpc, to avoid situation when cancel request
1070 can't processed due all i/o threads stay in wait lock.
1074 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1078 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1082 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1086 Description: Update to RHEL4 latest kernel.
1088 Severity : enhancement
1090 Description: Build SLES10 patchless client fails
1091 Details : The configure was broken by run ./configure with
1092 --with-linux-obj=.... argument for patchless client. When the
1093 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1094 can't search header adequately. Use absolute path such as
1095 -I($LINUX)/include instead.
1099 Description: interrupt oig_wait produce painc on resend.
1100 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1101 private set, and this produce situation when rq_ptlrpcd_data not
1102 copyed to new allocated request and triggered LBUG on assert
1103 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1104 ptlrpcd set if request is changed and can be safety changed to use
1109 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1110 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1111 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1112 these macros and replace them with their difinitions instead.
1116 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1117 Details : Update which_patch & target file for SLES10 latest kernel.
1120 Description: Remove client patches from SLES 10 kernel.
1121 Details : This causes SLES 10 clients to behave as patchless clients
1122 even on a Lustre-patched (server) kernel.
1124 Severity : enhancement
1126 Description: self-adjustable client's lru lists
1127 Details : use adaptive algorithm for managing client cached locks lru
1128 lists according to current server load, other client's work
1129 pattern, memory activities, etc. Both, server and client
1130 side namespaces provide number of proc tunables for controlling
1133 Severity : enhancement
1135 Description: light-weight GSS support
1136 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1137 and reduce performance overhead.
1139 Severity : enhancement
1141 Description: Linux keyring support
1142 Details : Support using service of Linux keyring for Lustre GSS internal
1143 context refresh/cache mechanism.
1147 Description: Fix errors in lfs documentation
1148 Details : Fixes man pages
1152 Description: don't use GFP_* in generic Lustre code.
1153 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1157 Description: obdclass is limited by single OBD_ALLOC(idarray)
1158 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1162 Description: warnings with build patchless client with vanila 2.6.19 and up
1163 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1164 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1168 Frequency : Only for SLES
1169 Description: sanity_quota fail test_1
1170 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1171 file, which makes TSTID[2] inunique.
1175 Frequency : Only for SLES9
1176 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1177 Details : Update target/ChangeLog/which_patch .
1181 Frequency : Only for test_10 in sanity-quota.sh
1182 Description: a bug in quota test code
1183 Details : checking a test flag in a wrong place
1187 Frequency : Only for test_13 in sanity-quota.sh
1188 Description: a bug in quota test script
1189 Details : shouldn't check overfull allocated quota
1193 Description: jbd statistics
1194 Details : Port older jbd statistics patch for sles10
1196 Severity : enhancement
1198 Description: Add printing inode info into message about error in writepage.
1203 Description: reply_lock_interpret crash due to race with it and lock cancel.
1204 Details : Do not replay locks that are being cancelled. Do not reference
1205 locks by their address during replay, just by their handle.
1209 Frequency : When flocks are used.
1210 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1211 Details : Flock locks might destroy just granted lock if it could be merged
1212 with another existing flock, this is done in completion handler,
1213 so teach ldlm_cli_enquque_fini that this is a valid case for
1219 Description: Oops in read and write path when failing to allocate lock.
1220 Details : Check if lock allocation failed and return error back.
1224 Description: lstripe command fails for valid OST index
1225 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1226 instead of lov->desc.ld_active_tgt_count.
1230 Description: sanity.sh failed test 103
1231 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1236 Description: df doesn't work properly if diskfs blocksize != 4K
1237 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1241 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1242 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1244 Severity : enhancement
1246 Description: Adaptive timeouts
1247 Details : RPC timeouts adapt to changing server load and network
1248 conditions to reduce resend attempts and improve recovery time.
1252 Description: llapi_file_create() does not allow some changes
1253 Details : add llapi_file_open() that allows specifying the mode and
1254 open flags, and also returns an open file handle.
1258 Description: merge and cleanup kernel patches.
1259 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1263 Description: Add journal checksum support.(Kernel part)
1264 Details : The journal checksum feature adds two new flags i.e
1265 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1266 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1267 indicates that the commit block contains the checksum for
1268 the blocks described by the descriptor blocks. Now commit
1269 record can be sent to disk without waiting for descriptor
1270 blocks to be written to disk. This behavior is controlled
1271 using JBD2_FEATURE_ASYNC_COMMIT flag.
1275 Description: OSS needs mutliple precreate threads
1276 Details : Add ability to start more than one create thread per OSS.
1280 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1281 Details : execute lfs setstripe on client
1285 Description: mds_obd_create error creating tmp object
1286 Details : When the user sets quota on root, llog will be affected and can't
1287 create files and write files.
1290 Frequency : Always on ia64 patchless client, and possibly others.
1292 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1293 Details : This allows the patchless client to be loaded on architectures
1294 without this export.
1298 Description: RedHat Update kernel for RHEL5
1299 Details : Add the kernel config file for RHEL5.
1303 Description: RedHat Update kernel for RHEL5
1304 Details : Modify the kernel config file more closer RHEL5.
1308 Description: Build failure against Centos5 (RHEL5)
1309 Details : Define PAGE_SIZE when it isn't present.
1313 Description: test_7 on sanity-quota.sh can't be used on separate machines
1314 Details : add facet to handle it
1318 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1319 Details : with NFS, the anon dentry's parent was set to itself in
1320 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1321 corresponding dentry other than use rec->ur_name.
1323 Severity : enhancement
1325 Description: lfs setstripe enhancement
1326 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1330 Description: enable data checksumming by default
1331 Details : enable checksum by default, allow --disable-checksum
1332 configure option and "-o nochecksum" mount option. Checksums
1333 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1337 Description: lustre support for RHEL5
1338 Details : Add support for RHEL5.
1342 Description: Client eviction due to failover config
1343 Details : after a connection loss, the lustre client should attempt to
1344 reconnect to the last active server first before trying the
1345 other potential connections.
1349 Description: when mds and osts use different quota unit(32bit and 64bit),
1350 quota will be released repeatly.
1351 Details : void sending multiple quota reqs to mds, which will keep the
1352 status between the reqs.
1356 Description: osts not allocated evenly to files
1357 Details : change the condition to increase offset_idx
1360 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1362 Description: Data corruption for OSTs that are formatted larger than 2TB
1364 Details : When generating the bio request for lustre file writes the
1365 sector number would overflow a temporary variable before being
1366 used for the IO. The data reads correctly from Lustre (which
1367 will overflow in a similar manner) but other file data or
1368 filesystem metadata may be corrupted in some cases.
1372 Description: Tune the kernel for good SCSI performance.
1373 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1374 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1379 Description: rewrite ext2-derived code in obdclass/uuid.c
1380 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1381 from scratch preserving functionality.
1385 Description: "lfs find -obd UUID" prints directories
1386 Details : "lfs find -obd UUID" will return all directory names instead
1387 of just file names. It is incorrect because the directories
1388 do not reside on the OSTs.
1392 Description: use i_size_read and i_size_write in 2.6 port
1393 Details : replace inode->i_size access with i_size_read/write()
1396 Frequency : when removing large files
1398 Description: scheduling issue during removal of large Lustre files
1399 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1400 It causes scheduling issues when removing large files (17TB in the
1404 Frequency : only with liblustre clients on XT3
1406 Description: evictions taking too long
1407 Details : allow llrd to evict clients directly on OSTs
1410 Frequency : only on ppc
1412 Description: /proc/fs/lustre/devices broken on ppc
1413 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1414 We should make sure the type of 'index' is loff_t and then cast
1415 explicitly as needed below. Do not assign an explicitly cast
1419 Frequency : only for rhel5
1421 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1422 Details : Modify the target file & which_kernel.
1424 Severity : enhancement
1426 Description: omit set fsid for export NFS
1427 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1428 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1432 Frequency : Always, for kernels after 2.6.16
1433 Description: Fix warning idr_remove called for id=.. which is not allocated.
1434 Details : Last kernels save old s_dev before kill super and not allow
1435 to restore from callback - restore it before call kill_anon_super.
1439 Description: buffer overruns could theoretically occur
1440 Details : llapi_semantic_traverse() modifies the "path" argument by
1441 appending values to the end of the origin string, and a
1442 overrun may occur. Adding buffer overrun check in liblustreapi.
1446 Description: Fix error on 'ls .' at the top of the Lustre mount.
1447 Details : Don't revalidate dentry if it is a root dentry.
1451 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1452 Details : Modify vm-tunables-rhel4.patch.
1456 Description: Kernel config for 2.6.18-vanilla.
1457 Details : Modify targets/2.6-vanilla.target.in.
1458 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1459 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1460 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1461 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1465 Description: improve handling recoverable errors
1466 Details : if request processig with error which can be recoverable on server
1467 request should be resend, otherwise page released from cache and
1472 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1473 Details : Modify target file & which_patch.
1474 A flaw was found in the IA32 system call emulation provided
1475 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1476 value could be stored in the %RAX register, which could trigger an
1477 out-of-bounds system call table access. An untrusted local user
1478 could exploit this flaw to run code in the kernel
1479 (ie a root privilege escalation). (CVE-2007-4573).
1483 Description: O_DIRECT bypasses client statistics.
1484 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1485 read_ahead_stats were updated. copied the stats section from
1486 osc_send_oap_rpc() into async_internal().
1490 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1494 Description: change order of libsysio includes
1495 Details : '#include sysio.h' should always come before '#include xtio.h'
1499 Description: handle "serious error: objid * already exists" more gracefully
1500 Details : If LAST_ID value on disk is smaller than the objects existing in
1501 the O/0/d* directories, it indicates disk corruption and causes an
1502 LBUG(). If the object is 0-length, then we should use the existing
1503 object. This will help to avoid a full fsck in most cases.
1505 Severity : enhancement
1507 Description: adapt the lustre_config script to support the upgrade case
1508 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1509 server targets to 1.6 in parallel.
1513 Description: To avoid grant space > avaible space when the disk is almost
1514 full. Without this patch you might see the error "grant XXXX >
1515 available" or some LBUG about grant, when the disk is almost
1517 Details : In filter_check_grant, for non_grant cache write, we should
1518 check the left space by if (*left > ungranted + bytes), instead
1519 of (*left > ungranted), because only we are sure the left space
1520 is enough for another "bytes", then the ungrant space should be
1521 increase. In client, we should update cl_avail_grant only there
1522 is OBD_MD_FLGRANT in the reply.
1526 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1527 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1528 described in CVE-2007-4573. This problem could allow untrusted
1529 local users to gain root access.
1532 Frequency : when using O_DIRECT and quotas
1534 Description: Incorrect file ownership on O_DIRECT output files
1535 Details : block usage reported by 'lfs quota' does not take into account
1536 files that have been written with O_DIRECT.
1541 Description: touch file failed when fs is not full
1542 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1543 otherwise we can get ENOSP while fs is not full.
1547 Description: parallel lock callbacks
1548 Details : Instead of sending blocking and completion callbacks as separated
1549 requests, adding them to a set and sending in parallel.
1552 Frequency : only for Cray XT3
1553 Bugzilla : 12829/13455
1554 Description: Changing primary group doesn't change the group lustre assigns to
1556 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1557 overridden with the primary group provided by the group upcall,
1558 whereas the supplied fsgid can be trusted if it is in the list of
1559 supplementary groups returned by the group upcall.
1561 Severity : enhancement
1563 Description: Allow masking D_WARNING, D_ERROR messages from console
1564 Details : Console messages can now be disabled via lnet.printk.
1568 Description: User code with malformed file open parameter crashes client node
1569 Details : Before packing join_file req, all the related reference should be
1570 checked carefully in case some malformed flags cause fake
1571 join_file req on client.
1575 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1576 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1577 mode is equal to the granted mode. However, it should also check
1578 that LDLM_FL_CP_REQD is not set.
1582 Description: Heavy nfs access might result in deadlocks
1583 Details : After ELC code landed, it is now improper to enqueue any mds
1584 locks under och_sem, because enqueue might want to decide to
1585 cancel open locks for same inode we are holding och_sem for.
1589 Description: Client eviction while running blogbench
1590 Details : A lot of unlink operations with concurrent I/O can lead to a
1591 deadlock causing evictions. To address the problem, the number of
1592 oustanding OST_DESTROY requests is now throttled to
1593 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1594 callbacks are processed in priority.
1598 Description: enable ACLs on MDS by default
1599 Details : ACLs must be enabled on MDS by default.
1602 Frequency : PPC/PPC64 only
1604 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1605 Details : fix duplicated definitions between asm-ppc64/types.h and
1606 lustre_types.h on PPC.
1609 Frequency : PPC/PPC64 only
1611 Description: asm-ppc/segment.h does not exist
1612 Details : fix compile issue on PPC.
1616 Description: better handle error messages in extents code
1619 Frequency : RHEL4 only
1621 Description: mkfs is very slow on IA64/RHEL4
1622 Details : A performance regression has been discovered in the MPT Fusion
1623 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1624 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1625 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1627 Severity : enhancement
1629 Description: SNMP support enhancement
1630 Details : Adding total number of sampled request for an MDS node in snmp
1633 Severity : enhancement
1635 Description: Optimize ldlm waiting list processing for PR extent locks
1636 Details : When processing waiting list for read extent lock and meeting read
1637 lock that is same or wider to it that is not contended, skip
1638 processing rest of the list and immediatelly return current
1639 status of conflictness, since we are guaranteed there are no
1640 conflicting locks in the rest of the list.
1644 Description: Time out and refuse to reconnect
1645 Details : When the failover node is the primary node, it is possible
1646 to have two identical connections in imp_conn_list. We must
1647 compare not conn's pointers but NIDs, otherwise we can defeat
1648 connection throttling.
1652 Description: port llog fixes from b1_6 into HEAD
1653 Details : Port llog reference couting and some llog cleanups from b1_6
1654 (bug 10800) into HEAD, for protect from panic and access to already
1655 free llog structures.
1659 Description: Detect stride IO mode in read-ahead
1660 Details : When a client does stride read, read-ahead should detect that and
1661 read-ahead pages according to the detected stride pattern.
1665 Description: data checksumming impacts single node performance
1666 Details : add support for several checksum algorithm. Currently, only CRC32
1667 and Adler-32 are supported. The checksum type can be changed on
1668 the fly via /proc/fs/lustre/osc/*/checksum_type.
1672 Description: use adler32 for page checksums
1673 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1678 Description: build for x2 fails
1679 Details : fix compile issue on Cray systems.
1683 Description: Properly match for duplicate locks
1684 Details : Due to different lock order from skiplists code, we need to
1685 traverse entire list for now
1688 Frequency : only on PPC/SLES10
1690 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1691 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1692 be defined. Add a hack in mkfs_lustre.c to work around this compile
1697 Description: LASSERT on MDS when client holding flock lock dies
1698 Details : ldlm pool logic depends on number of granted locks equal to
1699 number of released locks which is not true for flock locks, so
1700 just exclude such locks from consideration.
1704 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1705 Details : Use fsfilt_sync() for both the whole filesystem sync and
1706 individual file sync to eliminate dangerous inode locking
1707 with I_LOCK that can lead to a deadlock.
1711 Description: performance in 1.6.3
1712 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1713 sure that 1MB requests are not fragmented by the block layer.
1715 Severity : enhancement
1717 Description: organize the server-side client stats on per-nid basis
1718 Details : Change the structure of stats under obdfilter and mds to
1726 The "uuid"s file would list the uuids of _active_ exports.
1727 And the clear entry is to clear all stats and stale nids.
1729 Severity : enhancement
1731 Description: eliminate client locks in face of contention
1732 Details : file contention detection and lockless i/o implementation
1733 for contended files.
1737 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1740 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1742 Description: MDS may lose file striping (and hence file data) in some cases
1743 Details : If there are additional extended attributes stored on the MDS,
1744 in particular ACLs, SELinux, or user attributes (if user_xattr
1745 is specified for the client mount options) then there is a risk
1746 of attribute loss. Additionally, the Lustre file striping
1747 needs to be larger than default (e.g. striped over all OSTs),
1748 and an additional attribute must be stored initially in the
1749 inode and then increase in size enough to be moved to the
1750 external attribute block (e.g. ACL growing in size) for file
1755 Description: skiplist implementation simplification
1756 Details : skiplists are used to group compatible locks on granted list
1757 that was implemented as tracking first and last lock of each
1758 lock group the patch changes that to using doubly linked lists
1762 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1763 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1764 because LASSERT on both the data supplied by a client, and the
1765 data on disk is dangerous and incorrect.
1767 Severity : enhancement
1769 Description: Slow truncate/writes to huge files at high offsets.
1770 Details : Directly associate cached pages to lock that protect those pages,
1771 this allows us to quickly find what pages to write and remove
1772 once lock callback is received.
1776 Description: more ldlm soft lockups
1777 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1778 starve other threads from the resource lock for a long time in
1779 case of long waiting queue, so change the debug level from
1780 D_OTHER to the less frequently used D_INFO.
1782 Severity : enhancement
1784 Description: add -gid, -group, -uid, -user options to lfs find
1788 Description: Hung threads in invalidate_inode_pages2_range
1789 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1790 once one is completed. As a result, some RPCs are stuck in the
1791 queue and are never sent.
1795 Description: filter threads hungs on waiting journal commit
1796 Details : Cleanup filter group llog code, then only filter group llog will
1797 be only created in the MDS/OST syncing process.
1801 Description: Procfs and llog threads access destoryed import sometimes.
1802 Details : Sync the import destoryed process with procfs and llog threads by
1803 the import refcount and semaphore.
1805 Severity : enhancement
1807 Description: openlock cache of b1_6 port to HEAD
1812 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1813 Details : If insertion of an extent fails, then discard the inode
1814 preallocation and free data blocks else it can lead to duplicate
1819 Description: don't always update ctime in ext3_xattr_set_handle()
1820 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1821 In some cases the ctime should not be updated, for example for
1822 2.0->1.8 compatibility it is necessary to delete an xattr and it
1823 should not update the ctime.
1827 Bugzilla : 15713/16362
1828 Description: Assertion in iopen_connect_dentry in 1.6.3
1829 Details : looking up an inode via iopen with the wrong generation number can
1830 populate the dcache with a disconneced dentry while the inode
1831 number is in the process of being reallocated. This causes an
1832 assertion failure in iopen since the inode's dentry list contains
1833 both a connected and disconnected dentry.
1837 Description: assertion failure in ldlm_handle2lock()
1838 Details : fix a race between class_handle_unhash() and class_handle2object()
1839 introduced in lustre 1.6.5 by bug 13622.
1844 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1845 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1850 Description: oops in sd_iostats_seq_show()
1851 Details : unloading/reloading the scsi low level driver triggers a kernel
1852 bug when trying to access the sd iostat file.
1857 Description: Kernel panics during QLogic driver reload
1858 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1859 patch, causing memory corruption.
1864 Description: journal_dev option does not work in b1_6
1865 Details : pass mount option during pre-mount.
1867 Severity : enhancement
1869 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1870 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1871 extent information of a file. It can be used to map logical blocks
1872 in a file to physical blocks in the block device.
