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),
7 2.6.22.14 vanilla (kernel.org).
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.21 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.6.sun1
12 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
13 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
14 removed cwd "./" (refer to Bugzilla 14399).
15 * File join has been disabled in this release, refer to Bugzilla 16929.
19 Description: Fix in ptlrpc_expire_one_request() to print the signed time
22 Severity : enhancement
24 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
27 Severity : enhancement
29 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
31 Severity : enhancement
33 Description: correctly handle big reply message.
34 Details : send LNet event if reply is bigger then buffer and adjust this buffer
39 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
41 Severity : enhancement
43 Description: Add support for OEL5.
47 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
51 Description: client could not reconnect to OST because of an active request.
52 Details : The client now retries to reconnect to the same server, if a
53 connect request failed with EBUSY or -EAGAIN. Test case 35b
54 conf-sanity.sh has been created.
58 Description: Don't allow make backward step on assignin osc next id.
59 Details : race between allocation next id and ll_sync thread can be cause
60 of set wrong osc next id and can be kill valid ost objects.
64 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
68 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
72 Description: Error handling in osc_statfs_interpret() has been improved.
73 Details : Check in osc_statfs_interpret() for EBADR.
77 Description: Do not send statfs() requests to OSTs disabled by administrator.
78 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
80 Severity : enhancement
82 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
84 Severity : enhancement
86 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
88 Severity : enhancement
90 Description: Enhancement for lfs(1) command to use numeric uid/gid.
94 Description: Improved cleanup procedure for conf-sanity test_46a.
98 Description: Do not start bulk transfer, if client tried to reconnect.
99 Details : Do not start bulk transfer, if client tried to reconnect,
100 but failed due to active rpcs. Check for exp_abort_active_req
101 has been implemented.
105 Description: Prevent inconsistences between linux and lustre mount structures.
106 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
107 Make the sleep interruptible.
109 Severity : enhancement
111 Description: provide server to client comms path
112 Details : server to client communications path via new LDLM_SET_INFO rpc
116 Description: truncate starts GFP_FS allocation under transaction causing deadlock
117 Details : ldiskfs_truncate calls grab_cache_page which may start page
118 allocation under an open transaction. This may lead to
119 calling prune_icache with consequent lustre reentrance.
123 Description: send correctly lsm on open replay
124 Details : MDS is trust to LSM size on replay open, but client can set wrong size
129 Description: enable client interface failover
130 Details : When a child reconnects from another NID, properly update export
131 nid hash position and ldlm reverse import.
133 Severity : enhancement
135 Description: Communicate OST degraded/readonly state via statfs to MDS
136 Details : Flags in the statfs returned from OSTs indicate whether the
137 OST is in a degraded RAID state, or if the filesystem has
138 turned read-only after a filesystem error is detected.
143 Description: don't panic if EPROTO was hit when reading symlink
144 Details : correctly handling request reference in error cases.
146 Severity : enhancement
148 Description: Add LustreNetLink, a kernel-userspace communcation path.
150 Severity : enhancement
152 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
157 Description: access to llog context before init.
158 Details : move handling CATALOGS file at osc layer and forbid access to llog
162 Frequency : rare, only when client is writing to OST on local node
164 Description: Avoid deadlock for local client writes
165 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
166 memory freeing context. This allows OST threads to set the
167 PF_MEMALLOC flag on task structures in order to allocate memory
168 from reserved pools and complete IO.
169 Use GFP_HIGHUSER for OST allocations for non-local client writes,
170 so that the OST threads generate memory pressure and allow
171 inactive pages to be reclaimed.
173 Severity : enhancement
175 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
180 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
181 Details : move ldlm namespace creation in setup phase to avoid grab
182 _lprocfs_lock with cli_sem held.
186 Description: Temporarily disable grant shrink.
187 Details : Disable the feature for debugging.
190 Frequency : only when formatting test filesystems
192 Description: Unable to run several mkfs.lustre on loop devices at the same time
193 Details : mkfs.lustre returns error 256 on the concurrent loop devices
194 formatting. The solution is to proper handle the error.
196 Severity : enhancement
198 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
200 Severity : enhancement
202 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
207 Description: resolve race between obd_disconnect and class_disconnect_exports
208 Details : if obd_disconnect will be called to already disconnected export he
209 forget release one reference and osc module can't unloaded.
211 Severity : enhancement
213 Description: Allow tuning service thread via /proc
214 Details : For each service a new
215 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
216 created that can be used to set min/max thread counts, and get the
217 current number of running threads.
222 Descriptoin: don't return error if have partially created objects for file.
223 Details : lov_update_create_set uses set->set_success as index for created
224 objects, so if some requests will be failed, they will have hole
225 at end of array and qos_shrink_lsm for allocate correct lsm.
227 Severity : enhancement
229 Description: Update OFED support to 1.4.1
234 Description: Reduce small size read RPC
235 Details : Set read-ahead limit for every file and only do read-ahead when
236 available read-ahead pages are bigger than 1M to avoid small size
239 Severity : enhancement
241 Description: move AT tunable parameters for more consistent usage
242 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
244 Severity : enhancement
246 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
247 Details : allow skip disconnected ost for send statfs request and hide error
253 Description: fix racy locking of mballoc block bitmaps causing BUG
254 Details : The locking of the mballoc buddy bitmap and the in-memory
255 block bitmap was using two different spin locks in some
256 cases. This made it possible to incorrectly access the
257 mballoc bitmap while another process was modifying it,
258 causing a sanity assertion to fail. While no on-disk corruption
259 was reported, there was some risk of this happening.
262 Frequency : rare, on llog test 6
264 Descriptoin: don't allow connect to already connected import
265 Details : allowing connect to already connected import is hide connecting problem.
270 Descriptoin: allow kill process which wait statahead result
271 Details : for some reasons 'ls' can stick in waiting result from statahead,
272 in this case need way for kill this process.
274 Severity : enhancement
276 Description: Add state history info file, enhance import info file
277 Details : Track import connection state changes in a new osc/mdc proc file;
278 add overview-type data to the osc/mdc import proc file.
280 Severity : enhancement
282 Description: MDS create should not wait for statfs RPC while holding DLM lock.
284 Severity : enhancement
286 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
291 Descriptoin: OST grant too much space to client even there are not enough space.
292 Details : Client will shrink its grant cache to OST if there are no write
293 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
294 this grant cache if there are already not enough avaible space
295 (left_space < total_clients * 32M).
298 Frequency : start MDS on uncleanly shutdowned MDS device
300 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
301 Details : stay in waiting mds<>ost recovery finished produce random bugs
302 due race between two ll_sync thread for one lov target. send
303 ACTIVATE event only if connect realy finished and import have
307 Frequency : rare, connect and disconnect target at same time
309 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
310 Details : don't call obd_disconnect under lov_lock. this long time
311 operation and can block ptlrpcd which answer to connect request.
316 Descriptoin: don't lose wakeup for imp_recovery_waitq
317 Details : recover_import_no_retry or invalidate_import and import_close can
318 both sleep on imp_recovery_waitq, but we was send only one wakeup
322 Frequency : always with long access acl
324 Descriptoin: mds can't pack reply with long acl.
325 Details : mds don't control size of acl but they limited by reint/getattr
328 Severity : enhancement
330 Description: Update to SLES10 kernel-2.6.16.60-0.33.
332 Severity : enhancement
334 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
337 Frequency : start MDS on uncleanly shutdowned MDS device
339 Descriptoin: aborting recovery hang on MDS
340 Details : don't throttle destroy RPCs for the MDT.
343 Frequency : on remount
345 Description: external journal device not working after the remount
346 Details : clear dev_rdonly flag for external journal devices in
352 Description: shutdown vs evict race
353 Details : client_disconnect_export vs connect request race.
354 if client will evicted at this time - we start invalidate
355 thread without referece to import and import can be freed
359 Frequency : rare, need acl's on inode.
361 Description: client can't handle ost additional correctly
362 Details : if ost was added after client connected to mds client can have
363 hit lnet_try_match_md ... to big messages to wide striped files.