1876 Description: LDLM soft lockups - improvement
1877 Details : It is be possible to send the lock handle along with each read
1878 or write request because the client is already doing a lock match
1879 itself so there isn't any reason the OST should have to re-do that
1883 Frequency : only on Cray X2
1885 Description: X2 build failures
1886 Details : fix build failures on Cray X2.
1890 Description: xid & resent requests
1891 Details : Initialize RPC XID from clock at startup (randomly if clock is
1894 Severity : enhancement
1896 Description: Add lustre_start utility to start or stop multiple Lustre servers
1901 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1902 Details : In case of memory pressure, list_del() can be called twice on
1903 req->rq_history_list, causing a kernel oops.
1907 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1908 Details : fix stack overflow in the distributed lock manager by defering
1909 export eviction after a failed AST to the elt thread instead of
1910 handling it in the dlm interpret routine.
1914 Description: Convert some comments to new format.
1915 Details : Update documenting comments to match doxygen conventions.
1919 Description: Grammar fixes.
1920 Details : A couple of trivial spelling fixes.
1924 Description: OSD_COUNTERS-mandatory
1925 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1926 to simplify the code.
1930 Description: simplify cmm_device freeing logic.
1931 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1935 Description: Add lockdep support to dt_object_operations locking interface.
1936 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1937 parameter indicating lock ordering. Update mdd code to use new
1942 Description: Introduce failloc constants for lockless IO tests.
1943 Details : Add two new failloc constants to test lockless IO. Only one of
1944 them in implemented---another is checked in yet to be landed
1949 Description: Add lockdep support for inode mutex.
1950 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1951 in the situations where Lustre has to lock more than one inode
1956 Description: Add optional invariants checking support.
1957 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1958 configure switch. This macro is to be used for consistency and
1959 sanity checks that are too expensive to be left in `production'
1964 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1965 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1970 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1971 Details : Use spin_lock_nested() in (the only) situation where more than
1972 one ldlm_lock is locked simultaneously. Also, fix possible
1973 dead-lock in ldlm_lock_change_resource() by enforcing particular
1978 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1979 Details : Instead of specifying each ldlm_lock call-back through separate
1980 parameter, wrap them into struct ldlm_callback_suite.
1984 Description: Kill join_lru obd method and its callers.
1985 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1986 in memory---a requirement implemented through ->o_join_lru() obd
1987 method in HEAD. Remove this method and its users.
1991 Description: Add asynchronous ldlm ENQUEUE completion handler.
1992 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1993 case that stock ldlm_completion_ast() cannot handle as it waits
1994 until lock is granted. Introduce new ldlm_completion_ast_async()
1995 for this. Also comment ldlm_completion_ast().
1999 Description: ldlm_error <-> errno conversion.
2000 Details : Add functions to map (rather arbitrary) between LDLM error codes
2001 and standard errno values. CLIO needs this to prevent LDLM specific
2002 constants from escaping ldlm and osc.
2006 Description: Kill unused ldlm_handle2lock_ns() function.
2007 Details : Kill unused ldlm_handle2lock_ns() function.
2011 Description: Add lu_ref support to ldlm_lock
2012 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2013 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2014 and ldlm_resource. LDLM interface has to be changed, because code that
2015 releases a reference on a lock, has to "know" what reference this is.
2016 In the most frequent case
2018 lock = ldlm_handle2lock(handle);
2020 LDLM_LOCK_PUT(lock);
2022 no changes are required. When any other reference (received _not_ from
2023 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2024 instead of LDLM_LOCK_PUT().
2026 Arguably, changes are pervasive, and interface requires some discipline
2027 for proper use. On the other hand, it was very instrumental in finding
2028 a few leaked lock references.
2032 Description: Add ldlm_lock_addref_try().
2033 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2034 attempts to addref a lock that might be being canceled
2039 Description: Add ldlm_weigh_callback().
2040 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2041 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2042 hard-coded `number of pages' logic.
2046 Description: Add lockdep annotations to llog code.
2047 Details : Use appropriately tagged _nested() locking calls in the places
2048 where llog takes more than one ->lgh_lock lock.
2052 Description: Add loi_kms_set().
2053 Details : Wrap kms updates into a helper function.
2057 Description: Constify instances of struct lsm_operations.
2058 Details : Constify instances of struct lsm_operations.
2062 Description: lu_conf support.
2063 Details : On a server, a file system object is uniquely identified
2064 by a fid, which is sufficient to locate and load all object
2065 state (inode). On a client, on the other hand, more data are
2066 necessary instantiate an object. Change lu_object_find() and
2067 friends to take additional `lu_conf' argument describing object.
2068 Typically this includes layout information.
2072 Description: lu_context fixes.
2073 Details : Introduce new lu_context functions that are needed on the client
2074 side, where some system threads (ptlrpcd) are shared by multiple
2075 modules, and so cannot be stopped during module shutdown.
2079 Description: Add start and stop methods to lu_device_type_operations.
2080 Details : Introduce two new methods in lu_device_type_operations, that are
2081 invoked when first instance of a given type is created and last one
2082 is destroyed respectively. This is need by CLIO.
2086 Description: Add lu_ref support to struct lu_device.
2087 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2088 track leaked references.
2092 Description: Introduce lu_kmem_descr.
2093 Details : lu_kmem_descr and its companion interface allow to create
2094 and destroy a number of kmem caches at once.
2098 Description: Fix lu_object finalization race.
2099 Details : Fix a race between lu_object_find() finding an object and its
2100 concurrent finalization. This race is (most likely) not possible
2101 on the server, but might happen on the client.
2105 Description: Introduce lu_ref interface.
2106 Details : lu_ref is a debugging module allowing to track references to
2107 a given object. It is quite cpu expensive, and has to be
2108 explicitly enabled with --enable-lu_ref. See usage description
2113 Description: Factor lu_site procfs stats into a separate function.
2114 Details : Separate lu_site stats printing code into a separate function
2115 that can be reused on a client.
2119 Description: Constify instances of struct {lu,dt,md}_device_operations.
2120 Details : Constify instances of struct {lu,dt,md}_device_operations.
2124 Description: Introduce struct md_site and move meta-data specific parts of
2125 struct lu_site here.
2126 Details : Move md-specific fields out of struct lu_site into special struct
2127 md_site, so that lu_site can be used on a client.
2131 Description: Kill mdd_lov_destroy().
2132 Details : Remove unused mdd code.
2136 Description: Add st_block checking to multistat.c.
2137 Details : Add st_block checking to multistat.c.
2141 Description: Add lu_ref support to struct obd_device.
2142 Details : Add lu_ref tracking to obd_device.
2146 Description: Kill obd_set_fail_loc().
2147 Details : Remove unused code.
2151 Description: Add special type for ptlrpc_request interpret functions.
2152 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2153 there. Actual usage will be in CLIO.
2157 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2158 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2159 doesn't work with lockdep.
2163 Description: Add rwv.c test program.
2164 Details : New testing program exercising readv(2) and writev(2) (Qian).
2168 Description: Add sendfile.c test program.
2169 Details : New testing program exercising sendfile(2) (Jay).
2173 Description: Ratelimit a message that can be very frequent.
2174 Details : Ratelimit a memory allocation failure message that can
2179 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2180 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2184 Description: Liblustre build fixes.
2185 Details : Liblustre build fixes.
2189 Description: libcfs: add cfs_{need,cond}_resched() interface.
2190 Details : libcfs: add cfs_{need,cond}_resched() definition and
2191 implementations for Linux, NT, and liblustre.
2193 Severity : enhancement
2195 Description: More exported tunables for mballoc
2196 Details : Add support for tunable preallocation window and new tunables for
2197 large/small requests
2201 Description: Detect corruption of block bitmap and checking for preallocations
2202 Details : Checks validity of on-disk block bitmap. Also it does better
2203 checking of number of applied preallocations. When corruption is
2204 found, it turns filesystem readonly to prevent further corruptions.
2208 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2209 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2210 window to 0. If it does hit cache miss, and read-pattern is still
2211 stride-io mode, does not reset the stride window, but also does not
2212 increase the stride window length in this case.
2216 Frequency : only for big-endian servers
2217 Description: Check if system is big-endian while mounting fs with extents feature
2218 Details : Mounting a filesystem with extents feature will fail on big-endian
2219 systems since ext3-based ldiskfs is not supported on big-endian
2220 systems. This can be over-riden with "bigendian_extents" mount option.
2222 Severity : enhancement
2224 Description: The root squash functionality
2225 Details : A security feature, which is to prevent users from being able
2226 to mount lustre on their desktop, run as root, and delete
2227 all of the files in the filesystem. The goal is accomplished by
2228 remapping user id (UID) and group id (GID) of the root user to
2229 a UID and GID specified by the system administartor via Lustre
2230 configuration management server (MGS). The functionality also
2231 allows to specify sets of clients for which the remapping does
2236 Description: Excessive recovery window
2237 Details : With AT enabled, the recovery window can be excessively long (6000+
2238 seconds). To address this problem, we no longer use
2239 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2240 timeout no longer depends on the service time, it is set to
2241 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2242 time via pb_service_time.
2246 Description: Watchdog triggered on MDS failover
2247 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2248 the OSTs know that the MDS "UUID" can be reused for the same export
2249 from a different NID, so we do not need to wait for the export to be
2253 Frequency : rare, only if using MMP with Linux RAID
2255 Description: MMP doesn't work with Linux RAID
2256 Details : While using HA for Lustre servers with Linux RAID, it is possible
2257 that MMP will not detect multiple mounts. To make this work we
2258 need to unplug the device queue in RAID when the MMP block is being
2259 written. Also while reading the MMP block, we should read it from
2260 disk and not the cached one.
2262 Severity : enhancement
2264 Description: open file using fid
2265 Details : A file can be opened using just its fid, like
2266 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2269 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2272 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2273 Details : add proper locking for creation of jbd_4k slab cache
2277 Description: MMP check in ext3_remount() fails without displaying any error
2278 Details : When multiple mount protection fails during remount, proper error
2281 Severity : enhancement
2283 Description: Allow stripe size to be up to 4G-64k
2284 Details : Fix math logic to allow large stripe sizes.
2288 Description: add check for >8TB ldiskfs filesystems
2289 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2290 Don't allow >8TB ldiskfs filesystems to be mounted without
2291 force_over_8tb mount option
2296 Descriptoin: OSS complains "too many blocks" and crashes
2297 Details : In filter_commitrw_write():
2298 1. lquota_chkquota();
2299 2. obdo_from_inode();
2300 3. lquota_pending_commit();
2301 "2." has changed uid and gid in this case, then caused
2302 lquota_chkquota() and lquota_pending_commit() to use
2303 different uid and gid. That is the root of the bug.
2307 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2308 Details : Detect and print out the 'stripe count = -1' case properly.
2312 Description: Out or order replies might be lost on replay
2313 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2314 request with tid smaller then one currently being added, add it
2315 to the start, not end of the list.
2317 --------------------------------------------------------------------------------
2319 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2321 * Support for kernels:
2322 2.6.5-7.283 (SLES 9),
2323 2.6.9-55.EL (RHEL 4),
2324 2.6.16.46-0.14 (SLES 10),
2325 2.6.18.8 vanilla (kernel.org)
2326 * Client support for unpatched kernels:
2327 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2328 2.6.16 - 2.6.22 vanilla (kernel.org)
2329 * Due to recently discovered recovery problems, we do not recommend
2330 using patchless RHEL 4 clients with this or any earlier release.
2331 * Recommended e2fsprogs version: 1.39.cfs8
2332 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2333 * Starting with this release, the ldiskfs backing filesystem required
2334 by Lustre is now in its own package, lustre-ldiskfs. This package
2335 should be installed. It is versioned separately from Lustre and
2336 may be released separately in future.
2340 Description: block reactivating mgc import until all deactivates complete
2341 Details : Fix race when failing back MDT/MGS to itself (testing)
2344 Frequency : at statup only
2346 Description: mds_lov_synchronize race leads to various problems
2347 Details : simultaneous MDT->OST connections at startup can cause the
2348 sync to abort, leaving the OSC in a bad state.
2350 Severity : enhancement
2352 Description: add optional extra BUILD_VERSION info
2353 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2354 to override the lustre version.
2357 Frequency : 2.6.18 servers only
2359 Description: ll_kern_mount() doesn't release the module reference
2360 Details : The ldiskfs module reference count never drops down to 0
2361 because ll_kern_mount() doesn't release the module reference.
2366 Description: server LBUG when using old ost_num_threads parameter
2367 Details : Accept the old ost_num_threads parameter but warn that it
2368 is deprecated, and fix an off-by-one error that caused an LBUG.
2373 Description: Transient SCSI error results in persistent IO issue
2374 Details : iobuf->dr_error is not reinitialized to 0 between two
2378 Frequency : sometimes when underlying device returns I/O errors
2380 Description: OSTs not going read-only during write failures
2381 Details : OSTs are not remounted read-only when the journal commit threads
2382 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2383 instead of the ext3 wrappers.
2387 Description: poor connect scaling with increasing client count
2388 Details : Don't run filter_grant_sanity_check for more than 100 exports
2389 to improve scaling for large numbers of clients.
2392 Frequency : SLES10 only
2394 Description: sanity-quota.sh quotacheck failed: rc = -22
2395 Details : Quotas cannot be enabled on SLES10.
2398 Frequency : liblustre clients only
2400 Description: getdirentries does not give error when run on compute nodes
2401 Details : getdirentries does not fail when the size specified as an argument
2402 is too small to contain at least one entry
2404 Severity : enhancement
2406 Description: Add LNET router traceability for debug purposes
2407 Details : If a checksum failure occurs with a router as part of the
2408 IO path, the NID of the last router that forwarded the bulk data
2409 is printed so it can be identified.
2414 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2415 Details : Due to a race condition, liblustre clients were occasionally
2416 evicted incorrectly.
2418 Severity : enhancement
2420 Description: lfs setstripe use optional parameters instead of postional
2423 Severity : enhancement
2425 Description: Nanosecond timestamp support for ldiskfs
2426 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2427 resolution timestamps. There is not yet support for this at
2428 the Lustre filesystem level.
2431 Frequency : during server recovery
2433 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2434 Details : request with rq_no_resend flag not awake l_wait_event if they get
2438 Frequency : nfs export on patchless client
2440 Description: connectathon hang when test nfs export over patchless client
2441 Details : Disconnected dentry cannot be found with lookup, so we do not need
2442 to unhash it or make it invalid
2445 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2446 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2447 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2452 Description: open req refcounting wrong on reconnect
2453 Details : If reconnect happened between getting open reply from server and
2454 call to mdc_set_replay_data in ll_file_open, we will schedule
2455 replay for unreferenced request that we are about to free.
2456 Subsequent close will crash in variety of ways.
2457 Check that request is still eligible for replay in
2458 mdc_set_replay_data().
2463 Description: disable writes to filesystem when reading health_check file
2464 Details : the default for reading the health_check proc file has changed
2465 to NOT do a journal transaction and write to disk, because this
2466 can cause reads of the /proc file to hang and block HA state
2467 checking on a healthy but otherwise heavily loaded system. It
2468 is possible to return to the previous behaviour during configure
2469 with --enable-health-write.
2471 Severity : enhancement
2473 Description: 64-bit inode version
2474 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2475 to the inode. This will be required for version-based recovery.
2480 Description: MDS fails to start if a duplicate client export is detected
2481 Details : in some rare cases it was possible for a client to connect to
2482 an MDS multiple times. Upon recovery the MDS would detect this
2483 and fail during startup. Handle this more gracefully.
2485 Severity : enhancement
2487 Description: Add -o localflock option to simulate old noflock
2489 Details : This will achieve local-only flock/fcntl locks
2495 Description: log_commit_thread vs filter_destroy race leads to crash
2496 Details : Take import reference before releasing llog record semaphore
2501 Description: Wrong request locking in request set processing
2502 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2503 imp_delayed_list, in this place should be used imp_lock.
2506 Frequency : when reconnection
2508 Description: Grant Leak when osc reconnect to OST
2509 Details : When osc reconnect ost, OST(filter) should check whether it
2510 should grant more space to client by comparing fed_grant and
2511 cl_avail_grant, and return the granted space to client instead
2512 of "new granted" space, because client will call osc_init_grant
2513 to update the client grant space info.
2516 Frequency : when client reconnect to OST
2518 Description: Grant Leak when osc do resend and replay bulk write
2519 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2520 bulk write request, because the grant info while be sync between
2521 OSC and OST when reconnect, and we should ignore the grant info
2522 these of resend/replay write req.
2527 Description: Grant space more than avaiable left space sometimes.
2528 Details : When then OST is about to be full, if two bulk writing from
2529 different clients came to OST. Accord the avaliable space of the
2530 OST, the first req should be permitted, and the second one
2531 should be denied by ENOSPC. But if the seconde arrived before
2532 the first one is commited. The OST might wrongly permit second
2533 writing, which will cause grant space > avaiable space.
2536 Frequency : when client is evicted
2538 Description: Grant might be wrongly erased when osc is evicted by OST
2539 Details : when the import is evicted by server, it will fork another
2540 thread ptlrpc_invalidate_import_thread to invalidate the
2541 import, where the grant will be set to 0. While the original
2542 thread will update the grant it got when connecting. So if
2543 the former happened latter, the grant will be wrongly errased
2544 because of this race.
2549 Description: Checking Stale with correct fid
2550 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2551 to check whether it is stale, because sometimes, we want the
2552 enqueue happened anyway, and op_data.fid2 will not be initialized.
2554 Severity : enhancement
2556 Description: update patchless client
2557 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2560 Frequency : only with 2.4 kernel
2562 Description: random memory corruption
2563 Details : size of struct ll_inode_info is to big for union inode.u and this
2564 can be cause of random memory corruption.
2569 Description: Memory leak in recovery
2570 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2571 It should also check obd_fail before fsfilt_start, otherwise if
2572 fsfilt_start return -EROFS,(failover mds during mds recovery).
2573 then the req will return with repmsg->transno = 0 and rc = EROFS.
2574 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2575 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2576 be freed no matter whether fsfilt_commit success or not.
2579 Frequency : only with huge count clients
2581 Description: Prevents from taking the superblock lock in llap_from_page for
2583 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2584 for llap_from_page call in ll_removepage prevents from taking the
2585 superblock lock for a soon died page.
2590 Description: Not check open intent error before release open handle
2591 Details : in some rare cases, the open intent error is not checked before
2592 release open handle, which may cause
2593 ASSERTION(open_req->rq_transno != 0), because it tries to release
2594 the failed open handle.
2599 Description: Set cat log bitmap only after create log success.
2600 Details : in some rare cases, the cat log bitmap is set too early. and it
2601 should be set only after create log success.
2605 Description: Accessing a block bevice can re-enable I/O when Lustre is
2606 tearing down a device.
2607 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2612 Description: service threads may hog cpus when there are a lot of requests
2613 Details : Insert cond_resched to give other threads a chance to use some CPU
2618 Description: the cat log was not initialized in recovery
2619 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2620 unlink log on mds will not be initialized until mds post
2621 recovery. And also in mds post recovery, the unlink log will
2622 initialization will be done asynchronausly, so there will be race
2623 between add unlink log and unlink log initialization.
2627 Description: brw_stats were being printed incorrectly
2628 Details : brw_stats were being printed as log2 but all of them were not
2629 recorded as log2. Also remove some code duplication arising from
2630 filter_tally_{read,write}.
2634 Frequency : rare, only in recovery.
2635 Description: ASSERTION(req->rq_type != LI_POISON) failed
2636 Details : imp_lock should be held while iterating over imp_sending_list for
2637 prevent destroy request after get timeout in ptlrpc_queue_wait.
2641 Description: replay-single.sh test 52 fails
2642 Details : A lock's skiplist need to be cleanup when it being unlinked
2643 from its resource list.
2647 Description: Short directio read returns full requested size rather than
2649 Details : Direct I/O operations should return actual amount of bytes
2650 transferred rather than requested size.
2652 Severity : enhancement
2654 Description: metadata RPC reduction (e.g. for rm performance)
2655 Details : decrease the amount of synchronous RPC between clients and servers
2656 by canceling conflicing lock before the operation on the client
2657 and packing thier handles into the main operation RPC to server.
2659 Severity : enhancement
2661 Description: Async OSC create to avoid the blocking unnecessarily.
2662 Details : If a OST has no remain object, system will block on the creating
2663 when need to create a new object on this OST. Now, ways use
2664 pre-created objects when available, instead of blocking on an
2665 empty osc while others are not empty. If we must block, we block
2666 for the shortest possible period of time.
2670 Description: improve handling recoverable errors
2671 Details : if request processig with error which can be recoverable on server
2672 request should be resend, otherwise page released from cache and
2675 Severity : enhancement
2677 Description: refine locking for avoid write wrong info into lov_objid file
2678 Details : fix possible races with add new target and write/update data in
2681 --------------------------------------------------------------------------------
2683 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2688 Frequency : on some architectures
2690 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2691 Details : Uninitialized flags sometimes cause configuration commands to
2695 Frequency : patchless clients only
2697 Description: missing __iget() symbol export
2698 Details : The __iget() symbol export is missing. To avoid the need for
2699 this on patchless clients the deathrow inode reaper is turned
2700 off, and we depend on the VM to clean up old inodes. This
2701 dependency was during via the fix for bug 12181.
2703 --------------------------------------------------------------------------------
2705 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2707 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2708 INTEROPERATE with older versions automatically. In many cases a
2709 special upgrade step is needed. Please read the
2710 user documentation before upgrading any part of a live system.
2711 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2712 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2713 * WARNING: Lustre configuration and startup changes are required with
2714 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2718 Severity : enhancement
2720 Description: MountConf
2721 Details : Lustre configuration is now managed via mkfs and mount
2722 commands instead of lmc and lconf. New obd types (MGS, MGC)
2723 are added for dynamic configuration management. See
2724 https://mail.clusterfs.com/wikis/lustre/MountConf for
2727 Severity : enhancement
2729 Description: dynamic OST addition
2730 Details : OSTs can now be added to a live filesystem
2732 Severity : enhancement
2734 Description: startup order invariance
2735 Details : MDTs and OSTs can be started in any order. Clients only
2736 require the MDT to complete startup.
2738 Severity : enhancement
2740 Description: parallel, asynchronous orphan cleanup
2741 Details : orphan cleanup is now performed in separate threads for each
2742 OST, allowing parallel non-blocking operation.
2744 Severity : enhancement
2746 Description: optimized stripe assignment
2747 Details : stripe assignments are now made based on ost space available,
2748 ost previous usage, and OSS previous usage, in order to try
2749 to optimize storage space and networking resources.
2751 Severity : enhancement
2753 Description: Permanently set tunables
2754 Details : All writable /proc/fs/lustre tunables can now be permanently
2755 set on a per-server basis, at mkfs time or on a live system.
2757 Severity : enhancement
2759 Description: Lustre message v2
2760 Details : Add lustre message format v2.
2762 Severity : enhancement
2764 Description: client OST exclusion list
2765 Details : Clients can be started with a list of OSTs that should be
2766 declared "inactive" for known non-responsive OSTs.
2770 Description: SPEC SFS validation failure on NFS v2 over lustre.
2771 Details : Changes the blocksize for regular files to be 2x RPC size,
2772 and not depend on stripe size.
2774 Severity : enhancement
2776 Description: Multiple MD RPCs in flight.
2777 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2778 To avoid overly-overloading MDS, introduce a limit on number of
2779 MDS RPCs in flight for a single client and add /proc controls
2780 to adjust this limit.
2782 Severity : enhancement
2784 Description: client read/write statistics
2785 Details : Add client read/write call usage stats for performance
2786 analysis of user processes.
2787 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2788 file access. extents_stats shows chunk size distribution.
2789 extents_stats_per_process show chunk size distribution per
2792 Severity : enhancement
2794 Description: mds statistics
2795 Details : Add detailed mds operations statistics in
2796 /proc/fs/lustre/mds/*/stats.
2800 Description: Failure of copying files with lustre special EAs.
2801 Details : Client side always return success for setxattr call for lustre
2802 special xattr (currently only "trusted.lov").
2807 Description: Refcount LNET uuids
2808 Details : The global LNET uuid list grew linearly with every startup;
2809 refcount repeated list entries instead of always adding to
2812 Severity : enhancement
2814 Description: Dynamic service threads
2815 Details : Within a small range, start extra service threads
2816 automatically when the request queue builds up.
2819 Frequency : mixed-endian client/server environments
2821 Description: mixed-endian crashes
2822 Details : The new msg_v2 system had some failures in mixed-endian
2825 Severity : enhancement
2827 Description: Easy OST removal
2828 Details : OSTs can be permanently deactivated with e.g. 'lctl
2829 conf_param lustre-OST0001.osc.active=0'
2831 Severity : enhancement
2833 Description: MGS proc entries
2834 Details : Added basic proc entries for the MGS showing what filesystems
2837 Severity : enhancement
2839 Description: provide MGS failover
2840 Details : Added config lock reacquisition after MGS server failover.
2842 Severity : enhancement
2844 Description: add Linux 2.4 support
2845 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2849 Description: a large application tries to do I/O to the same resource and dies
2850 in the middle of it.
2851 Details : Check the req->rq_arrival time after the call to
2852 ost_brw_lock_get(), but before we do anything about
2853 processing it & sending the BULK transfer request. This
2854 should help move old stale pending locks off the queue as
2855 quickly as obd_timeout.
2858 Frequency : when an incorrect nid is specified during startup
2860 Description: ptlrpc connect to non-existant node causes kernel crash
2861 Details : LNET can't be re-entered from an event callback, which
2862 happened when we expire a message after the export has been
2863 cleaned up. Instead, hand the zombie cleanup off to another
2866 Severity : enhancement
2868 Description: plain/inodebits lock performance improvement
2869 Details : Grouping plain/inodebits in granted list by their request modes
2870 and bits policy, thus improving the performance of search through
2874 Frequency : only if OST filesystem is corrupted
2876 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2877 Details : for a short time RPCs with bulk IO are in the replay list,
2878 but replay of bulk IOs is unimplemented. If the OST filesystem
2879 is corrupted due to disk cache incoherency and then replay is
2880 started it is possible to trip an assertion. Avoid putting
2881 committed RPCs into the replay list at all to avoid this issue.
2884 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2886 Description: System hang on startup
2887 Details : This bug allowed the liblustre (e.g. catamount) client to
2888 return to the app before handling all startup RPCs. This
2889 could leave the node unresponsive to lustre network traffic
2890 and manifested as a server ptllnd timeout.
2892 Severity : enhancement
2894 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2895 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2896 boolean to control whether to print peer debug info when a
2897 client's RPC times out.
2900 Frequency : only for kernels with patches from Lustre below 1.4.3
2902 Description: Remove old rdonly API
2903 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2906 Frequency : only for devices with external journals
2908 Description: Set external device read-only also
2909 Details : During a commanded failover stop, we set the disk device
2910 read-only while the server shuts down. We now also set any
2911 external journal device read-only at the same time.
2914 Frequency : when upgrading from 1.4 while trying to change parameters
2916 Description: The wrong (new) MDC name was used when setting parameters for
2917 upgraded MDT's. Also allows changing of OSC (and MDC)
2918 parameters if --writeconf is specified at tunefs upgrade time.
2921 Frequency : when setting specific ost indicies
2923 Description: QOS code breaks on skipped indicies
2924 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2925 created with --index need not be sequential.
2927 Severity : enhancement
2929 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2930 Details : The uninit_groups feature works in conjunction with the kernel
2931 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2932 up the pass1 processing of e2fsck. This is a read-only feature
2933 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2934 mount filesystems that have had this feature enabled.
2936 Severity : enhancement
2938 Description: Improve multi-block allocation algorithm to avoid fragmentation
2939 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2940 allocation locality and avoid filesystem fragmentation.
2942 ------------------------------------------------------------------------------
2944 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2946 * Support for kernels:
2947 2.4.21-47.0.1.EL (RHEL 3)
2948 2.6.5-7.283 (SLES 9)
2949 2.6.9-42.0.10.EL (RHEL 4)
2950 2.6.12.6 vanilla (kernel.org)
2951 2.6.16.27-0.9 (SLES 10)
2952 * Recommended e2fsprogs version: 1.39.cfs5
2954 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2958 Frequency : occasional, depends on client load and configuration
2959 Bugzilla : 12181, 12203
2960 Description: data loss for recently-modified files
2962 Details : In some cases it is possible that recently written or created
2963 files may not be written to disk in a timely manner (this should
2964 normally be within 30s unless client IO load is very high).
2965 The problem appears as zero-length files or files that are a
2966 multiple of 1MB in size after a client crash or client eviction
2967 that are missing data at the end of the file.
2969 This problem is more likely to be hit on clients where files are
2970 repeatedly created and unlinked in the same directory, clients
2971 have a large amount of RAM, have many CPUs, the filesystem has
2972 many OSTs, the clients are rebooted frequently, and/or the files
2973 are not accessed by other nodes after being written.
2975 The presence of the problem can be detected by looking at
2976 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2977 smaller than the second (nr_unused) then dirty files will not
2978 be flushed automatically to disk. "sync; sleep 10" should be
2979 run several times on the node before unmounting it to update
2980 Lustre (this is also safe to run on nodes without this problem).
2982 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2983 that can cause this same problem, so customers using that kernel
2984 also need to update the kernel in addition to Lustre. In order
2985 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2987 It is normal that files written just before a client crash (less
2988 than 30s) may not yet have been flushed to disk, even for local
2992 Frequency : frequent on thin XT3 nodes
2994 Description: UUID collision on thin XT3 Linux nodes
2995 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2996 randomly, since we relied on an insufficiently-seeded PRNG.
3001 Description: OSS hangs after "All ost request buffers busy"
3002 Details : A deadlock between quota and journal operations caused OSS
3003 hangs after printing "All ost request buffers busy."
3006 Frequency : always on liblustre builds
3008 Description: Cleanup compiler warnings on liblustre
3011 Frequency : always on liblustre builds on XT3
3013 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3017 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3018 > 32000 subdirectories
3019 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3020 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3021 subdirectory count crosses 32000. This will aid e2fsck to
3022 correctly handle more than 32000 subdirectories.
3025 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3027 Description: System hang on startup
3028 Details : This bug allowed the liblustre (e.g. catamount) client to
3029 return to the app before handling all startup RPCs. This
3030 could leave the node unresponsive to lustre network traffic
3031 and manifested as a server ptllnd timeout.
3033 Severity : enhancement
3035 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3036 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3037 boolean to control whether to print peer debug info when a
3038 client's RPC times out.
3043 Description: make O_SYNC working on 2.6 kernels
3044 Details : 2.6 kernels use different method for mark pages for write,
3045 so need add a code to lustre for O_SYNC work.
3050 Description: Failure to close file and release space on NFS
3051 Details : Put inode details into lock acquired in ll_intent_file_open.
3052 Use mdc_intent_lock in ll_intent_open to properly
3053 detect all kind of errors unhandled by mdc_enqueue.
3058 Description: proc file read during shutdown sometimes raced obd removal,
3060 Details : Add lock to prevent obd access after proc file removal.
3063 Frequency : Only for files larger than 4GB on 32-bit clients.
3065 Description: improperly doing page alignment of locks
3066 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3067 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3072 Description: RPCs being resent when they shouldn't be
3073 Details : Some RPCs that should not be resent are being resent. This
3074 can cause inconsistencies in the RPC state machine. Do not
3075 resend such requests.
3078 Frequency : rare, only with NFS export
3080 Description: Crash on NFS re-export node
3081 Details : under very unusual load conditions an assertion is hit in
3082 ll_intent_file_open()
3085 Frequency : only if OST filesystem is corrupted
3087 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3088 Details : for a short time RPCs with bulk IO are in the replay list,
3089 but replay of bulk IOs is unimplemented. If the OST filesystem
3090 is corrupted due to disk cache incoherency and then replay is
3091 started it is possible to trip an assertion. Avoid putting
3092 committed RPCs into the replay list at all to avoid this issue.
3097 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3098 Details : Large single O_DIRECT read and write calls can fail to allocate
3099 a sufficiently large buffer to process the request. In case of
3100 allocation failure the allocation is retried with a smaller
3101 buffer and broken into smaller requests.
3103 Severity : enhancement
3105 Description: Add -o localflock option to simulate old noflock behaviour.
3106 Details : This will achieve local-only flock/fcntl locks coherentness.
3111 Description: versioning check is incomplete
3112 Details : Checking the version difference of client vs. server, report
3113 error if the gap is too big.
3118 Description: add support PG_writeback bit
3119 Details : add support for PG_writeback bit for Lustre, for more carefull
3120 work with page cache in 2.6 kernel. This also fix some deadlocks
3121 and remove hack for work O_SYNC with 2.6 kernel.
3123 Severity : enhancement
3125 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3126 Details : The uninit_groups feature works in conjunction with the kernel
3127 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3128 up the pass1 processing of e2fsck. This is a read-only feature
3129 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3130 mount filesystems that have had this feature enabled.
3132 Severity : enhancement
3134 Description: Improve multi-block allocation algorithm to avoid fragmentation
3135 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3136 allocation locality and avoid filesystem fragmentation.
3138 ------------------------------------------------------------------------------
3140 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3142 * Support for kernels:
3143 2.6.9-42.0.3.EL (RHEL 4)
3144 2.6.5-7.276 (SLES 9)
3145 2.4.21-47.0.1.EL (RHEL 3)
3146 2.6.12.6 vanilla (kernel.org)
3147 2.6.16.21-0.8 (SLES10)
3148 * Recommended e2fsprogs version: 1.39.cfs2-0
3150 * The backwards-compatible /proc/sys/portals symlink has been removed
3151 in this release. Before upgrading, please ensure that you change
3152 any configuration scripts or /etc/sysctl.conf files that access
3153 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3154 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3155 in advance of the upgrade on any system running Lustre 1.4.6 or
3156 newer, since /proc/sys/lnet was added in that version.
3157 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3161 Frequency : only when quota is used
3163 Description: avoid scanning export list for quota master
3164 Details : Change the algorithms to avoid scanning export list in order
3165 to improve the efficiency.
3168 Frequency : MDS failover only, very rarely
3170 Description: "went back in time" messages on mds failover
3171 Details : The greatest transno may be lost when the current operation
3172 finishes with an error (transno==0) and the client's last_rcvd
3173 record is over-written. Save the greatest transno in the
3174 mds_last_transno for this case.
3177 Frequency : always for specific kernels and striping counts
3179 Description: client may get "Matching packet too big" without ACL support
3180 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3181 when trying to access files in certain configurations. The
3182 clients should in fact be denied when mounting because they do
3183 not understand ACLs.
3186 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3188 Description: many clients connecting with IO in progress causes connect timeouts
3189 Details : Avoid synchronous journal commits to avoid delays caused by many
3190 clients connecting/disconnecting when bulk IO is in progress.
3191 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3192 OST_IO_PORTAL to avoid delays behind potentially many pending
3196 Frequency : occasionally with multiple writers to a single file
3198 Description: shared writes to file may result in wrong size reported by stat()
3199 Details : Allow growing of kms when extent lock is cancelled
3202 Frequency : always with random mmap IO to multi-striped file
3204 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3205 Details : Only if the hole is at the end of OST object so that kms is too
3206 small. Fix is to increase kms accordingly in ll_nopage.
3209 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3211 Description: writes to a missing object would leak memory on the OST
3212 Details : If there is an inconsistency between the MDS and OST filesystems,
3213 such that the MDS references an object that doesn't exist, writes
3214 to that object will leak memory due to incorrect cleanup in the
3215 error handling path, eventually running out of memory on the OST.
3220 Description: Creating too long symlink causes lustre errors
3221 Details : Check symlink and name lengths before sending requests to MDS.
3224 Frequency : only if flock is enabled (not on by default)
3226 Description: posix locks not released on fd closure on 2.6.9+
3227 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3228 kernels, this caused such locks not to be released on fd close and
3229 then assertions on fs unmount about still used locks.
3232 Frequency : MDS failover only, very rarely
3234 Description: clients may get ASSERTION(granted_lock != NULL)
3235 Details : When request was taking a long time, and a client was resending
3236 a getattr by name lock request. The were multiple lock requests
3237 with the same client lock handle and
3238 mds_getattr_name->fixup_handle_for_resent_request found one of the
3239 lock handles but later failed with ASSERTION(granted_lock != NULL).
3244 Description: handle->h_buffer_credits > 0, assertion failure
3245 Details : h_buffer_credits is zero after truncate, causing assertion
3246 failure. This patch extends the transaction or creates a new
3250 Frequency : NFS re-export or patchless client
3251 Bugzilla : 11179, 10796
3252 Description: Crash on NFS re-export node (__d_move)
3253 Details : We do not want to hash the dentry if we don't have a lock.
3254 But if this dentry is later used in d_move, we'd hit uninitialised
3255 list head d_hash, so we just do this to init d_hash field but
3256 leave dentry unhashed.
3259 Frequency : NFS re-export or patchless client
3261 Description: NFS exports has problem with symbolic link
3262 Details : lustre client didn't properly install dentry when re-exported
3263 to NFS or running patchless client.
3266 Frequency : NFS re-export or patchless client
3268 Description: Various nfs/patchless fixes.
3269 Details : fixes reuse disconected alias for lookup process - this fixes
3270 warning "find_exported_dentry: npd != pd",
3271 fix permission error with open files at nfs.
3272 fix apply umask when do revalidate.
3275 Frequency : occasional
3277 Description: Crash on NFS re-export node
3278 Details : calling clear_page() on the wrong pointer triggered oops in
3279 generic_mapping_read().