364 in this case need teach client to handle config events about add
365 lov target and update client max ea size at that event.
367 Severity : enhancement
369 Description: Changelogs
370 Details : Changelogs are a lightweight mechanism to track filesystem
371 metadata and namespace changes. The changelog is recorded
372 permanently on the MDTs, and is periodically "consumed" / purged
373 when records are no longer needed.
375 Severity : enhancement
377 Description: compact fld format with extents
378 Details : Store range of seq rather than every seq in FLD. Seq
379 controller update FLD rather than clients. In Case of CMD, mdt0
380 has FLD, all other metadata server act as non persistent proxy
381 for FLD queries and cache fld entries in fld cache.
386 Description: don't skip ost target if they assigned to file
387 Details : Drop slow OSCs if we can, but not for requested start idx.
388 This means "if OSC is slow and it is not the requested
389 start OST, then it can be skipped, otherwise skip it only
390 if it is inactive/recovering/out-of-space.
394 Description: more cleanup in mds_lov
395 Details : not send LOV EA under replay, we can't know about they size at this
396 time. Don't allow client connect to mds before any ost connected,
397 for avoid problems with LOV EA size and returning EIO to client.
399 Severity : enhancement
401 Description: Interoperability at server side (Disk interoperability)
403 Severity : enhancement
405 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
407 Severity : enhancement
409 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
411 Severity : enhancement
413 Description: New client IO stack (CLIO).
415 Severity : enhancement
417 Description: Commit on sharing. Eliminate inter-client dependencies between
418 uncommitted transactions by doing transaction commits.
419 Thereby clients may recovery independently.
422 Frequency : Create a symlink file with a very long name
424 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
425 Details : If there is no extra space in the request for early cancels,
426 ldlm_req_handles_avail() returns 0 instead of a negative value.
428 Severity : enhancement
430 Description: Add /proc entry for import status
431 Details : The mdc, osc, and mgc import directories now have
432 an import directory that contains useful import data for debugging
435 Severity : enhancement
437 Description: Re-disable certain /proc logging
438 Details : Enable and disable client's offset_stats, extents_stats and
439 extents_stats_per_process stats logging on the fly.
442 Frequency : Only on FC kernels 2.6.22+
444 Description: oops in statahead
445 Details : Do not drop reference count for the dentry from VFS when lookup,
446 VFS will do that by itself.
448 Severity : enhancement
450 Description: Generic /proc file permissions
451 Details : Set /Proc file permissions in a more generic way to enable non-
452 root users operate on some /proc files.
456 Description: Hitting mdc_commit_close() ASSERTION
457 Details : Properly handle request reference release in
458 ll_release_openhandle().
462 Description: quota recovery deadlock during mds failover
463 Details : This patch includes att18982, att18236, att18237 in bz14840.
465 1. fix osts hang when mds does failover with quotaon
466 2. prevent watchdog storm when osts threads wait for the
471 Frequency : only patchless client
472 Description: add workaround for race between add/remove dentry from hash
474 Severity : enhancement
476 Description: Allow OST glimpses to return PW locks
480 Description: LBUG when llog conf file is full
481 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
485 Description: Prevent import from entering FULL state when server in recovery
489 Description: service mount cannot take device name with ":"
490 Details : Only when device name contains ":/" will mount treat it as
496 Description: replace ptlrpcd with the statahead thread to interpret the async
497 statahead RPC callback
501 Frequency : on recovery
502 Description: I/O failures after umount during fail back
503 Details : if client reconnected to restarted server we need join to recovery
504 instead of find server handler is changed and process self
505 eviction with cancel all locks.
507 Severity : enhancement
509 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
511 Severity : enhancement
513 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
515 Severity : enhancement
517 Description: Upcall on Lustre log has been dumped
518 Details : Allow for a user mode script to be called once a Lustre log has
519 been dumped. It passes the filename of the dumped log to the
520 script, the location of the script can be specified via
521 /proc/sys/lnet/debug_log_upcall.
526 Description: avoid idr_remove called for id which is not allocated.
527 Details : Move assigment s_dev for clustered nfs to end of initialization,
528 to avoid problem with error handling.
533 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
534 Details : When connection is reused this not moved from CONN_UNUSED_HASH
535 into CONN_USED_HASH and this prodice warning when put connection
536 again in unused hash.
539 Severity : enhancement
541 Description: File striping can now be set to use an arbitrary pool of OSTs.
543 Severity : enhancement
545 Description: Export bytes_read/bytes_write count on OSC/OST.
549 Description: Early reply size mismatch, MGC loses connection
550 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
551 the connect flags are properly negotiated.
556 Description: quotas are not honored with O_DIRECT
557 Details : all writes with the flag O_DIRECT will use grants which leads to
558 this problem. Now using OBD_BRW_SYNC to guard this.
562 Description: add quota statistics
563 Details : 1. sort out quota proc entries and proc code.
564 2. add quota statistics
566 Severity : enhancement
568 Description: enable quota support for HEAD.
572 Description: Properly propagate oinfo flags from lov to osc for statfs
573 Details : restore missing copy oi_flags to lov requests.
575 Severity : enhancement
577 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
578 : l_getgroups(8), lst(8), routerstat(8)
579 Details : included man pages for llobdstat(8), llstat(8),
580 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
582 Severity : enhancement
584 Description: configure's --enable-quota should check the
585 : kernel .config for CONFIG_QUOTA
586 Details : configure is terminated if --enable-quota is passed but
587 : no quota support is in kernel
591 Description: Remove portals compatibility
592 Details : Remove portals compatibility, not interoperable with releases
597 Description: Resolve device initialization race
598 Details : Prevent proc handler from accessing devices added to the
599 obd_devs array but yet be intialized.
601 Severity : enhancement
603 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
605 Severity : enhancement
607 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
612 Description: Using wrong pointer in osc_brw_prep_request
613 Details : Access to array[-1] can produce panic if kernel compiled with
614 CONFIG_PAGE_ALLOC enabled
618 Description: Client runs out of low memory
619 Details : Consider only lowmem when counting initial number of llap pages
623 Description: *optional* service tags registration
624 Details : if the "service tags" package is installed on a Lustre node
625 When the filesystem is mounted, a local-node service tag will
626 be created. See http://inventory.sun.com/ for more information
627 about the Service Tags asset management system.
631 Description: Kernel BUG tries to release flock
632 Details : Lustre does not destroy flock lock before last reference goes
633 away. So always drop flock locks when client is evicted and
634 perform unlock regardless of successfulness of speaking to MDS.
638 Description: add refcount protection for osc callbacks, avoid panic on shutdown
642 Description: sanity test 65a fails if stripecount of -1 is set
643 Details : handle -1 striping on filesystem in ll_dirstripe_verify
648 Description: ASSERTION(CheckWriteback(page,cmd)) failed
649 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
652 Severity : enhancement
654 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
658 Description: do not process already freed flock
659 Details : flock can possibly be freed by another thread before it reaches
660 to ldlm_flock_completion_ast.
664 Description: LBUG during stress test
665 Details : Need properly lock accesses the flock deadlock detection list.
669 Description: oops in page fault handler
670 Details : kernel page fault handler can return two special 'pages' in error
671 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
675 Description: timeout with invalidate import.
676 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
677 should be handled by ptlrpcd. This produce long age waiting and
678 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
680 Severity : enhancement
682 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
686 Description: enable MGS and MDT services start separately
687 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
688 a co-located MGS without starting the MGS, which is a complement
689 to 'nosvc' mount option.
693 Frequency : after recovery
694 Description: precreate to many object's after del orphan.
695 Details : del orphan st in oscc last_id == next_id and this triger growing
696 count of precreated objects. Set flag LOW to skip increase count
697 of precreated objects.
701 Frequency : rare, on clear nid stats
702 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
703 Details : when clean nid stats sometimes try destroy live entry,
704 and this produce panic in free.
708 Description: Stack overflow during MDS log replay
709 ease stack pressure by using a thread dealing llog_process.