3282 Frequency : rarely, using O_DIRECT IO
3284 Description: unaligned directio crashes client with LASSERT
3285 Details : check for unaligned buffers before trying any requests.
3288 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3290 Description: stale data returned from RAID cache
3291 Details : If only a small amount of IO is done to the RAID device before
3292 reading it again it is possible to get stale data from the RAID
3293 cache instead of reading it from disk.
3296 Frequency : always for sles10 kernel
3298 Description: sles10 support
3299 Details : ll_follow_link: compile fixes and using of nd_set_link
3300 under newer kernels.
3303 Frequency : depends on arch, kernel and compiler version, always on sles10
3306 Description: recursive or deep enough symlinks cause stack overflow
3307 Details : getting rid of large stack-allocated variable in
3311 Frequency : depends on hardware
3313 Description: lustre write performance loss in the SLES10 kernel
3314 Details : the performance loss is caused by using of write barriers in the
3315 ext3 code. The SLES10 kernel turns barrier support on by
3316 default. The fix is to undo that change for ldiskfs.
3318 ------------------------------------------------------------------------------
3320 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3322 * Support for kernels:
3323 2.6.9-42.0.3EL (RHEL 4)
3324 2.6.5-7.276 (SLES 9)
3325 2.4.21-47.0.1.EL (RHEL 3)
3326 2.6.12.6 vanilla (kernel.org)
3330 Frequency : quota enabled and large files being deleted
3332 Description: releasing more than 4GB of quota at once hangs OST
3333 Details : If a user deletes more than 4GB of files on a single OST it
3334 will cause the OST to spin in an infinite loop. Release
3335 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3340 Description: statfs data retrieved from /proc may be stale or zero
3341 Details : When reading per-device statfs data from /proc, in the
3342 {kbytes,files}_{total,free,avail} files, it may appear
3343 as zero or be out of date.
3346 Frequency : systems with MD RAID1 external journal devices
3348 Description: lconf's call to blkid is confused by RAID1 journal devices
3349 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3350 of returning all block devices that match the journal UUID.
3353 Frequency : always, for aggregate stripe size over 4GB
3355 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3356 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3357 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3360 Frequency : always if "lfs find" used on a local file/directory
3362 Description: "lfs find" segfaults if used on a local file/directory
3363 Details : The case where a directory component was not specified wasn't
3364 handled correctly. Handle this properly.
3367 Frequency : always on ppc64
3369 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3370 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3371 perform endianness conversion.
3374 Frequency : rarely (truncate to non-zero file size after write under load)
3375 Bugzilla : 10730, 10687
3376 Description: Files padded with zeros to next 4K multiple
3377 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3378 it is possible that files that are truncated to a non-zero size
3379 immediately after being written are filled with zero bytes beyond
3380 the truncated size. No file data is lost.
3382 Severity : enhancement
3384 Description: Allow recovery/failover for liblustre clients.
3385 Details : liblustre clients were unaware of failover configurations until
3388 Severity : enhancement
3390 Description: user file locks should fail when not mounting with flock option
3391 Details : Set up an error-returning stub in ll_file_operations.lock field
3392 to prevent incorrect behaviour when client is mounted without
3393 flock option. Also, set up properly f_op->flock field for
3397 Frequency : always on ia64
3399 Description: "lfs df" loops on printing out MDS statfs information
3400 Details : The obd_ioctl_data was not initialized and in some systems
3401 this caused a failure during the ioctl that did not return
3402 an error. Initialize the struct and return an error on failure.
3405 Frequency : SLES 9 only
3407 Description: Error of copying files with lustre special EAs as root
3408 Details : Client side always return success for setxattr call for lustre
3409 special xattr (currently only "trusted.lov").
3412 Frequency : rarely on clusters with both ia64+i386 clients
3414 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3415 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3416 clients are doing concurrent writes to the same file it is
3417 possible that the ia64 clients may LASSERT because the OST
3418 extent locks are not PAGE_SIZE aligned. Ensure that grown
3419 locks are always aligned on the request boundary.
3422 Frequency : specific use, occasional
3424 Description: Overwriting in use executable truncates on-disk binary image
3425 Details : If one node attempts to overwrite an executable in use by
3426 another node, we now correctly return ETXTBSY instead of
3427 truncating the file.
3429 Severity : enhancement
3431 Description: Async OSC create to avoid the blocking unnecessarily.
3432 Details : If a OST has no remain object, system will block on the creating
3433 when need to create a new object on this OST. Now, ways use
3434 pre-created objects when available, instead of blocking on an
3435 empty osc while others are not empty. If we must block, we block
3436 for the shortest possible period of time.
3441 Description: chmod on Lustre root is propagated to other clients
3442 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3443 invalid by the follow_mount time.
3446 Frequency : liblustre clients only
3448 Description: Race in 'instant cancel' lock handling could lead to such locks
3449 never to be granted in case of SMP MDS
3450 Details : Do not destroy not yet granted but cbpending locks in
3454 Frequency : replay/resend of open
3456 Description: non null lock assetion failure in mds_intent_policy
3457 Details : Trying to replay/resend lockless open requests resulted in
3458 mds_open() returning 0 with no lock. Now it sets a flag if
3459 a lock is going to be returned.
3461 Severity : enhancement
3463 Description: Checksum enhancements
3464 Details : New checksum enhancements allow for resending RPCs that failed
3467 Severity : enhancement
3469 Description: Tunables on number of dirty pages in cacche
3470 Details : Allow to set limit on number of dirty pages cached.
3475 Description: client crash on unmount - lock still has references
3476 Details : In some error handling cases it was possible to leak a lock
3477 reference on a client while accessing a file. This was not
3478 harmful to the client during operation, but would cause the
3479 client to crash when the filesystem is unmounted.
3482 Frequency : specific case, rare
3484 Description: ETXTBSY on mds though file not in use by client
3485 Details : ETXTBSY is no longer incorrectly returned when attempting to
3486 chmod or chown a directory that the user previously tried to
3487 execute or a currently-executing binary.
3490 Frequency : extremely rare except on liblustre-based clients
3492 Description: Lustre space not freed when files are deleted
3493 Details : Clean up open-unlinked files after client eviction. Previously
3494 the unlink was skipped and the files remained as orphans.
3499 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3500 Details : In some cases it was possible to send a blocking callback to
3501 a client doing a glimpse, even though that client didn't get
3502 a lock granted. When the glimpse lock is cancelled on the OST
3503 the freed lock is left on the waiting list and corrupted the list.
3506 Frequency : all core dumps
3508 Description: Broke core dumps to lustre
3509 Details : Negative dentry may be unhashed if parent does not have UPDATE
3510 lock, but some callers, e.g. do_coredump, expect dentry to be
3511 hashed after successful create, hash it in ll_create_it.
3513 ------------------------------------------------------------------------------
3515 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3517 * Support for kernels:
3518 2.6.9-42.0.2.EL (RHEL 4)
3519 2.6.5-7.276 (SLES 9)
3520 2.4.21-40.EL (RHEL 3)
3521 2.6.12.6 vanilla (kernel.org)
3525 Frequency : always on RHEL 3
3527 Description: Number of open files grows over time
3528 Details : The number of open files grows over time, whether or not
3529 Lustre is started. This was due to a filp leak introduced
3530 by one of our kernel patches.
3532 ------------------------------------------------------------------------------
3534 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3536 * Support for kernels:
3537 2.6.9-42.EL (RHEL 4)
3538 2.6.5-7.276 (SLES 9)
3539 2.4.21-40.EL (RHEL 3)
3540 2.6.12.6 vanilla (kernel.org)
3545 Bugzilla : 5719, 9635, 9792, 9684
3546 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3547 Details : If a server is under heavy load and cannot reply to new
3548 connection requests before the client resends the (re)connect,
3549 the connection handling code can behave badly if two service
3550 threads are concurrently handing separate (re)connections from
3551 the same client. Add better locking to the connection handling
3552 code, and ensure that only a single connection will be processed
3553 for a given client UUID, even if the lock is dropped.
3555 Severity : enhancement
3557 Description: add TCP zero-copy support to kernel
3558 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3559 sends if the hardware supports scatter-gather and checksumming.
3560 This allows socklnd to do client-write and server-read more
3561 efficiently and reduce CPU utilization from skbuf copying.
3564 Frequency : only if NFS exporting from client
3566 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3567 Details : If setting attributes on a file created under NFS that had
3568 never been opened it would be possible to oops the client
3569 if the file had no objects.
3572 Frequency : always for liblustre
3574 Description: liblustre client does MDS+OSTs setattr RPC for each write
3575 Details : When doing a write from a liblustre client, the client
3576 incorrectly issued an RPC to the MDS and each OST the file was
3577 striped over in order to update the timestamps. When writing
3578 with small chunks and many clients this could overwhelm the MDS
3579 with RPCs. In all cases it would slow down the write because
3580 these RPCs are unnecessary.
3582 Severity : enhancement
3584 Description: allow number of MDS service threads to be changed at module load
3585 Details : It is now possible to change the number of MDS service threads
3586 running. Adding "options mds mds_num_threads={N}" to the MDS's
3587 /etc/modprobe.conf will set the number of threads for the next
3588 time Lustre is restarted (assuming the "mds" module is also
3589 reloaded at that time). The default number of threads will
3590 stay the same, 32 for most systems.
3595 Description: OST crash if filesystem is unformatted or corrupt
3596 Details : If an OST is started on a device that has never been formatted
3597 or if the filesystem is corrupt and cannot even mount then the
3598 error handling cleanup routines would dereference a NULL pointer.
3603 Description: NULL pointer deref in llap_from_page.
3604 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3605 incorrect) mapping if the page was truncated/reclaimed while it was
3606 searched for. Check for this condition and skip such pages when
3607 doing readahead. Introduce extra check to llap_from_page() to
3608 verify page->mapping->host is non-NULL (so page is not anonymous).
3611 Frequency : Sometimes when using sys_sendfile
3613 Description: "page not covered by a lock" warnings from ll_readpage
3614 Details : sendfile called ll_readpage without right page locks present.
3615 Now we introduced ll_file_sendfile that does necessary locking
3616 around call to generic_file_sendfile() much like we do in
3620 Frequency : with certain MDS communication failures at client mount time
3622 Description: NULL pointer deref after failed client mount
3623 Details : a client connection request may delayed by the network layer
3624 and not be sent until after the PTLRPC layer has timed out the
3625 request. If the client fails the mount immediately it will try
3626 to clean up before the network times out the request. Add a
3627 reference from the request import to the obd device and delay
3628 the cleanup until the network drops the request.
3631 Frequency : occasionally during client (re)connect
3633 Description: assertion failure during client (re)connect
3634 Details : processing a client connection request may be delayed by the
3635 client or server longer than the client connect timeout. This
3636 causes the client to resend the connection request. If the
3637 original connection request is replied in this interval, the
3638 client may trip an assertion failure in ptlrpc_connect_interpret()
3639 which thought it would be the only running connect process.
3642 Frequency : only with obd_echo servers and clients that are rebooted
3644 Description: kernel BUG accessing uninitialized data structure
3645 Details : When running an obd_echo server it did not start the ping_evictor
3646 thread, and when a client was evicted an uninitialized data
3647 structure was accessed. Start the ping_evictor in the RPC
3648 service startup instead of the OBD startup.
3650 Severity : enhancement
3651 Bugzilla : 10193 (patchless)
3652 Description: Remove dependency on various unexported kernel interfaces.
3653 Details : No longer need reparent_to_init, exit_mm, exit_files,
3654 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3657 Frequency : rare (only users of deprecated and unsupported LDAP config)
3659 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3660 Details : LDAP apparently contains 'lustreName' attributes instead of
3661 'name'. A simple remapping of the name is sufficient.
3664 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3666 Description: waiting_locks_callback trips kernel BUG if client is evicted
3667 Details : Running with the dump_on_timeout debug flag turned on makes
3668 it possible that the waiting_locks_callback() can try to dump
3669 the Lustre kernel debug logs from an interrupt handler. Defer
3670 this log dumping to the expired_lock_main() thread.
3672 Severity : enhancement
3674 Description: Support NFS exporting on 2.6 kernels.
3675 Details : Implement non-rawops metadata methods for NFS server to use without
3676 changing NFS server code.
3679 Frequency : very rare (synthetic metadata workload only)
3681 Description: two racing renames might cause an MDS thread to deadlock
3682 Details : Running the "racer" program may cause one MDS thread to rename
3683 a file from being the source of a rename to being the target of
3684 a rename at exactly the same time that another thread is doing
3685 so, and the second thread has already enqueued these locks after
3686 doing a lookup of the target and is trying to relock them in
3687 order. Ensure that we don't try to re-lock the same resource.
3690 Frequency : only very large systems with liblustre clients
3692 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3693 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3694 all OSTs in parallel. This allows the eviction of stale liblustre
3695 clients to proceed much faster than if they were done in series,
3696 and also offers similar improvements for other set_info RPCs.
3701 Description: excessive CPU usage during initial read phase on client
3702 Details : During the initial read phase on a client, it would agressively
3703 retry readahead on the file, consuming too much CPU and impacting
3704 performance (since 1.4.5.8). Improve the readahead algorithm
3705 to avoid this, and also improve some other common cases (read
3706 of small files in particular, where "small" is files smaller than
3707 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3712 Description: MDS crash when receiving packet with unknown intent.
3713 Details : Do not LBUG in unknown intent case, just return -EFAULT
3715 Severity : enhancement
3716 Bugzilla : 9293, 9385
3717 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3718 Details : Do not serialize getattr (non-intent version) and statfs.
3721 Frequency : occasional, when OST network is overloaded/intermittent
3723 Description: client evicted by OST after bulk IO timeout
3724 Details : If a client sends a bulk IO request (read or write) the OST
3725 may evict the client if it is unresposive to its data GET/PUT
3726 request. This is incorrect if the network is overloaded (takes
3727 too long to transfer the RPC data) or dropped the OST GET/PUT
3728 request. There is no need to evict the client at all, since
3729 the pinger and/or lock callbacks will handle this, and the
3730 client can restart the bulk request.
3733 Frequency : Always when mmapping file with no objects
3735 Description: client crashes when mmapping file with no objects
3736 Details : Check that we actually have objects in a file before doing any
3737 operations on objects in ll_vm_open, ll_vm_close and
3743 Description: Request leak when working with deleted CWD
3744 Details : Introduce advanced request refcount tracking for requests
3745 referenced from lustre intent.
3747 Severity : Enhancement
3749 Description: Cache open file handles on client.
3750 Details : MDS now will return special lock along with openhandle, if
3751 requested and client is allowed to hold openhandle, even if unused,
3752 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3753 closing files for every read/write openration.
3755 Severity : Enhancement
3757 Description: Cache open negative dentries on client when possible.
3758 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3759 negative dentries on lock revocation.
3764 Description: Remounting a client read-only wasn't possible with a zconf mount
3765 Details : It wasn't possible to remount a client read-only with llmount.
3767 Severity : enhancement
3768 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3769 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3770 MPICH is included to add Lustre-specific ADIO interfaces.
3771 This is based closely on the UFS ADIO layer and only differs
3772 in file creation, in order to allow the OST striping to be set.
3773 This is user-contributed code and not supported by CFS.
3778 Description: extended inode attributes (immutable, append-only) work improperly
3779 when 2.4 and 2.6 kernels are used on client/server or vice versa
3780 Details : Introduce kernel-independent values for these flags.
3782 Severity : enhancement
3785 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3786 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3787 granularity. Now, Fractional MB granularity is be supported,
3788 this is very useful for low memory system.
3790 Severity : enhancement
3792 Description: Getattr by fid
3793 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3794 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3795 have locks and avoids getattr rpc requests for it on every stat.
3798 Frequency : Always, for filesystems larger than 2TB
3800 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3801 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3802 filesystems larger than 2TB to be mounted with mballoc enabled.
3805 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3807 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3808 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3809 filesystems larger than 2TB is mounted then the kernel will
3810 silently corrupt the start of the filesystem. CONFIG_LBD is
3811 enabled for all CFS-supported kernels, but the possibility of
3812 this happening with a modified kernel config exists.
3814 Severity : enhancement
3816 Description: add client O_DIRECT support for 2.6 kernels
3817 Details : It is now possible to do O_DIRECT reads and writes to files
3818 in the Lustre client mountpoint on 2.6 kernel clients.
3820 Severity : enhancement
3822 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3823 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3824 parallel, not waiting for response from every OST before sending
3825 a rpc to the next OST.
3830 Description: setattr vs write race when updating file timestamps
3831 Details : Client processes that update a file timestamp into the past
3832 right after writing to the file (e.g. tar) it is possible that
3833 the updated file modification time can be reset to the current
3834 time due to a race between processing the setattr and write RPC.
3836 Severity : enhancement
3838 Description: Bring 'lfs find' closer in line with regular Linux find.
3839 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3840 -print0 options and obtains all the needed info through the lustre
3843 Severity : enhancement
3845 Description: support up to 1024 configured devices on one node
3846 Details : change obd_dev array from statically allocated to dynamically
3847 allocated structs as they are first used to reduce memory usage
3852 Description: Flush dirty partially truncated pages during truncate
3853 Details : Immediatelly flush partially truncated pages in filter_setattr,
3854 this way we completely avoid having any pages in page cache on OST
3855 and can retire ugly workarounds during writes to flush such pages.
3860 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3861 Details : protect inode from truncation within vfs_unlink() context
3862 just take a reference before calling vfs_unlink() and release it
3863 when parent's i_sem is free.
3868 Description: last_id value checked outside lock on OST caused LASSERT failure
3869 Details : If there were multiple MDS->OST object precreate requests in
3870 flight, it was possible that the OST's last object id was checked
3871 outside a lock and incorrectly tripped an assertion. Move checks
3872 inside locks, and discard old precreate requests.
3875 Frequency : always, if extents are used on OSTs
3877 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3878 Details : OSTs using the extents format would not zero the high 16 bits of
3879 the index physical block number. This is not a problem for any
3880 OST filesystems smaller than 16TB, and no kernels support ext3
3881 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3882 new/modified files) and can be fixed for existing filesystems
3883 with e2fsprogs-1.39-cfs1.
3888 Description: import connection selection may be incorrect if timer wraps
3889 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3890 import connections to be ignored if the 32-bit jiffies counter
3891 wraps. Use a 64-bit jiffies counter.
3894 Frequency : very large clusters immediately after boot
3896 Description: LNET request buffers exhausted under heavy short-term load
3897 Details : If a large number of client requests are generated on a service
3898 that has previously never seen so many requests it is possible
3899 that the request buffer growth cannot keep up with the spike in
3900 demand. Instead of dropping incoming requests, they are held in
3901 the LND until the RPC service can accept more requests.
3904 Frequency : Sometimes during replay
3906 Description: Assertion failure in ll_local_open after replay.
3907 Details : If replay happened on an open request reply before we were able
3908 to set replay handler, reply will become not swabbed tripping the
3909 assertion in ll_local_open. Now we set the handler right after
3910 recognising of open request
3913 Frequency : very rare
3915 Description: kernel reports "badness in vsnprintf"
3916 Details : Reading from the "recovery_status" /proc file in small chunks
3917 may cause a negative length in lprocfs_obd_rd_recovery_status()
3918 call to vsnprintf() (which is otherwise harmless). Exit early
3919 if there is no more space in the output buffer.
3921 Severity : enhancement
3923 Description: clear OBD RPC statistics by writing to them
3924 Details : It is now possible to clear the OBD RPC statistics by writing
3925 to the "stats" file.
3930 Description: Client mtime is not the same on different clients after utimes
3931 Details : In some cases, the client was using the utimes() syscall on
3932 a file cached on another node. The clients now validate the
3933 ctime from the MDS + OSTs to determine which one is right.
3938 Description: Inability to activate failout mode
3939 Details : lconf script incorrectly assumed that in pythong string's numeric
3940 value is used in comparisons.
3943 Frequency : always with multiple stripes per file
3945 Description: Inefficient object allocation for mutli-stripe files
3946 Details : When selecting which OSTs to stripe files over, for files with
3947 a stripe count that divides evenly into the number of OSTs,
3948 the MDS is always picking the same starting OST for each file.
3949 Return the OST selection heuristic to the original design.
3954 Description: mount failures may take full timeout to return an error
3955 Details : Under some heavy load conditions it is possible that a
3956 failed mount can wait for the full obd_timeout interval,
3957 possibly several minutes, before reporting an error.
3958 Instead return an error as soon as the status is known.
3960 Frequency : quota enabled and large files being deleted
3962 Description: releasing more than 4GB of quota at once hangs OST
3963 Details : If a user deletes more than 4GB of files on a single OST it
3964 will cause the OST to spin in an infinite loop. Release
3965 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3970 Description: statfs data retrieved from /proc may be stale or zero
3971 Details : When reading per-device statfs data from /proc, in the
3972 {kbytes,files}_{total,free,avail} files, it may appear
3973 as zero or be out of date.
3976 Frequency : systems with MD RAID1 external journal devices
3978 Description: lconf's call to blkid is confused by RAID1 journal devices
3979 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3980 of returning all block devices that match the journal UUID.
3983 Frequency : always, for aggregate stripe size over 4GB
3985 Description: assertion fails when trying to use 4GB stripe size
3986 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3987 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3990 Frequency : always on ppc64
3992 Description: the first write on an ext3 filesystem with mballoc got stuck
3993 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3994 perform endianness conversion.
3996 ------------------------------------------------------------------------------
3998 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4000 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4001 INTEROPERATE with older versions automatically. Please read the
4002 user documentation before upgrading any part of a live system.
4003 * WARNING: Lustre networking configuration changes are required with
4004 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4007 * Support for kernels:
4008 2.6.9-22.0.2.EL (RHEL 4)
4009 2.6.5-7.244 (SLES 9)
4010 2.6.12.6 vanilla (kernel.org)
4013 Severity : enhancement
4014 Bugzilla : 7981/8208
4015 Description: Introduced Lustre Networking (LNET)
4016 Details : LNET is new networking infrastructure for Lustre, it includes
4017 a reorganized network configuration mode (see the user
4018 documentation for full details) as well as support for routing
4019 between different network fabrics. Lustre Networking Devices
4020 (LNDs) for the supported network fabrics have also been
4021 created for this new infrastructure.
4023 Severity : enhancement
4024 Description: Introduced Access control lists
4025 Details : clients can set ACLs on files and directories in order to have
4026 more fine-grained permissions than the standard Unix UGO+RWX.
4027 The MDS must be started with the "-o acl" mount option.
4029 Severity : enhancement
4030 Description: Introduced filesystem quotas
4031 Details : Administrators may now establish per-user quotas on the
4034 Severity : enhancement
4036 Description: Configuration change for the XT3
4037 The PTLLND is now used to run Lustre over Portals on the XT3
4038 The configure option(s) --with-cray-portals are no longer used.
4039 Rather --with-portals=<path-to-portals-includes> is used to
4040 enable building on the XT3. In addition to enable XT3 specific
4041 features the option --enable-cray-xt3 must be used.
4046 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4047 Details : A race between allocating a new llcd and re-getting the llcd_lock
4048 allowed another thread to grab newly-allocated llcd.
4050 Severity : enhancement
4052 Description: 2.6 OST async journal commit and locking fix to improve performance
4053 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4054 2.6 kernels are now async as they already were in 2.4 kernels so
4055 that they can commit concurrently with the network bulk transfer.
4056 For block-allocated files the filter allocation semaphore is held
4057 to avoid filesystem fragmentation during allocation. BKL lock
4058 removed for 2.6 xattr operations where it is no longer needed.
4063 Description: lconf incorrectly determined whether two IP networks could talk
4064 Details : In some more complicated routing and multiple-network
4065 configurations, lconf will avoid trying to make a network
4066 connection to a disjoint part of the IP space. It was doing the
4067 math incorrectly for one set of cases.
4072 Description: Fix for potential infinite loop processing records in an llog.
4073 Details : If an llog record is corrupted/zeroed, it is possible to loop
4074 forever in llog_process(). Validate the llog record length
4075 and skip the remainder of the block on error.
4078 Frequency : occasional (liblustre only)
4080 Description: liblustre could not open files whose last component is a symlink
4081 Details : sysio_path_walk() would incorrectly pass the open intent to
4082 intermediate path components.
4085 Frequency : rare (liblustre only with non-standard tuning)
4086 Bugzilla : 7201 (7350)
4087 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4088 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4089 after it has been released, only do this for referenced locks
4091 Severity : enhancement
4093 Description: specifying an (invalid) directory default stripe_size of -1
4094 would reset the directory default striping
4095 Details : stripe_size -1 was used internally to signal directory stripe
4096 removal, now use "all default" to signal dir stripe removal
4097 as a directory striping of "all default" is not useful
4100 Frequency : common for large clusters running liblustre clients
4102 Description: doing an ls when liblustre clients are running is slow
4103 Details : sending a glimpse AST to a liblustre client waits for every AST
4104 to time out, as liblustre clients will not respond. Since they
4105 cannot cache data we refresh the OST lock LVB from disk instead.
4107 Severity : enhancement
4109 Description: doing an ls at the same time as file IO can be slow
4110 Details : enqueue and other "small" requests can be blocked behind many
4111 large IO requests. Create a new OST IO portal for non-IO
4112 requests so they can be processed faster.
4115 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4117 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4118 created file with mode 000
4119 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4120 allow an ftruncate() as a non-root user to a file with mode 000.
4121 Lustre now respects this flag to disable mode checks when
4122 truncating a file owned by the user
4125 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4128 Description: Revoking locks from clients that went dead or catatonic might take
4130 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4131 cancellation of such locks instant on servers without waiting for
4132 any reply from clients. Clients drops these locks when cancel
4133 notification from server is received without replying.
4136 Frequency : liblustre-only, when liblustre client dies or becomes busy
4138 Description: Doing ls on Linux clients can take a long time with active
4140 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4141 granting such locks to it in the first place if possible. Locks
4142 are taken by proxy on the OST during the read or write and
4143 dropped immediately afterward. Add connect flags handling, do
4144 not grant locks to liblustre clients for glimpse ASTs.
4146 Severity : enhancement
4148 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4149 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4150 sequential reads and behaves badly with applications doing random
4151 reads. Improve it by reading ahead at least read region, and
4152 avoiding excessive large RPC for small reads.
4154 Severity : enhancement
4156 Description: Creating more than 1000 files for a single job may cause a load
4157 imbalance on the OSTs if there are also a large number of OSTs.
4158 Details : qos_prep_create() uses an OST index reseed value that is an
4159 even multiple of the number of available OSTs so that if the
4160 reseed happens in the middle of the object allocation it will
4161 still utilize the OSTs as uniformly as possible.
4166 Description: OST or MDS may oops in ping_evictor_main()
4167 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4168 but doesn't restart at beginning of obd_exports_timed list
4171 Severity : enhancement
4173 Description: improve by-nid export eviction on the MDS and OST
4174 Details : allow multiple exports with the same NID to be evicted at one
4175 time without re-searching the exports list.
4178 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4180 Description: MDS may oops in groups_free()
4181 Details : in rare race conditions a newly allocated group_info struct is
4182 freed again, and this can be NULL. The 2.4 compatibility code
4183 for groups_free() checked for a NULL pointer, but 2.6 did not.
4186 Frequency : common for liblustre clients doing little filesystem IO
4187 Bugzilla : 9352, 7313
4188 Description: server may evict liblustre clients accessing contended locks
4189 Details : if a client is granted a lock or receives a completion AST
4190 with a blocking AST already set it would not reply to the AST
4191 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4194 Frequency : lfs setstripe, only systems with more than 160 OSTs
4196 Description: unable to set striping with a starting offset beyond OST 160
4197 Details : llapi_create_file() incorrectly limited the starting stripe
4198 index to the maximum single-file stripe count.
4201 Frequency : LDAP users only
4203 Description: lconf did not handle in-kernel recovery with LDAP properly
4204 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4206 Severity : enhancement
4208 Description: bind OST threads to NUMA nodes to improve performance
4209 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4210 node and do their allocations there to localize memory access
4212 Severity : enhancement
4214 Description: llmount can determine client NID directly from Myrinet (GM)
4215 Details : the client NID code from gmnalnid was moved directly into
4216 llmount, removing the need to use this or specifying the
4217 client NID explicitly when mounting GM clients with zeroconf
4220 Frequency : if client is started with down MDS
4222 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4223 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4224 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4225 ptlrpc_import_delayed_req() because we didn't check if the
4226 request was interrupted, and we also didn't break out of the
4227 event loop if there was no timeout
4232 Description: data loss during non-page-aligned writes to a single file from
4233 both multiple nodes and multiple threads on one node at same time
4234 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4235 inconsistency led to false short-reads, that were cached and later
4236 used by ->prepare_write() to fill in partially written page,
4237 leading to data loss.
4240 Frequency : always, if lconf --abort_recovery used
4242 Description: lconf --abort_recovery fails with 'Operation not supported'
4243 Details : lconf was attempting to abort recovery on the MDT device and not
4246 Severity : enhancement
4248 Description: remove cleanup logs
4249 Details : replace lconf-generated cleanup logs with lustre internal
4250 cleanup routines. Eliminates the need for client-cleanup and
4253 Severity : enhancement
4255 Description: add support for EAs (user and system) on lustre filesystems
4256 Details : it is now possible to store extended attributes in the Lustre
4257 client filesystem, and with the user_xattr mount option it
4258 is possible to allow users to store EAs on their files also
4260 Severity : enhancement
4262 Description: Add possibility (config option) to show minimal available OST free
4264 Details : When compiled with --enable-mindf configure option, statfs(2)
4265 (and so, df) will return least minimal free space available from
4266 all OSTs as amount of free space on FS, instead of summary of
4267 free spaces of all OSTs.
4269 Severity : enhancement
4271 Description: do not expand extent locks acquired on OST-side
4272 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4273 by server: they are not cached anyway.
4276 Frequency : when mmap is used/binaries executed from Lustre
4278 Description: Unmmap pages before throwing them away from read cache.
4279 Details : llap_shrink cache now attempts to unmap pages before discarding
4280 them (if unmapping failed - do not discard). SLES9 kernel has
4281 extra checks that trigger if this unmapping is not done first.
4286 Description: lconf didn't resolve symlinks before checking to see whether a
4287 given mountpoint was already in use
4290 Frequency : when migrating failover services
4291 Bugzilla : 6395, 9514
4292 Description: When migrating a subset of services from a node (e.g. failback
4293 from a failover service node) the remaining services would
4294 time out and evict clients.
4295 Details : lconf --force (implied by --failover) sets the global obd_timeout
4296 to 5 seconds in order to quickly disconnect, but this caused
4297 other RPCs to time out too quickly. Do not change the global
4298 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4300 Severity : enhancement
4301 Frequency : if MDS is started with down OST
4302 Bugzilla : 9439,5706
4303 Description: Allow startup/shutdown of an MDS without depending on the
4304 availability of the OSTs.
4305 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4306 Add appropriate locking and lov-osc refcounts for safe
4307 cleaning. Add osc abort_inflight calls in case the
4308 synchronize never started.
4311 Frequency : occasional (Cray XT3 only)
4313 Description: root not authorized to access files in CRAY_PORTALS environment
4314 Details : The client process capabilities were not honoured on the MDS in
4315 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4316 been accessed by an authorized user then root was able to access
4317 the file on the local client also. The root user capabilities
4318 are now allowed on the MDS, as this environment has secure UID.
4321 Frequency : occasional
4323 Description: ldiskfs "too long searching" message happens too often
4324 Details : A debugging message (otherwise harmless) prints too often on
4325 the OST console. This has been reduced to only happen when
4326 there are fragmentation problems on the filesystem.
4331 Description: Division by zero in statfs when all OSCs are inactive
4332 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4333 lov_statfs divides by value returned by lov_get_stripecnt().
4337 Bugzilla : 9489, 3273
4338 Description: First write from each client to each OST was only 4kB in size,
4339 to initialize client writeback cache, which caused sub-optimal
4340 RPCs and poor layout on disk for the first writen file.
4341 Details : Clients now request an initial cache grant at (re)connect time
4342 and so that they can start streaming writes to the cache right
4343 away and always do full-sized RPCs if there is enough data.
4344 If the OST is rebooted the client also re-establishes its grant
4345 so that client cached writes will be honoured under the grant.
4350 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4351 Details : Now I/O RPCs go to different portal number and (presumably) fast
4352 lock requests (and glimses) and other RPCs get their own service
4353 threads pool that should be able to service those RPCs
4356 Severity : enhancement
4358 Description: Ability to exchange lustre version between client and servers and
4359 issue warnings at client side if client is too old. Also for
4360 liblustre clients there is ability to refuse connection of too old
4362 Details : New 'version' field is added to connect data structure that is
4363 filled with version info. That info is later checked by server and
4367 Frequency : rare, liblustre only.
4368 Bugzilla : 9296, 9581
4369 Description: Two simultaneous writes from liblustre at offset within same page
4370 might proceed at the same time overwriting eachother with stale
4372 Details : I/O lock withing llu_file_prwv was released too early, before data
4373 actually was hitting the wire. Extended lock-holding time until
4374 server acknowledges receiving data.
4377 Frequency : extremely rare. Never observed in practice.
4379 Description: avoid generating lustre_handle cookie of 0.
4380 Details : class_handle_hash() generates handle cookies by incrementing
4381 global counter, and can hit 0 occasionaly (this is unlikely, but
4382 not impossible, because initial value of cookie counter is
4383 selected randonly). Value of 0 is used as a sentinel meaning
4384 "unassigned handle" --- avoid it. Also coalesce two critical
4385 sections in this function into one.
4387 Severity : enhancement
4389 Description: allow liblustre clients to delegate truncate locking to OST
4390 Details : To avoid overhead of locking, liblustre client instructs OST to
4391 take extent lock in ost_punch() on client's behalf. New connection
4392 flag is added to handle backward compatibility.
4394 Severity : enhancement
4395 Bugzilla : 4928, 7341, 9758
4396 Description: allow number of OST service threads to be specified
4397 Details : a module parameter allows the number of OST service threads
4398 to be specified via "options ost ost_num_threads={N}" in the
4399 OSS's /etc/modules.conf or /etc/modprobe.conf.
4403 Bugzilla : 6146, 9635, 9895
4404 Description: servers crash with bad pointer in target_handle_connect()
4405 Details : In rare cases when a client is reconnecting it was possible that
4406 the connection request was the last reference for that export.
4407 We would temporarily drop the export reference and get a new
4408 one, but this may have been the last reference and the export
4409 was just destroyed. Get new reference before dropping old one.
4411 Severity : enhancement
4412 Frequency : if client is started with failover MDS
4414 Description: Allow multiple MDS hostnames in the mount command
4415 Details : Try to read the configuration from all specified MDS
4416 hostnames during a client mount in case the "primary"
4419 Severity : enhancement
4421 Description: Stop sending data to evicted clients as soon as possible.
4422 Details : Check if the client we are about to send or are sending data to
4423 was evicted already. (Check is done every second of waiting,
4424 for which l_wait_event interface was extended to allow checking
4425 of exit condition at specified intervals).
4428 Frequency : rare, normally only when NFS exporting is done from client
4430 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4432 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4435 Frequency : always, if srand() is called before liblustre initialization
4437 Description: Liblustre uses system PRNG disturbing its usage by user application
4438 Details : Introduce internal to lustre fast and high-quality PRNG for
4439 lustre usage and make liblustre and some other places in generic
4440 lustre code to use it.
4442 Severity : enhancement
4443 Bugzilla : 9477, 9557, 9870
4444 Description: Verify that the MDS configuration logs are updated when xml is
4445 Details : Check if the .xml configuration logs are newer than the config
4446 logs stored on the MDS and report an error if this is the case.
4447 Request --write-conf, or allow starting with --old_conf.
4449 Severity : enhancement
4451 Description: Handle symlinks in the path when checking if Lustre is mounted.
4452 Details : Resolve intermediate symlinks when checking if a client has
4453 mounted a filesystem to avoid duplicate client mounts.
4458 Description: lconf can hit an error exception but still return success.
4459 Details : The lconf command catches the Command error exception at the top
4460 level script context and will exit with the associated exit
4461 status, but doesn't ensure that this exit status is non-zero.
4466 Description: failure of ptlrpc thread startup can cause oops
4467 Details : Starting a ptlrpc service thread can fail if there are a large
4468 number of threads or the server memory is very fragmented.
4469 Handle this without oopsing.
4472 Frequency : always, only if liblustre and non-default acceptor port was used
4474 Description: liblustre cannot connect to servers with non-default acceptor port
4475 Details : tcpnal_set_default_params() was not called and was therefore
4476 ignoring the environment varaible TCPNAL_PORT, as well as other
4477 TCPNAL_ environment variables
4482 Description: two objects could be created on the same OST for a single file
4483 Details : If an OST is down, in some cases it was possible to create two
4484 objects on a single OST for a single file. No problems other
4485 than potential performance impact and spurious error messages.
4489 Bugzilla : 5681, 9562
4490 Description: Client may oops in ll_unhash_aliases
4491 Details : Client dcache may become inconsistent in race condition.
4492 In some cases "getcwd" can fail if the current directory is
4498 Description: Inode refcounting problems in NFS export code
4499 Details : link_raw functions used to call d_instantiate without obtaining
4500 extra inode reference first.
4504 Bugzilla : 9942, 9903
4505 Description: Referencing freed requests leading to crash, memleaks with NFS.
4506 Details : We used to require that call to ll_revalidate_it was always
4507 followed by ll_lookup_it. Also with revalidate_special() it is
4508 possible to call ll_revalidate_it() twice for the same dentry
4509 even if first occurence returned success. This fix changes semantic
4510 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4511 reference on a request referred from the intent.
4512 ll_intent_release() then releases such a request.
4515 Frequency : rare, normally benchmark loads only
4517 Description: unlinked inodes were kept in memory on the client
4518 Details : If a client is repeatedly creating and unlinking files it
4519 can accumulate a lot of stale inodes in the inode slab cache.