713 Description: wait until IO finished before start new when do lock cancel.
714 Details : VM protocol want old IO finished before start new, in this case
715 need wait until PG_writeback is cleared until check dirty flag and
716 call writepages in lock cancel callback.
718 Severity : enhancement
720 Description: using special macro for print time and cleanup in includes.
724 Description: mds_mfd_close() ASSERTION(rc == 0)
725 Details : In mds_mfd_close(), we need protect inode's writecount change
726 within its orphan write semaphore to prevent possible races.
730 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
734 Frequency : rare, on shutdown ost
735 Description: don't hit live lock with umount ost.
736 Details : shrink_dcache_parent can be in long loop with destroy dentries,
737 use shrink_dcache_sb instead.
741 Description: don't panic with use echo client
742 Details : echo client pass NULL as client nid pointer and this produce null
747 Description: fix build on ppc32
748 Details : compile code with -m64 flag produce wrong object file for ppc32.
752 Description: add message levels for liblustreapi
756 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
757 Details : Move the 'good_osts' check before the 'total_bavail' check. This
758 will result in an -EAGAIN and in the exit call path we call
759 alloc_rr() which will with increasing aggressiveness attempt to
760 aquire precreated objects on the minimum number of required OSCs.
764 Description: Use old size assignment to avoid deadlock
765 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
766 the scheduling while holding a spinlock. We do not need locking
767 for size in ll_update_inode() because size is only updated from
768 the MDS for directories or files without objects, so there is no
769 other place to do the update, and concurrent access to such inodes
770 are protected by the inode lock.
774 Description: resolve "_IOWR redefined" build error on SLES10
778 Description: dump the memory debugging after all modules are unloaded to
779 suppress false negative in conf_sanity test 39
781 Severity : enhancement
783 Description: build kernel-ib packages for OFED 1.3 in our release cycle
788 Description: fix SLES kernel versioning
789 Details : the kernel version for our SLES 10 kernel did not include a "-"
790 before the "smp" at the end. while this was not a problem in
791 general, it did mean that software trying to use the kernel
792 version to try to detect a vendor specific kernel would fail.
793 this was most evident by the OFED build scripts.
797 Description: Don't update lov_desc members until making sure they are valid
798 Details : When updating lov_desc members via proc fs, need fix their
799 validities before doing the real update.
803 Description: don't put request into delay list while invalidate in flight.
804 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
805 import in flight. this produce timeout for invalidate and sometimes
806 can cause stale data.
810 Frequency : on ppc only
811 Description: not convert ost objects for directory because it's not exist.
812 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
816 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
818 Severity : enhancement
820 Description: Update to support 2.6.22.14 vanilla kernel.
824 Frequency : rare, on recovery
825 Description: read procfs can produce deadlock in some situation
826 Details : Holding lprocfs lock which send rpc can produce block for destroy
827 obd objects and this also block reconnect with -EALREADY. This
828 isn't fix all lprocfs bugs - but make it rare.
830 Severity : enhancement
832 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
835 Frequency : frequent on X2 node
837 Description: mdc_set_open_replay_data LBUG
838 Details : Set replay data for requests that are eligible for replay.
842 Description: lustre_mgs: operation 101 on unconnected MGS
843 Details : When MGC is disconnected from MGS long enough, MGS will evict the
844 MGC, and late on MGC cannot successfully connect to MGS and a lot
845 of the error messages complaining that MGS is not connected.
849 Frequency : on network error
850 Description: panic with double free request if network error
851 Details : mdc_finish_enqueue is finish request if any network error ocuring,
852 but it's true only for synchronus enqueue, for async enqueue
853 (via ptlrpcd) this incorrect and ptlrpcd want finish request
856 Severity : enhancement
858 Description: client-side metadata stat-ahead during readdir(directory readahead)
859 Details : perform client-side metadata stat-ahead when the client detects
860 readdir and sequential stat of dir entries therein
863 Frequency : on start mds
865 Description: Implement get_info(last_id) in obdfilter.
868 Frequency : occasional
870 Description: Correctly check stale fid, not start epoch if ost not support SOM
871 Details : open with flag O_CREATE need set old fid in op_fid3 because
872 op_fid2 was overwritten with new generated fid, but mds can answer
873 with one of these two fids and both is not stale. Setattr
874 incorrectly started an epoch and assume will be called
875 done_writing, but without SOM done_writing ever being called.
878 Frequency : rare, depends on device drivers and load
880 Description: MDS or OSS nodes crash due to stack overflow
881 Details : Code changes in 1.8.0 increased the stack usage of some functions.
882 In some cases, in conjunction with device drivers that use a lot
883 of stack the MDS (or possibly OSS) service threads could overflow
884 the stack. One change which was identified to consume additional
885 stack has been reworked to avoid the extra stack usage.
888 Frequency : occasional
890 Description: Do not fail import if osc_interpret_create gets -EAGAIN
891 Details : If osc_interpret_create got -EAGAIN it immediately exits and
892 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
893 oscc_has_objects and see OSC has no objests and call
894 oscc_internal_create to resend create request.
896 Severity : enhancement
898 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
900 Severity : enhancement
902 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
905 Frequency : very rare
907 Description: Fix replay if there is an un-replied request and open
908 Details : In some cases, older replay request will revert the
909 mcd->mcd_last_xid on MDS which is used to record the client's
912 Severity : enhancement
914 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
916 Severity : enhancement
918 Description: Add rhel5 support to HEAD.
920 Serverity : enhancement
922 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
927 Description: Don't allow skipping OSTs if index has been specified.
928 Details : Don't allow skipping OSTs if index has been specified, make
929 locking in internal create lots better.
933 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
934 Details : access to bitfield in structure is always rounded to long
935 and this produce problem with not atomic change any bit.
939 Description: Lustre make rpms failed.
940 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
941 builds Lustre rpms from tarball.
944 Frequency : rare on shutdown ost
946 Description: If llog cancel was not send before clean_exports phase, this can
947 produce deadlock in llog code.
948 Details : If llog thread has last reference to obd and call class_import_put
949 this produce deadlock because llog_cleanup_commit_master wait when
950 last llog_commit_thread exited, but this never success because was
951 called from llog_commit_thread.
955 Description: allow userland application know is lost one of stripes.
956 Details : fill lvb_blocks with error code on ost and return it to
957 application if error flag found.
961 Description: NULL lov_tgts causing MDS oops
962 Details : more safe checks for NULL lov_tgts for avoid oops.
964 Severity : enhancement
966 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
970 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
971 Details : If an OST is down the MDS will hang indefinitely in
972 obd_statfs_rqset() waiting for the statfs data. While for
973 MDS QOS usage of statfs, it should not stuck in waiting.
975 Severity : enhancement
977 Description: remote_acl support
978 Details : Support ACL-based permission check for remote user.
979 Support setfacl/getfacl for remote user with the utils
980 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
981 the system "{s,g}etfacl" utils.
983 Severity : enhancement
985 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
987 Severity : enhancement
989 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
993 Description: make mgs_setparam() handle fsname containing dash
994 Details : fsname containing a dash does not work with lctl conf_param
996 Severity : enhancement
998 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1000 Severity : enhancement
1002 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1006 Frequency : rare, at shutdown
1007 Description: access already free / zero obd_namespace.
1008 Details : if client_disconnect_export was called without force flag set,
1009 and exist connect request in flight, this can produce access to
1010 NULL pointer (or already free pointer) when connect_interpret
1011 store ocd flags in obd_namespace.
1015 Frequency : only at startup
1016 Description: not alloc memory with spinlock held.
1017 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1018 if any spinlock held.
1020 Severity : enhancement
1022 Description: make lustre randomly fail allocating memory
1023 Details : Make lustre randomly failed allocating memory for testing purpose.
1025 Severity : enhancement
1027 Description: lost problems with lov objid file
1028 Details : Fixes some scability and access to not inited memory problems
1029 in work with lov objdid file.
1034 Description: lfs find does not continue on file error
1035 Details : Continue other files processing when a file/dir is absent.
1039 Description: Inconsistent usage of lustre_pack_reply()
1040 Details : Standardize the usage of lustre_pack_reply() such that it
1041 always generate a CERROR on failure.
1044 Frequency : occasional
1046 Description: MDS hang or stay in waiting lock
1047 Details : If client receive lock with CBPENDING flag ldlm need send lock
1048 cancel as separate rpc, to avoid situation when cancel request
1049 can't processed due all i/o threads stay in wait lock.
1053 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1057 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1061 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1065 Description: Update to RHEL4 latest kernel.
1067 Severity : enhancement
1069 Description: Build SLES10 patchless client fails
1070 Details : The configure was broken by run ./configure with
1071 --with-linux-obj=.... argument for patchless client. When the
1072 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1073 can't search header adequately. Use absolute path such as
1074 -I($LINUX)/include instead.
1078 Description: interrupt oig_wait produce painc on resend.
1079 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1080 private set, and this produce situation when rq_ptlrpcd_data not
1081 copyed to new allocated request and triggered LBUG on assert
1082 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1083 ptlrpcd set if request is changed and can be safety changed to use
1088 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1089 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1090 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1091 these macros and replace them with their difinitions instead.
1095 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1096 Details : Update which_patch & target file for SLES10 latest kernel.
1099 Description: Remove client patches from SLES 10 kernel.
1100 Details : This causes SLES 10 clients to behave as patchless clients
1101 even on a Lustre-patched (server) kernel.
1103 Severity : enhancement
1105 Description: self-adjustable client's lru lists
1106 Details : use adaptive algorithm for managing client cached locks lru
1107 lists according to current server load, other client's work
1108 pattern, memory activities, etc. Both, server and client
1109 side namespaces provide number of proc tunables for controlling
1112 Severity : enhancement
1114 Description: light-weight GSS support
1115 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1116 and reduce performance overhead.
1118 Severity : enhancement
1120 Description: Linux keyring support
1121 Details : Support using service of Linux keyring for Lustre GSS internal
1122 context refresh/cache mechanism.
1126 Description: Fix errors in lfs documentation
1127 Details : Fixes man pages
1131 Description: don't use GFP_* in generic Lustre code.
1132 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1136 Description: obdclass is limited by single OBD_ALLOC(idarray)
1137 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1141 Description: warnings with build patchless client with vanila 2.6.19 and up
1142 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1143 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1147 Frequency : Only for SLES
1148 Description: sanity_quota fail test_1
1149 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1150 file, which makes TSTID[2] inunique.
1154 Frequency : Only for SLES9
1155 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1156 Details : Update target/ChangeLog/which_patch .
1160 Frequency : Only for test_10 in sanity-quota.sh
1161 Description: a bug in quota test code
1162 Details : checking a test flag in a wrong place
1166 Frequency : Only for test_13 in sanity-quota.sh
1167 Description: a bug in quota test script
1168 Details : shouldn't check overfull allocated quota
1172 Description: jbd statistics
1173 Details : Port older jbd statistics patch for sles10
1175 Severity : enhancement
1177 Description: Add printing inode info into message about error in writepage.
1182 Description: reply_lock_interpret crash due to race with it and lock cancel.
1183 Details : Do not replay locks that are being cancelled. Do not reference
1184 locks by their address during replay, just by their handle.
1188 Frequency : When flocks are used.
1189 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1190 Details : Flock locks might destroy just granted lock if it could be merged
1191 with another existing flock, this is done in completion handler,
1192 so teach ldlm_cli_enquque_fini that this is a valid case for
1198 Description: Oops in read and write path when failing to allocate lock.
1199 Details : Check if lock allocation failed and return error back.
1203 Description: lstripe command fails for valid OST index
1204 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1205 instead of lov->desc.ld_active_tgt_count.
1209 Description: sanity.sh failed test 103
1210 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1215 Description: df doesn't work properly if diskfs blocksize != 4K
1216 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1220 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1221 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1223 Severity : enhancement
1225 Description: Adaptive timeouts
1226 Details : RPC timeouts adapt to changing server load and network
1227 conditions to reduce resend attempts and improve recovery time.
1231 Description: llapi_file_create() does not allow some changes
1232 Details : add llapi_file_open() that allows specifying the mode and
1233 open flags, and also returns an open file handle.
1237 Description: merge and cleanup kernel patches.
1238 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1242 Description: Add journal checksum support.(Kernel part)
1243 Details : The journal checksum feature adds two new flags i.e
1244 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1245 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1246 indicates that the commit block contains the checksum for
1247 the blocks described by the descriptor blocks. Now commit
1248 record can be sent to disk without waiting for descriptor
1249 blocks to be written to disk. This behavior is controlled
1250 using JBD2_FEATURE_ASYNC_COMMIT flag.
1254 Description: OSS needs mutliple precreate threads
1255 Details : Add ability to start more than one create thread per OSS.
1259 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1260 Details : execute lfs setstripe on client
1264 Description: mds_obd_create error creating tmp object
1265 Details : When the user sets quota on root, llog will be affected and can't
1266 create files and write files.
1269 Frequency : Always on ia64 patchless client, and possibly others.
1271 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1272 Details : This allows the patchless client to be loaded on architectures
1273 without this export.
1277 Description: RedHat Update kernel for RHEL5
1278 Details : Add the kernel config file for RHEL5.
1282 Description: RedHat Update kernel for RHEL5
1283 Details : Modify the kernel config file more closer RHEL5.
1287 Description: Build failure against Centos5 (RHEL5)
1288 Details : Define PAGE_SIZE when it isn't present.
1292 Description: test_7 on sanity-quota.sh can't be used on separate machines
1293 Details : add facet to handle it
1297 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1298 Details : with NFS, the anon dentry's parent was set to itself in
1299 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1300 corresponding dentry other than use rec->ur_name.
1302 Severity : enhancement
1304 Description: lfs setstripe enhancement
1305 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1309 Description: enable data checksumming by default
1310 Details : enable checksum by default, allow --disable-checksum
1311 configure option and "-o nochecksum" mount option. Checksums
1312 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1316 Description: lustre support for RHEL5
1317 Details : Add support for RHEL5.
1321 Description: Client eviction due to failover config
1322 Details : after a connection loss, the lustre client should attempt to
1323 reconnect to the last active server first before trying the
1324 other potential connections.
1328 Description: when mds and osts use different quota unit(32bit and 64bit),
1329 quota will be released repeatly.
1330 Details : void sending multiple quota reqs to mds, which will keep the
1331 status between the reqs.
1335 Description: osts not allocated evenly to files
1336 Details : change the condition to increase offset_idx
1339 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1341 Description: Data corruption for OSTs that are formatted larger than 2TB
1343 Details : When generating the bio request for lustre file writes the
1344 sector number would overflow a temporary variable before being
1345 used for the IO. The data reads correctly from Lustre (which
1346 will overflow in a similar manner) but other file data or
1347 filesystem metadata may be corrupted in some cases.
1351 Description: Tune the kernel for good SCSI performance.
1352 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1353 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1358 Description: rewrite ext2-derived code in obdclass/uuid.c
1359 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1360 from scratch preserving functionality.
1364 Description: "lfs find -obd UUID" prints directories
1365 Details : "lfs find -obd UUID" will return all directory names instead
1366 of just file names. It is incorrect because the directories
1367 do not reside on the OSTs.
1371 Description: use i_size_read and i_size_write in 2.6 port
1372 Details : replace inode->i_size access with i_size_read/write()
1375 Frequency : when removing large files
1377 Description: scheduling issue during removal of large Lustre files
1378 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1379 It causes scheduling issues when removing large files (17TB in the
1383 Frequency : only with liblustre clients on XT3
1385 Description: evictions taking too long
1386 Details : allow llrd to evict clients directly on OSTs
1389 Frequency : only on ppc
1391 Description: /proc/fs/lustre/devices broken on ppc
1392 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1393 We should make sure the type of 'index' is loff_t and then cast
1394 explicitly as needed below. Do not assign an explicitly cast
1398 Frequency : only for rhel5
1400 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1401 Details : Modify the target file & which_kernel.