4520 If there is no other client load running this can cause the
4521 client node to run out of memory. Instead flush old inodes
4522 from client cache that have the same inode number as a new inode.
4525 Frequency : SLES9 2.6.5 kernel and long filenames only
4526 Bugzilla : 9969, 10379
4527 Description: utime reports stale NFS file handle
4528 Details : SLES9 uses out-of-dentry names in some cases, which confused
4529 the lustre dentry revalidation. Change it to always use the
4533 Frequency : rare, unless heavy write-truncate concurrency is continuous
4534 Bugzilla : 4180, 6984, 7171, 9963, 9331
4535 Description: OST becomes very slow and/or deadlocked during object unlink
4536 Details : filter_destroy() was holding onto the parent directory lock
4537 while truncating+unlinking objects. For very large objects this
4538 may block other threads for a long time and slow overall OST
4539 responsiveness. It may also be possible to get a lock ordering
4540 deadlock in this case, or run out of journal credits because of
4541 the combined truncate+unlink. Solution is to do object truncate
4542 first in one transaction without parent lock, and then do the
4543 final unlink in a new transaction with the parent lock. This
4544 reduces the lock hold time dramatically.
4547 Frequency : rare, 2.4 kernels only
4549 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4550 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4551 before unmounting their filesystem in order to avoid deadlock.
4552 The kernel_locked() function in 2.4 kernels only checks whether
4553 the kernel lock is held, not whether it is this process that is
4554 holding it as 2.6 kernels do.
4559 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4560 Details : The client ptlrpc code may be trying to reconnect to a down
4561 server before a previous connection attempt has timed out.
4562 Increase the reconnect interval to be longer than the connection
4563 timeout interval to avoid sending duplicate connections to
4567 Frequency : echo_client brw_test command
4569 Description: fix echo_client to work with OST preallocated code
4570 Details : OST preallocation code (5137) didn't take echo_client IO path
4571 into account: echo_client calls filter methods outside of any
4572 OST thread and, hence, there is no per-thread preallocated
4573 pages and buffers to use. Solution: hijack pga pages for IO. As
4574 a byproduct, this avoids unnecessary data copying.
4578 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4579 Description: Client can oops in mdc_commit_close() after open replay
4580 Details : It was possible for the MDS to return an open request with no
4581 transaction number in mds_finish_transno() if the client was
4582 evicted, but without actually returning an error. Clients
4583 would later try to replay that open and may trip an assertion
4584 Simplify the client close codepath, and always return an error
4585 from the MDS in case the open is not successful.
4588 Frequency : rare, 2.6 OSTs only
4590 Description: OST may deadlock under high load on fragmented files
4591 Details : If there was a heavy load and highly-fragmented OST filesystems
4592 it was possible to have all the OST threads deadlock waiting on
4593 allocation of biovecs, because the biovecs were not released
4594 until the entire RPC IO was completed. Instead, release biovecs
4595 as soon as they are complete to ensure forward IO progress.
4597 Severity : enhancement
4599 Description: Support for specifying external journal device at mount
4600 Details : If an OST or MDS device is formatted with an external journal
4601 device, this device major/minor is stored in the ext3 superblock
4602 and may not be valid for failover. Allow detecting and
4603 specifying the external journal at mount time.
4608 Description: Mounting an MDS with pending unlinked files may cause oops
4609 Details : target_finish_recovery() calls mds_postrecov() which returned
4610 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4611 considers this an error and immediately begins cleaning up the
4612 lov, just after starting the mds_lov process
4614 Severity : enhancement
4616 Description: Implement 'lfs df' to report actual free space on per-OST basis
4617 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4618 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4619 usage of objects; '-h' to report in human readable format.
4621 ------------------------------------------------------------------------------
4623 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4630 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4631 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4632 at startup during orphan unlinking. Instead, skip buddy update
4633 before setup, as it will be regenerated after recovery is complete.
4638 Description: If an OST is inactive, its locks might reference stale inodes.
4639 Details : lov_change_cbdata() must iterate over all namespaces, even if
4640 they are inactive to clear inode references from the lock.
4642 Severity : enhancement
4643 Frequency : occasional, if non-standard max_dirty_mb used
4645 Description: Client will block write RPCs if not enough grant
4646 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4647 then the client will block writes while waiting for another RPC
4648 to complete instead of consuming its dirty limit. With change
4649 we get improved performance when max_dirty_mb is small.
4651 Severity : enhancement
4652 Bugzilla : 3389, 6253
4653 Description: Add support for supplementary groups on the MDS.
4654 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4655 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4656 lookups for user supplementary groups into a cache.
4660 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4661 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4662 for liblustre O_EXCL is enforced by MDS.
4666 Description: readdir never returns NULL in liblustre.
4667 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4668 dentry in struct dirent.
4672 Frequency : liblustre only, depends on application IO pattern
4673 Description: liblustre clients evicted if not contacting servers
4674 Details : Don't put liblustre clients into the ping_evictor list, so
4675 they will not be evicted by the pinger ever.
4677 Severity : enhancement
4679 Description: Add ability to evict clients by NID from MDS.
4680 Details : By echoing "nid:$NID" string into
4681 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4682 $NID would be instantly evicted from this MDS and from all active
4683 OSTs connected to it.
4687 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4688 Details : lookup_it_finish() used to query file size from OSTs that was not
4693 Description: service threads change working directory to that of init
4694 Details : Starting lustre service threads may pin the working directory
4695 of the parent thread, making that filesystem busy. Threads
4696 now change to the working directory of init to avoid this.
4700 Frequency : during shutdown only
4701 Description: shutdown with a failed MDS or OST can cause unmount to hang
4702 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4707 Frequency : 2.6 only
4708 Description: chmod/chown may include an extra supplementary group
4709 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4710 supplementary group and if none is specified this is used.
4713 Bugzilla : 5479 (6816)
4715 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4716 Details : If lookup is in progress on a file that is unlinked we might try
4717 to revalidate the inode and fail in revalidate after lookup is
4718 complete and ll_file_open() enqueues the open again but
4719 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4722 Frequency : always, if lconf --abort_recovery used
4724 Description: lconf --abort_recovery fails with 'Operation not supported'
4725 Details : lconf was attempting to abort recovery on the MDT device and not
4728 ------------------------------------------------------------------------------
4730 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4735 Frequency : rare (only unsupported configurations with a node running as an
4737 Bugzilla : 6514, 5137
4738 Description: Mounting a Lustre file system on a node running as an OST could
4740 Details : OSTs now preallocates memory needed to write out data at
4741 startup, instead of when needed, to avoid having to
4742 allocate memory in possibly low memory situations.
4743 Specifically, if the file system is mounted on on OST,
4744 memory pressure could force it to try to write out data,
4745 which it needed to allocate memory to do. Due to the low
4746 memory, it would be unable to do so and the node would
4747 become unresponsive.
4749 Severity : enhancement
4751 Description: Addition of lconf --service command line option
4752 Details : lconf now accepts a '--service <arg>' option, which is
4753 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4755 Severity : enhancement
4757 Description: Failover mode is now the default for OSTs.
4758 Details : By default, OSTs will now run in failover mode. To return to
4759 the old behaviour, add '--failout' to the lmc line for OSTs.
4761 Severity : enhancement
4763 Description: Health checks are now provided for MDS and OSTs
4764 Details : Additional detailed health check information on MSD and OSTs
4765 is now provided through the procfs health_check value.
4768 Frequency : occasional, depends on IO load
4770 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4771 after numerous create/delete cycles
4772 Details : The ext3 inode allocation policy would not allocate new inodes
4773 very well on the OSTs because there are no new directories
4774 being created. Instead we look for groups with free space if
4775 the parent directories are nearly full.
4780 Description: Network or server problems during mount may cause partially
4781 mounted clients instead of returning an error.
4782 Details : The config llog parsing code may overwrite the error return
4783 code during mount error handling, returning success instead
4789 Description: MDS can fail to allocate large reply buffers
4790 Details : After long uptimes the MDS can fail to allocate large reply
4791 buffers (e.g. zconf client mount config records) due to memory
4792 fragmentation or consumption by the buffer cache. Preallocate
4793 some large reply buffers so that these replies can be sent even
4794 under memory pressure.
4798 Frequency : rare (liblustre)
4799 Description: fsx running with liblustre complained that using truncate() to
4800 extend the file doesn't work. This patch corrects that issue.
4801 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4802 ATTR_SIZE and lsm use in llu_setattr_raw.
4806 Frequency : rare, only 2.6 kernels
4807 Description: Unusual file access patterns on the MDS may result in inode
4808 data being lost in very rare circumstances.
4809 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4810 code in the 2.6 kernel caused the inode and/or EA data not to
4811 be read from disk, causing single-file corruption.
4815 Frequency : rare, only 2.6 filesystems using extents
4816 Description: Heavy concurrent write and delete load may cause data corruption.
4817 Details : It was possible under high-load situations to have an extent
4818 metadata block in the block device cache from a just-unlinked
4819 file overwrite a newly-allocated data block. We now unmap any
4820 metadata buffers that alias just-allocated data blocks.
4824 Frequency : filesystems with default stripe_count larger than 77
4825 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4826 Details : lconf specifies an inode size of 4096 bytes when the default
4827 stripe_count is larger than 77. This conflicts with the default
4828 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4829 this case to avoid pinning too much memory for large EAs.
4831 ------------------------------------------------------------------------------
4833 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4838 Frequency : rare (extremely heavy IO load with hundreds of clients)
4840 Description: Client is evicted, gets IO error writing to file
4841 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4842 caused clients to be evicted for AST timeouts. The fixes in
4843 bug 5192 mean we no longer need to have such short AST timeouts
4844 so ldlm_timeout has been increased.
4847 Frequency : occasional during --force or --failover shutdown under load
4848 Bugzilla : 5949, 4834
4849 Description: Server oops/LBUG if stopped with --force or --failover under load
4850 Details : a collection of import/export refcount and cleanup ordering
4851 issues fixed for safer force cleanup
4854 Frequency : only filesystems larger than 120 OSTs
4855 Bugzilla : 5990, 6223
4856 Description: lfs getstripe would oops on a very large filesystem
4857 Details : lov_getconfig used kfree on vmalloc'd memory
4860 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4861 Bugzilla : 6242, 6243
4862 Description: reading from files that had been truncated to a non-zero size
4863 but never opened returned no data
4864 Details : ll_file_read() reads zeros from no-object files to EOF
4869 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4870 Details : There's little harm in aborting MDS/OSS recovery and letting it
4871 try again, so I removed the LASSERT and return an error instead.
4873 Severity : enhancement
4875 Description: New debugging infrastructure for tracking down data corruption
4876 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4877 (b) cover more of the client-side code path, and (c) try to narrow
4878 down where problems occurred
4882 Bugzilla : 3819, 4364, 4397, 6313
4883 Description: Racing close and eviction MDS could cause assertion in mds_close
4884 Details : It was possible to get multiple mfd references during close and
4885 client eviction, leading to one thread referencing a freed mfd.
4887 Severity: : enhancement
4888 Bugzilla : 3262, 6359
4889 Description: Attempts to reconnect to servers are now more aggressive.
4890 Details : This builds on the enhanced upcall-less recovery that was added
4891 in 1.4.2. When trying to reconnect to servers, clients will
4892 now try each server in the failover group every 10 seconds. By
4893 default, clients would previously try one server every 25 seconds.
4898 Description: After recovery, certain operations trigger a failed
4899 assertion on a client.
4900 Details : Failing over an mds, using lconf -d --failover, while a
4901 client was doing a readdir() call would cause the client to
4902 LBUG after recovery completed and the readdir() was resent.
4904 Severity : enhancement
4906 Description: Default groups are now added by lconf
4907 Details : You can now run lconf --group <servicename> without having to
4908 manually add groups with lmc.
4911 Frequency : occasional
4913 Description: Nodes with an elan id of 0 trigger a failed assertion
4916 Frequency : always when accessing e.g. tty/console device nodes
4918 Description: tty and some other devices nodes cannot be used on lustre
4919 Details : file's private_data field is used by device data and lustre
4920 values in there got lost. New field was added to struct file to
4921 store fs-specific private data.
4924 Frequency : when exporting Lustre via NFS
4926 Description: NFSD failed occasionally when looking up a path component
4927 Details : NFSD is looking up ".." which was broken in ext3 directories
4928 that had grown large enough to become hashed.
4931 Frequency : Clusters with multiple interfaces not on the same subnet
4933 Description: Nodes will repeatedly try to reconnect to an interface which it
4934 cannot reach and report an error to the log.
4935 Details : Extra peer list entries will be created by lconf with some peers
4936 unreachable. lconf now validates the peer before adding it.
4939 Frequency : Only if a default stripe is set on the filesystem root.
4941 Description: Setting a default stripe on the filesystem root prevented the
4942 filesystem from being remounted.
4943 Details : The client was sending extra request flags in the root getattr
4944 request and did not allocate a reply buffer for the dir EA.
4947 Frequency : occasional, higher if lots of files are accessed by one client
4948 Bugzilla : 6159, 6097
4949 Description: Client trips assertion regarding lsm mismatch/magic
4950 Details : While revalidating inodes the VFS looks up inodes with ifind()
4951 and in rare cases can find an inode that is being freed.
4952 The ll_test_inode() code will free the lsm during ifind()
4953 when it finds an existing inode and then the VFS later attaches
4954 this free lsm to a new inode.
4958 Bugzilla : 6422, 7030
4959 Description: MDS deadlock between mkdir and client eviction
4960 Details : Creating a new file via mkdir or mknod (starting a transaction
4961 and getting the ns lock) can deadlock with client eviction
4962 (gets ns lock and trying to finish a synchronous transaction).
4965 Frequency : occasional
4966 Description: While starting a server, the fsfilt_ext3 module could not be
4968 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4969 kernels. Previously, lconf would still use the ext3 name
4970 when trying to load modules. Now, it will correctly use
4971 ext3 on 2.4 and ldiskfs on 2.6.
4973 Severity : enhancement
4974 Description: The default stripe count has been changed to 1
4975 Details : The interpretation of the default stripe count (0, to lfs
4976 or lmc) has been changed to mean striping across a single
4977 OST, rather than all available. For general usage we have
4978 found a stripe count of 1 or 2 works best.
4980 Severity : enhancement
4981 Description: Add support for compiling against Cray portals.
4982 Details : Conditional compiling for some areas that are different
4986 Frequency : occasional
4987 Bugzilla : 6409, 6834
4988 Description: Creating files with an explicit stripe count may lead to
4989 a failed assertion on the MDS
4990 Details : If some OSTs are full or unavailable, creating files may
4991 trigger a failed assertion on the MDS. Now, Lustre will
4992 try to use other servers or return an error to the
4996 Frequency : occasional
4998 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4999 caused lock timeout during readahead (since 1.4.2).
5000 Details : Processes doing readahead might match a lock that hasn't been
5001 granted yet if there are overlapping and conflicting lock
5002 requests. The readahead process waits on ungranted lock
5003 (original lock is CBPENDING), while OST waits for that process
5004 to cancel CBPENDING read lock and eventually evicts client.
5006 Severity : enhancement
5008 Description: Initial enabling of flock support for clients
5009 Details : Implements fcntl advisory locking and file status functions.
5010 This feature is provided as an optional mount flag (default
5011 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5012 locking are implemented yet, and those that are are not guaranteed
5013 to be completely correct in production environments.
5014 mount -t lustre -o [flock|noflock] ...
5017 Frequency : occasional
5019 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5020 assertion in the ext3 JBD (journaling) layer.
5021 Details : The b_committed_data struct is protected by the big kernel lock
5022 in 2.4 kernels, serializing journal_commit_transaction() and
5023 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5024 access to this struct. In 2.6 kernels there is finer grained
5025 locking to improve SMP performance of the JBD layer.
5029 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5034 Description: merge and cleanup kernel patches.
5037 -----------------------------------------------------------------------------
5039 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5041 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5042 versions of Lustre. Please update all servers and clients to
5043 version 1.4.2 or later at the same time. You must also run
5044 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5045 to update the configuration logs.
5047 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5048 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5049 - swab llog records of type '0' so we get proper header size/idx (5861)
5050 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5051 - fix rename of one directory over another leaking an inode (5953)
5052 - avoid SetPageDirty on 2.6 (5981)
5053 - don't re-add just-being-destroyed locks to the waiting list (5653)
5054 - when creating new directories, inherit the parent's custom
5055 striping settings if present parent (3048)
5056 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5057 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5058 - don't hold client locks on temporary worklist from l_lru (5666)
5059 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5060 - automatically evict dead clients (5921)
5061 - Update file size properly in create+truncate+fstat case (6196)
5062 - Do not unhash mountpoint dentries, do not allow removal of
5064 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5065 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5066 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5067 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5068 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5069 - Correctly handle reads of files with no objects (6243)
5070 - lctl recover will also mark a device active if deactivate used (5933)
5072 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5073 - allow --write-conf on an MDS with different nettype than client (5619)
5074 - don't write config llogs to MDS for mounts not from that MDS (5617)
5075 - lconf should create multiple TCP connections from a client (5201)
5076 - init scripts are now turned off by default; run chkconfig --on
5077 lustre and chkconfig --on lustrefs to use them
5078 - upcalls are no longer needed for clients to recover to failover
5080 - add --abort-recovery option to lconf to abort recovery on device
5082 - add support for an arbitrary number of OSTs (3026)
5083 - Quota support protocol changes.