1403 Severity : enhancement
1405 Description: omit set fsid for export NFS
1406 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1407 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1411 Frequency : Always, for kernels after 2.6.16
1412 Description: Fix warning idr_remove called for id=.. which is not allocated.
1413 Details : Last kernels save old s_dev before kill super and not allow
1414 to restore from callback - restore it before call kill_anon_super.
1418 Description: buffer overruns could theoretically occur
1419 Details : llapi_semantic_traverse() modifies the "path" argument by
1420 appending values to the end of the origin string, and a
1421 overrun may occur. Adding buffer overrun check in liblustreapi.
1425 Description: Fix error on 'ls .' at the top of the Lustre mount.
1426 Details : Don't revalidate dentry if it is a root dentry.
1430 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1431 Details : Modify vm-tunables-rhel4.patch.
1435 Description: Kernel config for 2.6.18-vanilla.
1436 Details : Modify targets/2.6-vanilla.target.in.
1437 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1438 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1439 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1440 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1444 Description: improve handling recoverable errors
1445 Details : if request processig with error which can be recoverable on server
1446 request should be resend, otherwise page released from cache and
1451 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1452 Details : Modify target file & which_patch.
1453 A flaw was found in the IA32 system call emulation provided
1454 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1455 value could be stored in the %RAX register, which could trigger an
1456 out-of-bounds system call table access. An untrusted local user
1457 could exploit this flaw to run code in the kernel
1458 (ie a root privilege escalation). (CVE-2007-4573).
1462 Description: O_DIRECT bypasses client statistics.
1463 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1464 read_ahead_stats were updated. copied the stats section from
1465 osc_send_oap_rpc() into async_internal().
1469 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1473 Description: change order of libsysio includes
1474 Details : '#include sysio.h' should always come before '#include xtio.h'
1478 Description: handle "serious error: objid * already exists" more gracefully
1479 Details : If LAST_ID value on disk is smaller than the objects existing in
1480 the O/0/d* directories, it indicates disk corruption and causes an
1481 LBUG(). If the object is 0-length, then we should use the existing
1482 object. This will help to avoid a full fsck in most cases.
1484 Severity : enhancement
1486 Description: adapt the lustre_config script to support the upgrade case
1487 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1488 server targets to 1.6 in parallel.
1492 Description: To avoid grant space > avaible space when the disk is almost
1493 full. Without this patch you might see the error "grant XXXX >
1494 available" or some LBUG about grant, when the disk is almost
1496 Details : In filter_check_grant, for non_grant cache write, we should
1497 check the left space by if (*left > ungranted + bytes), instead
1498 of (*left > ungranted), because only we are sure the left space
1499 is enough for another "bytes", then the ungrant space should be
1500 increase. In client, we should update cl_avail_grant only there
1501 is OBD_MD_FLGRANT in the reply.
1505 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1506 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1507 described in CVE-2007-4573. This problem could allow untrusted
1508 local users to gain root access.
1511 Frequency : when using O_DIRECT and quotas
1513 Description: Incorrect file ownership on O_DIRECT output files
1514 Details : block usage reported by 'lfs quota' does not take into account
1515 files that have been written with O_DIRECT.
1520 Description: touch file failed when fs is not full
1521 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1522 otherwise we can get ENOSP while fs is not full.
1526 Description: parallel lock callbacks
1527 Details : Instead of sending blocking and completion callbacks as separated
1528 requests, adding them to a set and sending in parallel.
1531 Frequency : only for Cray XT3
1532 Bugzilla : 12829/13455
1533 Description: Changing primary group doesn't change the group lustre assigns to
1535 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1536 overridden with the primary group provided by the group upcall,
1537 whereas the supplied fsgid can be trusted if it is in the list of
1538 supplementary groups returned by the group upcall.
1540 Severity : enhancement
1542 Description: Allow masking D_WARNING, D_ERROR messages from console
1543 Details : Console messages can now be disabled via lnet.printk.
1547 Description: User code with malformed file open parameter crashes client node
1548 Details : Before packing join_file req, all the related reference should be
1549 checked carefully in case some malformed flags cause fake
1550 join_file req on client.
1554 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1555 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1556 mode is equal to the granted mode. However, it should also check
1557 that LDLM_FL_CP_REQD is not set.
1561 Description: Heavy nfs access might result in deadlocks
1562 Details : After ELC code landed, it is now improper to enqueue any mds
1563 locks under och_sem, because enqueue might want to decide to
1564 cancel open locks for same inode we are holding och_sem for.
1568 Description: Client eviction while running blogbench
1569 Details : A lot of unlink operations with concurrent I/O can lead to a
1570 deadlock causing evictions. To address the problem, the number of
1571 oustanding OST_DESTROY requests is now throttled to
1572 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1573 callbacks are processed in priority.
1577 Description: enable ACLs on MDS by default
1578 Details : ACLs must be enabled on MDS by default.
1581 Frequency : PPC/PPC64 only
1583 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1584 Details : fix duplicated definitions between asm-ppc64/types.h and
1585 lustre_types.h on PPC.
1588 Frequency : PPC/PPC64 only
1590 Description: asm-ppc/segment.h does not exist
1591 Details : fix compile issue on PPC.
1595 Description: better handle error messages in extents code
1598 Frequency : RHEL4 only
1600 Description: mkfs is very slow on IA64/RHEL4
1601 Details : A performance regression has been discovered in the MPT Fusion
1602 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1603 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1604 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1606 Severity : enhancement
1608 Description: SNMP support enhancement
1609 Details : Adding total number of sampled request for an MDS node in snmp
1612 Severity : enhancement
1614 Description: Optimize ldlm waiting list processing for PR extent locks
1615 Details : When processing waiting list for read extent lock and meeting read
1616 lock that is same or wider to it that is not contended, skip
1617 processing rest of the list and immediatelly return current
1618 status of conflictness, since we are guaranteed there are no
1619 conflicting locks in the rest of the list.
1623 Description: Time out and refuse to reconnect
1624 Details : When the failover node is the primary node, it is possible
1625 to have two identical connections in imp_conn_list. We must
1626 compare not conn's pointers but NIDs, otherwise we can defeat
1627 connection throttling.
1631 Description: port llog fixes from b1_6 into HEAD
1632 Details : Port llog reference couting and some llog cleanups from b1_6
1633 (bug 10800) into HEAD, for protect from panic and access to already
1634 free llog structures.
1638 Description: Detect stride IO mode in read-ahead
1639 Details : When a client does stride read, read-ahead should detect that and
1640 read-ahead pages according to the detected stride pattern.
1644 Description: data checksumming impacts single node performance
1645 Details : add support for several checksum algorithm. Currently, only CRC32
1646 and Adler-32 are supported. The checksum type can be changed on
1647 the fly via /proc/fs/lustre/osc/*/checksum_type.
1651 Description: use adler32 for page checksums
1652 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1657 Description: build for x2 fails
1658 Details : fix compile issue on Cray systems.
1662 Description: Properly match for duplicate locks
1663 Details : Due to different lock order from skiplists code, we need to
1664 traverse entire list for now
1667 Frequency : only on PPC/SLES10
1669 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1670 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1671 be defined. Add a hack in mkfs_lustre.c to work around this compile
1676 Description: LASSERT on MDS when client holding flock lock dies
1677 Details : ldlm pool logic depends on number of granted locks equal to
1678 number of released locks which is not true for flock locks, so
1679 just exclude such locks from consideration.
1683 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1684 Details : Use fsfilt_sync() for both the whole filesystem sync and
1685 individual file sync to eliminate dangerous inode locking
1686 with I_LOCK that can lead to a deadlock.
1690 Description: performance in 1.6.3
1691 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1692 sure that 1MB requests are not fragmented by the block layer.
1694 Severity : enhancement
1696 Description: organize the server-side client stats on per-nid basis
1697 Details : Change the structure of stats under obdfilter and mds to
1705 The "uuid"s file would list the uuids of _active_ exports.
1706 And the clear entry is to clear all stats and stale nids.
1708 Severity : enhancement
1710 Description: eliminate client locks in face of contention
1711 Details : file contention detection and lockless i/o implementation
1712 for contended files.