5084 - forward compatibility changes to wire structs (6007)
5085 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5086 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5087 - improved "lustre status" script
5088 - initialize blocksize for non-regular files (6062)
5089 - added --disable-server and --disable-client configure options (5782)
5090 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5091 - Vanilla 2.4.29 support
5092 - increase maximum number of obd devices to 520 (6242)
5093 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5094 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5095 SLES 9 2.6.5 kernels
5096 - limit stripes per file to 160 (the maximum EA size) (6093)
5098 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5101 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5102 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5103 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5104 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5105 - small ext3 extents cleanups and fixes (5733)
5106 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5107 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5108 - avoid needless client->OST connect, fix handle mismatch (5317)
5109 - fix DLM error path that led to out-of-sync client, long delays (5779)
5110 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5111 - fix several locking issues related to i_size (5492,5624,5654,5672)
5112 - don't move pending lock onto export if it is already evicted (5683)
5113 - fix kernel oops when creating .foo in unlinked directory (5548)
5114 - fix deadlock in obdfilter statistics vs. object create (5811)
5115 - use time_{before,after} to avoid timer jiffies wrap (5882)
5116 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5117 - Do not leak request if server was not able to process it (5154)
5118 - If mds_open unable to find parent dir, make that negative lookup(5154)
5119 - don't create new directories with extent-mapping (5909, 5936)
5121 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5122 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5123 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5124 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5125 - support for RHEL4 kernel on i686 (5773)
5126 - provide error messages when incompatible logs are encountered (5898)
5128 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5129 * version 1.4.0.10 (1.4.1 release candidate 1)
5131 - don't keep a lock reference when lock is not granted (4238)
5132 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5133 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5134 - drop import inflight refcount on signal_completed_replay error (5255)
5135 - unlock page after async write error during send (3677)
5136 - handle missing objects in filter_preprw_read properly (5265)
5137 - no transno return for symlink open, don't save no-trasno open (3440)
5138 - don't try to complete elan receive that already failed (4012)
5139 - free RPC server reply state on error (5406)
5140 - clean up thread from ptlrpc_start_thread() on error (5160)
5141 - readahead could read extra page into cache that wasn't ejected (5388)
5142 - prevent races in class_attach/setup/cleanup/detach (5260)
5143 - don't dereference de->d_inode after l_dput of de (5458)
5144 - use "int" for stripe value returned from lock_to_stripe (5544)
5145 - mballoc allocation and error-checking fixes in 2.6 (5504)
5146 - block device patches to fix I/O request sizes in 2.6 (5482)
5147 - look up hostnames for IB nals (5602)
5148 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5149 - don't start multiple acceptors for the same port (5277)
5150 - fix incorrect LASSERT in mds_getattr_name (5635)
5151 - export a proc file for general "ping" checking (5628)
5152 - fix "lfs check" to not block when the MDS is down (5628)
5154 - service request history (4965)
5155 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5156 - create an "evict_client" /proc entry on OSTs, like the MDS has
5157 - fix mount usage message, return errors per mount(8) (5168)
5158 - change grep [] to grep "[]" in tests so they work in more UMLs
5159 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5160 - remove ancient LOV_MAGIC_V0 EA support (5047)
5161 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5162 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5163 - allow readahead allocations to fail when low on memory (5383)
5164 - mmap locking landed again, after considerable improvement (2828)
5165 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5167 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5170 - send OST transaction number in read/write reply to free req (4966)
5171 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5172 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5173 - the watchdog thread now runs as interruptible (5246)
5174 - flock/lockf fixes (but it's still disabled, pending 5135)
5175 - don't use EXT3 constants in llite code (5094)
5176 - memory shortage at startup could cause assertion (5176)
5178 - reorganization of lov code
5179 - single portals codebase
5181 - add extents/mballoc support (5025)
5182 - direct I/O reads in the obdfilter (4048)
5183 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5185 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5188 - send OST transaction number in read/write reply to free req (4966)
5189 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5190 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5191 - the watchdog thread now runs as interruptible (5246)
5192 - handle missing objects in filter_preprw_read properly (5265)
5193 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5194 - drop import inflight refcount on signal_completed_replay error (5255)
5195 - unlock page after async write error during send (3677)
5196 - return original error code on reconstructed replies (3761)
5197 - no transno return for symlink open, don't save no-trasno open (3440)
5199 - add pid to ldlm debugging output (4922)
5200 - bump the watchdog timeouts -- we can't handle 30sec yet
5201 - extra debugging for orphan dentry/inode bug (5259)
5203 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5206 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5207 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5208 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5209 - fix NULL dereference in /proc/sys/portals/routes (4827)
5210 - allow failed mdc_close() operations to be interrupted (4561)
5211 - stop precreate on OST before MDS would time out on it (4778)
5212 - don't send partial-page writes before EOF from client (4410)
5213 - discard client grant for sub-page writes on large-page clients (4520)
5214 - don't free dentries not owned by NFS code, check generation (4806)
5215 - fix lsm leak if mds_create_objects() fails (4801)
5216 - limit debug_daemon file size, always print CERROR messages (4789)
5217 - use transno after validating reply (3892)
5218 - process timed out requests if import state changes (3754)
5219 - update mtime on OST during writes, return in glimpse (4829)
5220 - add mkfsoptions to LDAP (4679)
5221 - use ->max_readahead method instead of zapping global ra (5039)
5222 - don't interrupt __l_wait_event() during strace
5224 - add software watchdogs to catch hung threads quickly (4941)
5225 - make lustrefs init script start after nfs is mounted
5226 - fix CWARN/ERROR duplication (4930)
5227 - return async write errors to application if possible (2248)
5228 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5229 - print NAL number in %x format (4645)
5230 - update barely-supported suse-2.4.21-171 series (4842)
5231 - support for sles 9 %post scripts
5232 - support for building 2.6 kernel-source packages
5233 - support for sles km_* packages
5235 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5238 - ignore -ENOENT errors in osc_destroy (3639)
5239 - notify osc create thread that OSC is being cleaned up (4600)
5240 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5241 - reconstruct ost_handle() like mds_handle() (4657)
5242 - create a new thread to do import eviction to avoid deadlock (3969)
5243 - let lconf resolve symlinked-to devices (4629)
5244 - don't unlink "objects" from directory with default EA (4554)
5245 - hold socknal file ref over connect in case target is down (4394)
5246 - allow more than 32000 subdirectories in a single directory (3244)
5247 - fix blocks count for O_DIRECT writes (3751)
5248 - OST returns ENOSPC from object create when no space left (4539)
5249 - don't send truncate RPC if file size isn't changing (4410)
5250 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5251 - bind to privileged port in socknal and tcpnal (3689)
5253 - rate limit CERROR/CWARN console message to avoid overload (4519)
5254 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5255 - basic mmap support (3918)
5256 - kernel patch series update from b1_4 (4711)
5258 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5261 - avoid crash during MDS cleanup with OST shut down (2775)
5262 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5263 - don't use bad inodes on the MDS (3744)
5264 - dynamic object preallocation to improve recovery speed (4236)
5265 - don't hold spinlock over lock dumping or change debug flags (4401)
5266 - don't zero obd_dev when it is force cleaned (3651)
5267 - print grants to console if they go negative (4431)
5268 - "lctl deactivate" will stop automatic recovery attempts (3406)
5269 - look for existing locks in ldlm_handle_enqueue() (3764)
5270 - don't resolve lock handle twice in recovery avoiding race (4401)
5271 - revalidate should check working dir is a directory (4134)
5273 - don't always mark "slow" obdfilter messages as errors (4418)
5275 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5278 - don't close LustreDB during write_conf until it is done (3860)
5279 - fix typo in lconf for_each_profile (3821)
5280 - allow dumping logs from multiple threads at one time (3820)
5281 - don't allow multiple threads in OSC recovery (3812)
5282 - fix debug_size parameters (3864)
5283 - fix mds_postrecov to initialize import for llog ctxt (3121)
5284 - replace config semaphore with spinlock (3306)
5285 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5286 - don't allow enqueue to complete on a destroyed export (3822)
5287 - down write_lock before checking llog header bitmap (3825)
5288 - recover from lock replay timeout (3764)
5289 - up llog sem before sending rpc (3652)
5290 - reduce ns lock hold times when setting kms (3267)
5291 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5292 - fix NULL deref and obd_dev leak on setup error (3312)
5293 - replace some LBUG about llog ops with error handling (3841)
5294 - don't match INVALID dentries from d_lookup and spin (3784)
5295 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5296 - fix invalid assertion in ptlrpc_set_wait (3880)
5298 - add libwrap support for the TCP acceptor (3996)
5299 - add /proc/sys/portals/routes for non-root route listing (3994)
5300 - allow setting MDS UUID in .xml (2580)
5301 - print the stack of a process that LBUGs (4228)
5303 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5306 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5307 - make sure to unset replay flag from failed open requests (3440)
5308 - if default stripe count is 0, use OST count for inode size (3636)
5309 - update parent mtime/ctime on client for create/unlink (2611)
5310 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5311 - free recovery state on server during a forced cleanup (3571)
5312 - unregister_reply for resent reqs (3063)
5313 - loop back devices mounting and status check on 2.6 (3563)
5314 - fix resource-creation race that can provoke i_size == 0 (3513)
5315 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5316 - more debugging for page-accounting assertion (3746)
5317 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5318 - avoid deadlock after precreation failure (3758)
5319 - fix race and lock order deadlock in orphan handling (3450, 3750)
5320 - add validity checks when grabbing inodes from l_ast_data (3599)
5322 - add /proc/.../recovery_status to obdfilter (3428)
5323 - lightweight CDEBUG infrastructure, debug daemon (3668)
5324 - change default OSC RPC parameters to be better on small clusters
5325 - turn off OST read cache for files smaller than 32MB
5326 - install man pages and include them in rpms (3100)
5327 - add new init script for (un)mounting lustre filesystems (2593)
5328 - run chkconfig in %post for init scripts (3701)
5329 - drop scimac NAL (unmaintained)
5331 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5334 - clean kiobufs before and after use (3485)
5335 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5336 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5337 - change init script to fail more clearly if not run as root (1528)
5338 - allow clients to reconnect during replay (1742)
5339 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5340 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5341 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5342 - fix cond_resched (3554)
5343 - only evict unfinished clients after recovery (3515)
5344 - allow bulk resend, prevent data loss (3570)
5345 - dynamic ptlrpc request buffer allocation (2102)
5346 - don't allow unlinking open directory if it isn't empty (2904)
5347 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5348 - remove extraneous obd dereference causing LASSERT failure (3334)
5349 - don't use get_cycles() when creating temp. files on the mds (3156)
5350 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5351 - handle EEXIST for set-stripe, set proper directory name (3336)
5353 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5354 - fix ksocknal_fmb_callback() error messages (2918)
5356 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5359 - don't copy lvb into (possibly NULL) reply on error (2983)
5360 - don't deref dentry after dput, don't free lvb on error (2922)
5361 - use the kms to determine writeback rpc length (2947)
5362 - increment oti_logcookies when osc is inactive (2948)
5363 - update client's i_blocks count via lvb messages (2543)
5364 - handle intent open/close of special files properly (1557)
5365 - mount MDS with errors=remount-ro, like obdfilter (2009)
5366 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5367 - don't use cancelling-locks' kms values (2947)
5368 - use highest lock extent for kms, not last one (2925)
5369 - don't dereference ERR_PTR() dentry in error handling path (3107)
5370 - fix thread race in portals_debug_dumplog() (3122)
5371 - create lprocfs device entries at setup instead of at attach (1519)
5372 - common AST error handler, don't evict client on completion race (3145)
5373 - zero nameidata in detach_mnt in 2.6 (3118)
5374 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5375 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5376 - initialize RPC timeout timer earlier for 2.6 (3219)
5377 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5378 - print nal/nid for unknown nid (3258)
5379 - additional checks for oscc recovery before doing precreate (3284)
5380 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5381 - don't crash in mdc_close for bad permissions on open (3285)
5382 - zero i_rdev for non-device files (3147)
5383 - clear page->private before handing to FS, better assertion (3119)
5384 - tune the read pipeline (3236)
5385 - fix incorrect decref of invalidated dentry (2350)
5386 - provide read-ahead stats and refine rpc in flight stats (3328)
5387 - don't hold journal transaction open across create RPC (3313)
5388 - update atime on MDS at close time (3265)
5389 - close LDAP connection when recovering to avoid server load (3315)
5390 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5391 - don't leak open file on MDS after open resend (3325)
5392 - serialize filter_precreate and filter_destroy_precreated (3329)
5393 - loop device shouldn't call sync_dev() for nul device (3092)
5394 - clear page cache after eviction (2766)
5395 - resynchronize MDS->OST in background (2824)
5396 - refuse to mount the same filesystem twice on same mountpoint (3394)
5397 - allow llmount to create routes for mounting behind routers (3320)
5398 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5399 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5400 - fix rare problem with rename on htree directories (3417)
5402 - allow default OST striping configuration per directory (1414)
5403 - fix compilation for qswnal for 2.6 kernels (3125)
5404 - increase maximum number of MDS request buffers for large systems
5405 - change liblustreapi to be useful for external progs like lfsck (3098)
5406 - increase local configuration timeout for slow disks (3353)
5407 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5409 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5412 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5413 - don't overwrite extent policy data in reply if lock was blocked (2901)
5414 - drop filter export grants atomically with removal from device (2663)
5415 - del obd_self_export from work_list in class_disconnect_exports (2908)
5416 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5417 - swab reply message in mdc_close, other PPC fixes (2464)
5418 - fix destroying of named logs (2325)
5419 - overwrite old logs when running lconf --write_conf (2264)
5420 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5421 - fix race in target_handle_connect (2898)
5422 - mds_reint_create() should take same inode create lock (2926)
5423 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5424 - don't close files for self_export to avoid uninitialized obd (2936)
5425 - allow MDS with the same name as client node (2939)
5426 - hold dentry reference for closed log files for unlink (2325)
5427 - reserve space for all logs during transactions (2059)
5428 - don't evict page beyond end of stripe extent (2925)
5429 - don't oops on a deleted current working directory (2399)
5430 - handle hard links to targets without a parent properly (2517)
5431 - don't dereference NULL lock when racing during eviction (2867)
5432 - don't grow lock extents when lots of conflicting locks (2919)
5434 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5437 - account for cache space usage on clients to avoid data loss (974)
5438 - lfsck support in lustre kernel code (2349)
5439 - reduce journal credits needed for BRW writes (2370)
5440 - orphan handling to avoid losing space on client/server crashes
5441 - ptlrpcd can be blocked, stopping ALL progress (2477)
5442 - use lock value blocks to assist in proper KMS, faster stat (1021)
5443 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5444 - recovery for initial connections (2355)
5445 - fixes for mds_cleanup_orphans (1934)
5446 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5447 - block all file creations until orphan recovery completes (1901)
5448 - client remove rq_connection from request struct (2423)
5449 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5450 - recovery timer race (2670)
5451 - mdc_close recovey bug (2532)
5452 - ptlrpc cleanup bug (2710)
5453 - mds timeout on local locks (2588)
5454 - namespace lock held during RPCs (2431)
5455 - handle interrupted sync write properly (2503)
5456 - don't try to handle a message that hasn't been replied to (2699)
5457 - client assert failure during cleanup after abort recovery (2701)
5458 - leak mdc device after failed mount (2712)
5459 - ptlrpc_check_set allows timedout requests to complete (2714)
5460 - wait for inflight reqs when ptlrpcd finishes (2710)
5461 - make sure unregistered services are removed from the srv_list
5462 - reset bulk XID's when resending them (caught by 1138 test)
5463 - unregister_bulk after timeout
5464 - fix lconf error (2694)
5465 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5466 - readahead locks pages, leaves pending causing memory pressure (2673)
5467 - increase OST request buffers to 4096 on large machines (2729)
5468 - fix up permission of existing directories in simple_mkdir (2661)
5469 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5470 - don't assign transno to errored transactions (2742)
5471 - don't delete objects on OST if given a bogus objid from MDS (2751)
5472 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5473 - if rq_no_resend, then timeout request after recovery (2432)
5474 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5475 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5476 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5477 - MDS recovery shouldn't delete objects that it has given out (2730)
5478 - if enqueue arrives after completion, don't clobber LVB (2819)
5479 - don't unlock pages twice when trigger_group_io returns error (2814)
5480 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5481 - don't write pages to disk if there was an error (1450)
5482 - don't ping imports that have recovery disabled (2676)
5483 - take buffered bytes into account when balancing socknal conn (2817)
5484 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5485 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5486 - remove little-endian swabbing of llog records (1987)
5487 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5488 - retry reposting request buffers if they fail (1191)
5489 - grow extent at grant time to avoid granting a revoked lock (2809)
5490 - lock revoke doesn't evict page if covered by a second lock (2765)
5491 - disable VM readahead to avoid reading outside lock extents (2805)
5493 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5494 - updated kernel patches for hp-2.4.20 kernel (2681)
5496 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5499 - fix truncated write corruption (2366)
5500 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5502 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5503 - ensure len(uuid) < 37 in lmc (1171)
5504 - fix ia64 OOPS in llog_test (2255)
5505 - zero end of page at obdfilter for partial page writes (2648)
5506 - don't leave stale dentries around after renames (bug 2428)
5507 - fix timeouts when evicting a client with a single lock held (2642)
5508 - set deadline for the initial HELLO message to drain (2634)
5509 - print out dotted-quad IP addresses in the socknal (2302)
5511 - additional debugging for MDS client eviction problem (2443)
5512 - fix mkfsoptions support for osts (2603, 2604)
5514 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5517 - add series for the vanilla 2.6.0 kernel
5518 - add series for the vanilla 2.4.24 kernel
5519 - add series for a cray x86/64 UL kernel drop
5520 - fix xattr patches for the vanilla 2.4.19 series
5522 - generate true UUIDs in lmc (1171)
5523 - have portals stack dumping break in UML (2466)
5524 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5525 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5526 - wake callback waiting threads on client eviction (2460)
5527 - Add --ptldebug and --subsystem to lmc (1719)
5528 - update assertion to allow safe interrupt allocation
5529 - set rq_no_resend for cancel requests (2432)
5530 - recalculate ptlrpcd timeout after resend (2494)
5531 - call vfs_rmdir when removing pending directories (2368)
5532 - fix renaming a file to itself (2429)
5533 - lmc creates a default one-stripe lov (2454)
5534 - expand procfs space to handle large clusters (2326)
5535 - increase UML stack to avoid overflow
5536 - update lconf's list of debug and subsystem masks
5537 - fix lfs find --obd (2510)
5538 - /proc tunable for disabling filter read caching (2591)
5539 - stop rpm packages from altering slapd.conf (2301)
5540 - disable nagle in the socknal under 0conf (2578)
5541 - choose mds inode size based on stripe count (2572)
5542 - fix kernel-source rpm problems (2516)
5544 - add --disable-doc to avoid pdf generation (2421)
5545 - update documentation, tests, type-os, comments
5546 - avoid format warnings on ia64
5547 - remove the TOE NAL
5548 - tiny code cleanups by removing unused fields
5550 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5553 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5554 - remove the most blatant lies from BUILDING file (2371)
5555 - change default debug level to reasonable production setting
5556 - reduce client side cache size to reduce cache flush time
5557 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5558 - make TCP zerocopy and pinger support enabled by default (2476)