1716 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1719 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1721 Description: MDS may lose file striping (and hence file data) in some cases
1722 Details : If there are additional extended attributes stored on the MDS,
1723 in particular ACLs, SELinux, or user attributes (if user_xattr
1724 is specified for the client mount options) then there is a risk
1725 of attribute loss. Additionally, the Lustre file striping
1726 needs to be larger than default (e.g. striped over all OSTs),
1727 and an additional attribute must be stored initially in the
1728 inode and then increase in size enough to be moved to the
1729 external attribute block (e.g. ACL growing in size) for file
1734 Description: skiplist implementation simplification
1735 Details : skiplists are used to group compatible locks on granted list
1736 that was implemented as tracking first and last lock of each
1737 lock group the patch changes that to using doubly linked lists
1741 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1742 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1743 because LASSERT on both the data supplied by a client, and the
1744 data on disk is dangerous and incorrect.
1746 Severity : enhancement
1748 Description: Slow truncate/writes to huge files at high offsets.
1749 Details : Directly associate cached pages to lock that protect those pages,
1750 this allows us to quickly find what pages to write and remove
1751 once lock callback is received.
1755 Description: more ldlm soft lockups
1756 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1757 starve other threads from the resource lock for a long time in
1758 case of long waiting queue, so change the debug level from
1759 D_OTHER to the less frequently used D_INFO.
1761 Severity : enhancement
1763 Description: add -gid, -group, -uid, -user options to lfs find
1767 Description: Hung threads in invalidate_inode_pages2_range
1768 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1769 once one is completed. As a result, some RPCs are stuck in the
1770 queue and are never sent.
1774 Description: filter threads hungs on waiting journal commit
1775 Details : Cleanup filter group llog code, then only filter group llog will
1776 be only created in the MDS/OST syncing process.
1780 Description: Procfs and llog threads access destoryed import sometimes.
1781 Details : Sync the import destoryed process with procfs and llog threads by
1782 the import refcount and semaphore.
1784 Severity : enhancement
1786 Description: openlock cache of b1_6 port to HEAD
1791 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1792 Details : If insertion of an extent fails, then discard the inode
1793 preallocation and free data blocks else it can lead to duplicate
1798 Description: don't always update ctime in ext3_xattr_set_handle()
1799 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1800 In some cases the ctime should not be updated, for example for
1801 2.0->1.8 compatibility it is necessary to delete an xattr and it
1802 should not update the ctime.
1806 Bugzilla : 15713/16362
1807 Description: Assertion in iopen_connect_dentry in 1.6.3
1808 Details : looking up an inode via iopen with the wrong generation number can
1809 populate the dcache with a disconneced dentry while the inode
1810 number is in the process of being reallocated. This causes an
1811 assertion failure in iopen since the inode's dentry list contains
1812 both a connected and disconnected dentry.
1816 Description: assertion failure in ldlm_handle2lock()
1817 Details : fix a race between class_handle_unhash() and class_handle2object()
1818 introduced in lustre 1.6.5 by bug 13622.
1823 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1824 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1829 Description: oops in sd_iostats_seq_show()
1830 Details : unloading/reloading the scsi low level driver triggers a kernel
1831 bug when trying to access the sd iostat file.
1836 Description: Kernel panics during QLogic driver reload
1837 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1838 patch, causing memory corruption.
1843 Description: journal_dev option does not work in b1_6
1844 Details : pass mount option during pre-mount.
1846 Severity : enhancement
1848 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1849 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1850 extent information of a file. It can be used to map logical blocks
1851 in a file to physical blocks in the block device.
1855 Description: LDLM soft lockups - improvement
1856 Details : It is be possible to send the lock handle along with each read
1857 or write request because the client is already doing a lock match
1858 itself so there isn't any reason the OST should have to re-do that
1862 Frequency : only on Cray X2
1864 Description: X2 build failures
1865 Details : fix build failures on Cray X2.
1869 Description: xid & resent requests
1870 Details : Initialize RPC XID from clock at startup (randomly if clock is
1873 Severity : enhancement
1875 Description: Add lustre_start utility to start or stop multiple Lustre servers
1880 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1881 Details : In case of memory pressure, list_del() can be called twice on
1882 req->rq_history_list, causing a kernel oops.
1886 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1887 Details : fix stack overflow in the distributed lock manager by defering
1888 export eviction after a failed AST to the elt thread instead of
1889 handling it in the dlm interpret routine.
1893 Description: Convert some comments to new format.
1894 Details : Update documenting comments to match doxygen conventions.
1898 Description: Grammar fixes.
1899 Details : A couple of trivial spelling fixes.
1903 Description: OSD_COUNTERS-mandatory
1904 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1905 to simplify the code.
1909 Description: simplify cmm_device freeing logic.
1910 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1914 Description: Add lockdep support to dt_object_operations locking interface.
1915 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1916 parameter indicating lock ordering. Update mdd code to use new
1921 Description: Introduce failloc constants for lockless IO tests.
1922 Details : Add two new failloc constants to test lockless IO. Only one of
1923 them in implemented---another is checked in yet to be landed
1928 Description: Add lockdep support for inode mutex.
1929 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1930 in the situations where Lustre has to lock more than one inode
1935 Description: Add optional invariants checking support.
1936 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1937 configure switch. This macro is to be used for consistency and
1938 sanity checks that are too expensive to be left in `production'
1943 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1944 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1949 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1950 Details : Use spin_lock_nested() in (the only) situation where more than
1951 one ldlm_lock is locked simultaneously. Also, fix possible
1952 dead-lock in ldlm_lock_change_resource() by enforcing particular
1957 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1958 Details : Instead of specifying each ldlm_lock call-back through separate
1959 parameter, wrap them into struct ldlm_callback_suite.
1963 Description: Kill join_lru obd method and its callers.
1964 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1965 in memory---a requirement implemented through ->o_join_lru() obd
1966 method in HEAD. Remove this method and its users.
1970 Description: Add asynchronous ldlm ENQUEUE completion handler.
1971 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1972 case that stock ldlm_completion_ast() cannot handle as it waits
1973 until lock is granted. Introduce new ldlm_completion_ast_async()
1974 for this. Also comment ldlm_completion_ast().
1978 Description: ldlm_error <-> errno conversion.
1979 Details : Add functions to map (rather arbitrary) between LDLM error codes
1980 and standard errno values. CLIO needs this to prevent LDLM specific
1981 constants from escaping ldlm and osc.
1985 Description: Kill unused ldlm_handle2lock_ns() function.
1986 Details : Kill unused ldlm_handle2lock_ns() function.
1990 Description: Add lu_ref support to ldlm_lock
1991 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1992 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1993 and ldlm_resource. LDLM interface has to be changed, because code that
1994 releases a reference on a lock, has to "know" what reference this is.
1995 In the most frequent case
1997 lock = ldlm_handle2lock(handle);
1999 LDLM_LOCK_PUT(lock);
2001 no changes are required. When any other reference (received _not_ from
2002 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2003 instead of LDLM_LOCK_PUT().
2005 Arguably, changes are pervasive, and interface requires some discipline
2006 for proper use. On the other hand, it was very instrumental in finding
2007 a few leaked lock references.
2011 Description: Add ldlm_lock_addref_try().
2012 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2013 attempts to addref a lock that might be being canceled
2018 Description: Add ldlm_weigh_callback().
2019 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2020 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2021 hard-coded `number of pages' logic.
2025 Description: Add lockdep annotations to llog code.
2026 Details : Use appropriately tagged _nested() locking calls in the places
2027 where llog takes more than one ->lgh_lock lock.
2031 Description: Add loi_kms_set().
2032 Details : Wrap kms updates into a helper function.
2036 Description: Constify instances of struct lsm_operations.
2037 Details : Constify instances of struct lsm_operations.
2041 Description: lu_conf support.
2042 Details : On a server, a file system object is uniquely identified
2043 by a fid, which is sufficient to locate and load all object
2044 state (inode). On a client, on the other hand, more data are
2045 necessary instantiate an object. Change lu_object_find() and
2046 friends to take additional `lu_conf' argument describing object.
2047 Typically this includes layout information.
2051 Description: lu_context fixes.
2052 Details : Introduce new lu_context functions that are needed on the client
2053 side, where some system threads (ptlrpcd) are shared by multiple
2054 modules, and so cannot be stopped during module shutdown.
2058 Description: Add start and stop methods to lu_device_type_operations.
2059 Details : Introduce two new methods in lu_device_type_operations, that are
2060 invoked when first instance of a given type is created and last one
2061 is destroyed respectively. This is need by CLIO.
2065 Description: Add lu_ref support to struct lu_device.
2066 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2067 track leaked references.
2071 Description: Introduce lu_kmem_descr.
2072 Details : lu_kmem_descr and its companion interface allow to create
2073 and destroy a number of kmem caches at once.
2077 Description: Fix lu_object finalization race.
2078 Details : Fix a race between lu_object_find() finding an object and its
2079 concurrent finalization. This race is (most likely) not possible
2080 on the server, but might happen on the client.
2084 Description: Introduce lu_ref interface.
2085 Details : lu_ref is a debugging module allowing to track references to
2086 a given object. It is quite cpu expensive, and has to be
2087 explicitly enabled with --enable-lu_ref. See usage description
2092 Description: Factor lu_site procfs stats into a separate function.
2093 Details : Separate lu_site stats printing code into a separate function
2094 that can be reused on a client.
2098 Description: Constify instances of struct {lu,dt,md}_device_operations.
2099 Details : Constify instances of struct {lu,dt,md}_device_operations.
2103 Description: Introduce struct md_site and move meta-data specific parts of
2104 struct lu_site here.
2105 Details : Move md-specific fields out of struct lu_site into special struct
2106 md_site, so that lu_site can be used on a client.
2110 Description: Kill mdd_lov_destroy().
2111 Details : Remove unused mdd code.
2115 Description: Add st_block checking to multistat.c.
2116 Details : Add st_block checking to multistat.c.
2120 Description: Add lu_ref support to struct obd_device.
2121 Details : Add lu_ref tracking to obd_device.
2125 Description: Kill obd_set_fail_loc().
2126 Details : Remove unused code.
2130 Description: Add special type for ptlrpc_request interpret functions.
2131 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2132 there. Actual usage will be in CLIO.
2136 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2137 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2138 doesn't work with lockdep.
2142 Description: Add rwv.c test program.
2143 Details : New testing program exercising readv(2) and writev(2) (Qian).
2147 Description: Add sendfile.c test program.
2148 Details : New testing program exercising sendfile(2) (Jay).
2152 Description: Ratelimit a message that can be very frequent.
2153 Details : Ratelimit a memory allocation failure message that can
2158 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2159 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2163 Description: Liblustre build fixes.
2164 Details : Liblustre build fixes.
2168 Description: libcfs: add cfs_{need,cond}_resched() interface.
2169 Details : libcfs: add cfs_{need,cond}_resched() definition and
2170 implementations for Linux, NT, and liblustre.
2172 Severity : enhancement
2174 Description: More exported tunables for mballoc
2175 Details : Add support for tunable preallocation window and new tunables for
2176 large/small requests
2180 Description: Detect corruption of block bitmap and checking for preallocations
2181 Details : Checks validity of on-disk block bitmap. Also it does better
2182 checking of number of applied preallocations. When corruption is
2183 found, it turns filesystem readonly to prevent further corruptions.
2187 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2188 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2189 window to 0. If it does hit cache miss, and read-pattern is still
2190 stride-io mode, does not reset the stride window, but also does not
2191 increase the stride window length in this case.
2195 Frequency : only for big-endian servers
2196 Description: Check if system is big-endian while mounting fs with extents feature
2197 Details : Mounting a filesystem with extents feature will fail on big-endian
2198 systems since ext3-based ldiskfs is not supported on big-endian
2199 systems. This can be over-riden with "bigendian_extents" mount option.
2201 Severity : enhancement
2203 Description: The root squash functionality
2204 Details : A security feature, which is to prevent users from being able
2205 to mount lustre on their desktop, run as root, and delete
2206 all of the files in the filesystem. The goal is accomplished by
2207 remapping user id (UID) and group id (GID) of the root user to
2208 a UID and GID specified by the system administartor via Lustre
2209 configuration management server (MGS). The functionality also
2210 allows to specify sets of clients for which the remapping does
2215 Description: Excessive recovery window
2216 Details : With AT enabled, the recovery window can be excessively long (6000+
2217 seconds). To address this problem, we no longer use
2218 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2219 timeout no longer depends on the service time, it is set to
2220 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2221 time via pb_service_time.
2225 Description: Watchdog triggered on MDS failover
2226 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2227 the OSTs know that the MDS "UUID" can be reused for the same export
2228 from a different NID, so we do not need to wait for the export to be
2232 Frequency : rare, only if using MMP with Linux RAID
2234 Description: MMP doesn't work with Linux RAID
2235 Details : While using HA for Lustre servers with Linux RAID, it is possible
2236 that MMP will not detect multiple mounts. To make this work we
2237 need to unplug the device queue in RAID when the MMP block is being
2238 written. Also while reading the MMP block, we should read it from
2239 disk and not the cached one.
2241 Severity : enhancement
2243 Description: open file using fid
2244 Details : A file can be opened using just its fid, like
2245 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2248 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2251 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2252 Details : add proper locking for creation of jbd_4k slab cache
2256 Description: MMP check in ext3_remount() fails without displaying any error
2257 Details : When multiple mount protection fails during remount, proper error
2260 Severity : enhancement
2262 Description: Allow stripe size to be up to 4G-64k
2263 Details : Fix math logic to allow large stripe sizes.
2267 Description: add check for >8TB ldiskfs filesystems
2268 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2269 Don't allow >8TB ldiskfs filesystems to be mounted without
2270 force_over_8tb mount option
2275 Descriptoin: OSS complains "too many blocks" and crashes
2276 Details : In filter_commitrw_write():
2277 1. lquota_chkquota();
2278 2. obdo_from_inode();
2279 3. lquota_pending_commit();
2280 "2." has changed uid and gid in this case, then caused
2281 lquota_chkquota() and lquota_pending_commit() to use
2282 different uid and gid. That is the root of the bug.
2286 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2287 Details : Detect and print out the 'stripe count = -1' case properly.
2289 --------------------------------------------------------------------------------
2291 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2293 * Support for kernels:
2294 2.6.5-7.283 (SLES 9),
2295 2.6.9-55.EL (RHEL 4),
2296 2.6.16.46-0.14 (SLES 10),
2297 2.6.18.8 vanilla (kernel.org)
2298 * Client support for unpatched kernels:
2299 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2300 2.6.16 - 2.6.22 vanilla (kernel.org)
2301 * Due to recently discovered recovery problems, we do not recommend
2302 using patchless RHEL 4 clients with this or any earlier release.
2303 * Recommended e2fsprogs version: 1.39.cfs8
2304 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2305 * Starting with this release, the ldiskfs backing filesystem required
2306 by Lustre is now in its own package, lustre-ldiskfs. This package
2307 should be installed. It is versioned separately from Lustre and
2308 may be released separately in future.
2312 Description: block reactivating mgc import until all deactivates complete
2313 Details : Fix race when failing back MDT/MGS to itself (testing)
2316 Frequency : at statup only
2318 Description: mds_lov_synchronize race leads to various problems
2319 Details : simultaneous MDT->OST connections at startup can cause the
2320 sync to abort, leaving the OSC in a bad state.
2322 Severity : enhancement
2324 Description: add optional extra BUILD_VERSION info
2325 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2326 to override the lustre version.
2329 Frequency : 2.6.18 servers only
2331 Description: ll_kern_mount() doesn't release the module reference
2332 Details : The ldiskfs module reference count never drops down to 0
2333 because ll_kern_mount() doesn't release the module reference.