5559 - sync writes completed after process exits caused crashes (2319)
5560 - maintain correct mount count on the MDS (2356)
5561 - backout 1557, because 2316 wasn't really fixed
5562 - better file I/O statistics gathering in /proc
5563 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5564 - another kernel patch to fix zero-copy TCP function export
5565 - don't take duplicate lock when processing re-sent getattr (2420)
5566 - lctl uses obd_self_export instead of creating new conn (2353)
5567 - MDS/OST recovery case which requires object creation asserted (2425)
5568 - move lfs from /usr/sbin to /usr/bin in packages
5569 - fix race between mds_client_add and mds_client_free (2417)
5570 - use kmalloc instead of slabs in portals (2430)
5571 - don't create duplicate records when a failover MDS is present (2442)
5572 - remove unnecessary mount age check (2332)
5573 - don't remove directory inodes from locks prematurely (2451)
5574 - don't break if MDS service name is the same as hostname (2103)
5575 - fix races in client write RPC generation when cache full (2482)
5577 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5580 - remove now-unused request->rq_obd (278)
5581 - if an allocation fails, print out how much memory we've used (1933)
5582 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5583 - add the "configurable stack size" patch to most series files (1256)
5584 - ability to write large log records, for 100+ OST configs (2306)
5585 - fix NULL deref when filter_prep fails (2314)
5586 - fix operator precedence error in filter_sync
5587 - dynamic allocation of socknal TX descriptors (2315)
5588 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5589 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5590 - ensure that CWARN messages in Portals always get to the syslog
5591 - __init/__exit are not for prototype decls (ldlm_init/exit)
5592 - x86-64 compile warning fixes
5593 - fix gateway LMC keyword conflict (2318)
5594 - fix MDS lock inversions in getattr/reint paths (1844)
5595 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5596 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5597 - DLM race condition prevented some lock evictions (2328)
5598 - ENOMEM detection and retry on socknal sends (2230)
5599 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5600 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5601 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5602 - better error messages when a client is rejected during recovery (1505)
5603 - avoid cancelling locks which were never granted, after failure (2330)
5604 - fix i_sem/journal inversion in mds_client_add (2333)
5605 - fix truncate/getattr lock cycle deadlock (2334)
5606 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5607 - fix two rare exit paths which could leak an l_lock() ref (2321)
5608 - fencepost error in MDS/OST orphan recovery (2226)
5609 - make log record alignment 8 bytes (1988)
5610 - lstripe now fails when requested offset > ost_count (2237)
5611 - ensure that all kernel series have a complete list.h (1607)
5612 - fix crashes in special-file operations (2316)
5613 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5615 - allow configurable automake binary, for testing new versions
5616 - small update to the lfs documentation
5618 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5620 * fix negative export reference count in fsfilt_sync (2312)
5622 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5623 * release candidate 0.9.1
5625 - orphans are moved into the PENDING directory for possible recovery
5626 - replayed opens now open by fid for orphan/rename safety (1042)
5627 - last close of an orphan inode generates a transno (683)
5628 - chdir() and mount() now pin the directory entry (1020)
5629 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5630 - discard very old requests without processing them (1502)
5631 - remove some common, well-understood CERRORs (1505)
5632 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5633 - clear "grant" flags in OST replies until OST grant code lands (1644)
5634 - fix read performance by not clobbering i_blksize on client (1598)
5635 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5636 - make LRU size a /proc tunable, clears locks when reduced (707)
5637 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5638 - server locks take references on exports now (1558)
5639 - build fixes for 2.4.20-rh trees (1663)
5640 - return an error from lov_create if all OSCs are inactive (1751)
5641 - fix import levels when a reconnect happens without a timeout (1597)
5642 - exit early from mds_open if we get a lookup error (1749)
5643 - partial page read at EOF wouldn't wait for disk before sending (1642)
5644 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5645 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5646 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5647 - write/truncate lock inversion (1639)
5648 - Don't auto-load obdclass, portals modules during cleanup (1495)
5649 - fix timestamps from jumping to "now" (1763)
5650 - extra journal assertions (1648)
5651 - add an extra multiunlink test (1771)
5652 - fix read_record/write_record API (1776)
5653 - fix leak of offset_extent, possible incorrect i_size later (1772)
5654 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5655 - Debugging for the kqswnal_get_idle_tx problems (1820)
5656 - Allow recovery to be attempted multiple times (1536)
5657 - Write out MDS last_rcvd file after it is first created (1600)
5658 - Fix tx_descriptor leak in failed transmit situations (1827)
5659 - ext3 journaling fixes for assertion failure after IO error (1871)
5660 - class_export_put() on freed export after completion AST error (1896)
5661 - Fix revalidate looping in VFS (1322)
5662 - Don't access a freed export during MDS_REINT timeout (1521)
5663 - Add open-unlink recovery support on the MDS (1673,1764)
5664 - Return an error if no MDS data was read from last_rcvd (1946)
5665 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5666 - Don't setup a disk device that doesn't match exported UUID (317)
5667 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5668 - avoid committing NULL handle in force close
5669 - local.sh is now a one-stripe LOV configuration
5670 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5671 - fix ext3 htree duplicate directory entry corruption (1516)
5672 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5673 - update to new LOV EA format (2097)
5674 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5676 2003-06-15 Phil Schwan <phil@clusterfs.com>
5679 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5680 - per-import/export recovery handling (958, 931, 959)
5681 - multiple last-rcvd slots, for serving multiple FSes (949)
5682 - connections are again shared between multiple imp/exports (963, 964)
5683 - "umount -f" would hang if any requests needed to be sent (393, 978)
5684 - avoid pinning large req buffer by copying for queued messages (989)
5685 - add "uuid" to "lctl device" command to help upcalls (991)
5686 - "open" RPCs with transnos would confuse recovery counters (1037)
5687 - do proper endian conversion of all wire messages (288, 340, 891)
5688 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5689 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5690 - fix ptlrpc_connection leak in target_handle_connect (1174)
5691 - fix import refcounting bug in OST and MDS cleanup (1134)
5692 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5693 - fix very unlikely obd_types race condition (501)
5694 - remove osc_open hack for echo_client (1187)
5695 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5696 - a failure in read_inode2 leads to deadlock (1139)
5697 - cancel ack-locks as soon as transaction is committed (1072)
5698 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5699 - make sure to commitrw after any preprw to avoid deadlock (1162)
5700 - failing to execute a file in a lustre FS would lock inode (1203)
5701 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5702 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5703 - obd_uuid2tgt would walk off the end of the list (1255)
5704 - on IA64 the osc would give portals incorrect bulk size (1258)
5705 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5706 - fix lock inversion caused by new llite matching code (1282)
5707 - limit the number of dirty pages on a client to 10MB (1286)
5708 - timed out locks were not being corrected cancelled (1289)
5709 - fix O_DIRECT above 4GB on IA-32 (1292)
5710 * major user-visible changes
5711 - fail out/fail over policy now controlled by the upcall (993)
5713 - add OBD_PING to check server availability and failure (954)
5714 - lustre messages are now sent in sending host order (288, 340, 891)
5715 - add eadatalen to MDS getattr reply (340)
5716 - OST read replies may contain second buffer, with per-page status (593)
5718 2003-03-11 Phil Schwan <phil@clusterfs.com>
5721 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5722 - failed open()s could cause deadlock; fixed (867, 869)
5723 - stop cancelling OST locks when files are closed (481)
5724 - overlapping XID spaces caused network corruption (851, 853)
5725 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5726 - setattr_raw added, to do single-RPC, server-side setattrs
5727 - lmc/lconf syntax change for OST UUIDs
5728 - fix crashy race condition between ptlrpc_free_req and osc_close
5729 - don't use request in mdc_enqueue if we hit a timeout (889)
5730 - don't set the inode i_size for regular files from the MDS (896)
5731 - handle out of order completion AST (842)
5732 - don't LBUG if a lock request times out after receiving AST (913)
5733 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5734 - if a bad lock AST arrives, send an error instead of dropping entirely
5735 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5736 - fix leak in bulk IO when only partially completed (899, 900, 926)
5737 - fix O_DIRECT for ia64 (55)
5738 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5739 - C-z after timeout could hang a process forever; fixed (977)
5741 - client-side I/O cache (678, 924, 929, 941, 970)
5743 - READPAGE and SETATTRs which don't take server-side locks get
5746 2003-02-11 Phil Schwan <phil@clusterfs.com>
5749 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5750 - Fix fsfilt_extN_readpage() to read a full page of directory
5751 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5752 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5753 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5754 - Most intent operations were removed, in favour of a new RPC mode
5755 that does a single RPC to the server and bypasses most of the VFS
5756 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5757 - Aggressively cancel local locks on DLM servers
5758 - mds_reint_unlink sends EA to the client if it's the last nlink.
5759 client uses that EA to unlink OST objects.
5760 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5761 - recursive symlinks were fixed (439)
5762 - fixed NULL deref in DEBUG_REQ
5763 - filter_update_lastobjid no longer calls sync, which annoyed extN
5764 - fixed multi-client small-writes to a single file problem (445)
5765 - fixed mtime updates during file writes (607)
5766 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5767 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5768 - fixed Linux OST size reporting problem (444, 656)
5769 - OST now updates object mtime with writes or setattr (607, 619)
5770 - client verifies file size before zeroing page past EOF (445)
5771 - OST now writes last allocated objid to disk with allocation (108)
5772 - LOV on echo now works (409)
5774 - mds_reint_unlink sends a new buffer, with the EA included. this
5775 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5776 set if a regular file was being unlinked, and it was the last link
5777 - use PtlGet from the target for bulk writes (315)
5778 - OST now updates object mtime with writes or setattr (607, 619)
5779 - LDLM now has a grant-time callback to revalidate locked items, if
5781 - Many MDS operations were reorganized to combat race conditions
5783 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5784 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5786 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5789 - Fully reactivate OST imports after reconnection (512, others)
5790 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5791 - More graceful error handling for truncating on dead OST (515)
5792 - Don't error out unless we're actually accessing dead stripes (474)
5793 - Fix garbage sizes when stripes are missing (410)
5794 - LRU counters were broken, causing constant lock purge (433, 432)
5795 - garbage on read from stripes with failed OSTs (441)
5796 - mark OSCs as active before reconnecting during recovery (438)
5797 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5798 - lfind did not preserve OST order in output (443)
5799 - symlinks cause hung clients, incorrect data (439)
5800 - stop dereferencing request after dropping refcount (457)
5801 - don't LASSERT(spin_is_locked) on non-SMP (455)
5802 - fixes for many rename() bugs
5803 - fstat didn't correctly synchronize attributes (399)
5804 - server must handle lock cancellation during blocking AST prep (487)
5805 - bulk descriptors were free()d too soon (511)
5806 - fix paths in lconf, which would load incorrect modules (451, 507)
5807 - fix confusing lconf 'host not found' error message (386)
5808 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5809 - fix race condition in mdc_blocking_ast() for inode access (526)
5810 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5811 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5812 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5813 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5814 - fix lstripe in several ways (kernel side) (527)
5815 - fix request leak in ldlm_cli_enqueue (262)
5816 - incorrect OSC was marked inactive after OST failure
5817 - call mds_fs_cleanup before unmounting filesystem (524)
5818 - fix races between taking ns_lock and ldlm_lock_change_resource
5819 - fix races updating LOV export open file list
5820 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5821 - fix recovery NULL deref in ldlm_cli_cancel_unused
5822 - fix some DLM races by using new hash table for lock handles (419)
5823 - permit the client to specify desired inodes, at replay
5824 - duplicate requests when we queue them for replay reintegration
5825 - fix last_rcvd offset calculation
5826 - sync after each recovered transaction, so we always make progress
5827 - never, not always, ERESTART requests without transnos
5828 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5830 - skip replay if the MDS says that the client is already connected
5831 - don't check for a recovery-enabled export to match lctl's UUID
5832 - don't INC_USE_COUNT for phantom exports
5833 - don't crash when cleaning up phantom exports (567)
5834 - don't double-finish or set replay data for errored mdc_open requests
5835 - abort requests when they time out, so we don't get old replies
5836 - send/receive replies for AST messages again
5837 - if the client says that it doesn't have the lock, cancel it on the
5839 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5840 mark it as destroyed, it will all work out when decref is called
5841 - fix module use counts (22, 581)
5843 - ASTs now expect a reply (server cancels lock on error reply)
5845 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5848 - fix many simultaneous client startup (392)
5849 - fix dentry->d_it clobbering
5850 - credentials weren't being shipped for readdir/getattr operations
5851 - remove invalid assertions triggered during some concurrent MD
5853 - proper Lustre versions added (336, 389)
5854 - fix memory leak for create error case (398)
5855 - fix LOV locking bug that would get cli/srv out of sync
5856 - fix echo client over LOV (409)
5857 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5858 - fix double-O_EXCL intent crash (424)
5859 - avoid sending multiple lock CANCELs (352)
5861 - MDS can do multi-client recovery (modulo bugs in new code)
5863 - many updates, edits, cleanups
5865 2002-11-18 Phil Schwan <phil@clusterfs.com>
5868 - fix null d_it dereference (346)
5869 - fix full OST/dbench hang (333)
5870 - fix permission problem with file removal (286)
5871 - fix removal of OSCs from LOV when they fail
5872 - fix NULL deref during bulk timeout (214)
5873 - fix problems related to multiple filesystems on one MDS (241)
5874 - fixed serious subtle metadata locking bugs
5875 - free locks on clients when inodes are removed due to memory
5877 - fix inode pointer in lock data (285)
5878 - partial support for multiple MDS on a single host (241)
5879 - data locks weren't cancelled at clear_inode time (290, 311)
5880 - intent locks could lead to unbounded lock growth (205)
5881 - added a maximum lock count, an LRU list, and a flusher
5882 - fix multiple rename (365)
5883 - properly abstracted the echo client
5884 - OSC locked 1 byte too many; fixed
5885 - rewrote brw callback code:
5886 - fixed recovery bugs related to LOVs (306)
5887 - fixed too-many-pages-in-one-write crash (191)
5888 - fixed (again) crash in sync_io_timeout (214)
5889 - probably fixed callback-related race (385)
5891 - Add capability to MDS protocol
5892 - LDLM cancellations and callbacks on different portals
5894 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5897 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5898 - defer open object destruction to close time (601981, 138)
5899 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5900 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5901 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5902 - don't call obd_setattr in ll_file_release for destroyed objects
5903 * protocol change to lustre_msg: move |version| and add |flags|
5904 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5905 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5906 * added replay of create, unlink, link and rename operations during
5907 MDS failover; recovery should be much more robust now
5908 * remove failed OSCs from LOVs (only lov_create uses this so far)
5909 * the lustre-HOWTO was brought (more) up to date (582544)
5911 2002-10-23 Phil Schwan <phil@clusterfs.com>
5914 - in-use dentries weren't being reused properly (617851)
5915 - prevent multiple LDLM setup (599178)
5916 - fix LOV size calculations for truncate (617853)
5917 - fix client handling of MDS intent errors (POSIX)
5918 - fix permission bug in lovstripe.c test (624321)
5919 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5920 - truncate past end of file could corrupt data
5921 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5922 - a race in recovery could return ETIMEDOUT to apps (623947)
5923 - building outside the source directory was fixed
5924 * the lustre-HOWTO was brought (more) up to date (582544)
5925 * major progress was made on recovery functionality
5927 2002-10-10 Phil Schwan <phil@clusterfs.com>
5930 - recovery deadlock fix
5931 - rm -rf causes LBUG fix (617817)
5932 - file open by multiple tasks fix (618962)
5933 - directory permissions bugs (602707 and 620007)
5934 - journal_stop fixed with locking (611313)
5935 - O_APPEND failures resolved (618273, perhaps 614459)
5936 - lconf PATH fix (619770)
5937 - IA64 build fix (621450)
5938 - RPC buffer sizes scale with amount of memory
5940 2002-10-01 Phil Schwan <phil@clusterfs.com>
5943 - locks would be cancelled without throwing away data pages,
5944 resulting in inconsistent data (605627)
5945 - inode attributes were not always being refreshed (605627, 612449)
5946 - lconf now continues to cleanup after lctl reports an error
5947 - MDS now enforces user permissions (602707)
5948 - lprocfs cleanup fixed, but not yet enabled (614157)
5949 - fixed infinite server hang, should a client not respond to an AST
5950 - avoid going into recovery if user calls readlink() with a buffer
5951 that's too small (613941)
5952 - AST RPCs no longer require replies (614867) -- this may be changed
5953 - don't crash server if client sends an IOV that's too big (611336)
5954 - fixed lock conversion deadlock (611892)
5955 - fixed the following of symlinks (614622)
5956 * recovery: the server can remove locks from a client that dies, other
5957 clients can make progress
5958 * more extN patch fixes
5959 * compile-time configurable ptlrpc buffer allocations
5961 - collaborative read cache document
5962 - Lustre Lite Performance CDR document-in-progress
5964 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5967 - fix typo in patch-2.4.18
5969 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5972 - clear ptlrpc request each time in handle_incoming_request()
5973 - unlink of files now destroys the object on the OST
5975 2002-09-19 Peter Braam <braam@clusterfs.com>
5977 * add hard link support
5978 * change obdfile creation method
5979 * kernel patch changed
5981 2002-09-19 Peter Braam <braam@clusterfs.com>
5984 - stack overflow bug in extN fixed
5986 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5988 * documentation updates
5989 - add man pages for config tools
5990 - update tests/README to describe testing with new config tools
5991 - finish metadata API descriptions
5992 * bug fixes and cleanups
5993 - statfs workaround for 16TB limit
5994 - LOV stripe allocation improved, can stripe on subset of OSTs
5995 - LOV file size/IO offset was wrong for files > 4GB in size
5996 - object EA data was being dropped, caused files to be unreadable
5997 - memory overflow with non-LOV OST caused memory corruption
5998 - fixed regression tests to work with new config tools, obdfilter
5999 - fixed bug when directory size became larger than 1 block
6000 - fixed bug (for single client case) when PWD was deleted
6001 - invalidate local directory pages when doing intent-based ops
6002 - avoid LDLM oops when lock callback contained bad data
6004 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6006 * documentation updates
6007 * bug fixes and cleanups
6008 - configuration tools
6011 - 64-bit compile warnings
6012 - 64-bit internal statfs data
6014 * test_brw on persistent OST devices
6016 * lprocfs (disabled)
6018 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6020 * documentation updates
6021 * bug fixes and cleanups
6022 * configuration tools
6024 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6027 * many small fixes to 0_5_4
6028 * io/network handling
6029 * thinkos in MDS operations
6031 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6034 * crucial basic fixes to 0.5.3
6035 * IOR, Iozone work over Elan
6038 2002-08-07 Phil Schwan <phil@clusterfs.com>
6039 * version 0_5_3, our first alpha
6040 * we use the new Portals iovs
6041 * documentation updates
6042 * bug fixes and cleanups
6043 * small changes in the DLM wire protocol
6045 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6046 * version 0_5_1 with some initial stability,
6047 * locking on MD and file I/O.
6048 * documentation updates
6049 * several bug fixes since 0.5.0
6050 * small changes in wire protocol
6052 2002-07-18 Phil Schwan <phil@clusterfs.com>
6054 * delivered as Lustre Light Alpha
6055 * fixed a crash after handling invalid MDS requests
6056 * fixed directory pages for architectures with non-4k pages sizes
6058 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6059 * release version v0_4_4
6060 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6062 * Fixes a number of interruption problems with OST operations.
6063 * Update documentation for portals header changes
6064 * Move all wire protocol structs/defines to lustre_idl.h
6065 * Fixes symlink length bug.
6066 * Add tcpdump to repository.
6068 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6069 * release version v0_4_3
6070 * Fixes statfs for inodes on extN.
6071 * Fixes bug in runtests which would delete /etc/hosts.
6072 * Use 64-bit object IDs wherever possible (not into VFS though)
6073 Remove ost_get_info, which is unused by lustre, and out of date.
6075 2002-07-03 Peter Braam <braam@clusterfs.com>
6076 * release version v0_4_2 Fixes a lookup error (type not passed)
6077 * move forward to head of Portals
6078 * move forward to latest Lustre kernel
6080 2002-06-25 Peter Braam <braam@clusterfs.com>
6081 * release version v0_4_1. Hopefully stable on single node use.