2338 Description: server LBUG when using old ost_num_threads parameter
2339 Details : Accept the old ost_num_threads parameter but warn that it
2340 is deprecated, and fix an off-by-one error that caused an LBUG.
2345 Description: Transient SCSI error results in persistent IO issue
2346 Details : iobuf->dr_error is not reinitialized to 0 between two
2350 Frequency : sometimes when underlying device returns I/O errors
2352 Description: OSTs not going read-only during write failures
2353 Details : OSTs are not remounted read-only when the journal commit threads
2354 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2355 instead of the ext3 wrappers.
2359 Description: poor connect scaling with increasing client count
2360 Details : Don't run filter_grant_sanity_check for more than 100 exports
2361 to improve scaling for large numbers of clients.
2364 Frequency : SLES10 only
2366 Description: sanity-quota.sh quotacheck failed: rc = -22
2367 Details : Quotas cannot be enabled on SLES10.
2370 Frequency : liblustre clients only
2372 Description: getdirentries does not give error when run on compute nodes
2373 Details : getdirentries does not fail when the size specified as an argument
2374 is too small to contain at least one entry
2376 Severity : enhancement
2378 Description: Add LNET router traceability for debug purposes
2379 Details : If a checksum failure occurs with a router as part of the
2380 IO path, the NID of the last router that forwarded the bulk data
2381 is printed so it can be identified.
2386 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2387 Details : Due to a race condition, liblustre clients were occasionally
2388 evicted incorrectly.
2390 Severity : enhancement
2392 Description: lfs setstripe use optional parameters instead of postional
2395 Severity : enhancement
2397 Description: Nanosecond timestamp support for ldiskfs
2398 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2399 resolution timestamps. There is not yet support for this at
2400 the Lustre filesystem level.
2403 Frequency : during server recovery
2405 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2406 Details : request with rq_no_resend flag not awake l_wait_event if they get
2410 Frequency : nfs export on patchless client
2412 Description: connectathon hang when test nfs export over patchless client
2413 Details : Disconnected dentry cannot be found with lookup, so we do not need
2414 to unhash it or make it invalid
2417 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2418 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2419 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2424 Description: open req refcounting wrong on reconnect
2425 Details : If reconnect happened between getting open reply from server and
2426 call to mdc_set_replay_data in ll_file_open, we will schedule
2427 replay for unreferenced request that we are about to free.
2428 Subsequent close will crash in variety of ways.
2429 Check that request is still eligible for replay in
2430 mdc_set_replay_data().
2435 Description: disable writes to filesystem when reading health_check file
2436 Details : the default for reading the health_check proc file has changed
2437 to NOT do a journal transaction and write to disk, because this
2438 can cause reads of the /proc file to hang and block HA state
2439 checking on a healthy but otherwise heavily loaded system. It
2440 is possible to return to the previous behaviour during configure
2441 with --enable-health-write.
2443 Severity : enhancement
2445 Description: 64-bit inode version
2446 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2447 to the inode. This will be required for version-based recovery.
2452 Description: MDS fails to start if a duplicate client export is detected
2453 Details : in some rare cases it was possible for a client to connect to
2454 an MDS multiple times. Upon recovery the MDS would detect this
2455 and fail during startup. Handle this more gracefully.
2457 Severity : enhancement
2459 Description: Add -o localflock option to simulate old noflock
2461 Details : This will achieve local-only flock/fcntl locks
2467 Description: log_commit_thread vs filter_destroy race leads to crash
2468 Details : Take import reference before releasing llog record semaphore
2473 Description: Wrong request locking in request set processing
2474 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2475 imp_delayed_list, in this place should be used imp_lock.
2478 Frequency : when reconnection
2480 Description: Grant Leak when osc reconnect to OST
2481 Details : When osc reconnect ost, OST(filter) should check whether it
2482 should grant more space to client by comparing fed_grant and
2483 cl_avail_grant, and return the granted space to client instead
2484 of "new granted" space, because client will call osc_init_grant
2485 to update the client grant space info.
2488 Frequency : when client reconnect to OST
2490 Description: Grant Leak when osc do resend and replay bulk write
2491 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2492 bulk write request, because the grant info while be sync between
2493 OSC and OST when reconnect, and we should ignore the grant info
2494 these of resend/replay write req.
2499 Description: Grant space more than avaiable left space sometimes.
2500 Details : When then OST is about to be full, if two bulk writing from
2501 different clients came to OST. Accord the avaliable space of the
2502 OST, the first req should be permitted, and the second one
2503 should be denied by ENOSPC. But if the seconde arrived before
2504 the first one is commited. The OST might wrongly permit second
2505 writing, which will cause grant space > avaiable space.
2508 Frequency : when client is evicted
2510 Description: Grant might be wrongly erased when osc is evicted by OST
2511 Details : when the import is evicted by server, it will fork another
2512 thread ptlrpc_invalidate_import_thread to invalidate the
2513 import, where the grant will be set to 0. While the original
2514 thread will update the grant it got when connecting. So if
2515 the former happened latter, the grant will be wrongly errased
2516 because of this race.
2521 Description: Checking Stale with correct fid
2522 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2523 to check whether it is stale, because sometimes, we want the
2524 enqueue happened anyway, and op_data.fid2 will not be initialized.
2526 Severity : enhancement
2528 Description: update patchless client
2529 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2532 Frequency : only with 2.4 kernel
2534 Description: random memory corruption
2535 Details : size of struct ll_inode_info is to big for union inode.u and this
2536 can be cause of random memory corruption.
2541 Description: Memory leak in recovery
2542 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2543 It should also check obd_fail before fsfilt_start, otherwise if
2544 fsfilt_start return -EROFS,(failover mds during mds recovery).
2545 then the req will return with repmsg->transno = 0 and rc = EROFS.
2546 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2547 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2548 be freed no matter whether fsfilt_commit success or not.
2551 Frequency : only with huge count clients
2553 Description: Prevents from taking the superblock lock in llap_from_page for
2555 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2556 for llap_from_page call in ll_removepage prevents from taking the
2557 superblock lock for a soon died page.
2562 Description: Not check open intent error before release open handle
2563 Details : in some rare cases, the open intent error is not checked before
2564 release open handle, which may cause
2565 ASSERTION(open_req->rq_transno != 0), because it tries to release
2566 the failed open handle.
2571 Description: Set cat log bitmap only after create log success.
2572 Details : in some rare cases, the cat log bitmap is set too early. and it
2573 should be set only after create log success.
2577 Description: Accessing a block bevice can re-enable I/O when Lustre is
2578 tearing down a device.
2579 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2584 Description: service threads may hog cpus when there are a lot of requests
2585 Details : Insert cond_resched to give other threads a chance to use some CPU
2590 Description: the cat log was not initialized in recovery
2591 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2592 unlink log on mds will not be initialized until mds post
2593 recovery. And also in mds post recovery, the unlink log will
2594 initialization will be done asynchronausly, so there will be race
2595 between add unlink log and unlink log initialization.
2599 Description: brw_stats were being printed incorrectly
2600 Details : brw_stats were being printed as log2 but all of them were not
2601 recorded as log2. Also remove some code duplication arising from
2602 filter_tally_{read,write}.
2606 Frequency : rare, only in recovery.
2607 Description: ASSERTION(req->rq_type != LI_POISON) failed
2608 Details : imp_lock should be held while iterating over imp_sending_list for
2609 prevent destroy request after get timeout in ptlrpc_queue_wait.
2613 Description: replay-single.sh test 52 fails
2614 Details : A lock's skiplist need to be cleanup when it being unlinked
2615 from its resource list.
2619 Description: Short directio read returns full requested size rather than
2621 Details : Direct I/O operations should return actual amount of bytes
2622 transferred rather than requested size.
2624 Severity : enhancement
2626 Description: metadata RPC reduction (e.g. for rm performance)
2627 Details : decrease the amount of synchronous RPC between clients and servers
2628 by canceling conflicing lock before the operation on the client
2629 and packing thier handles into the main operation RPC to server.
2631 Severity : enhancement
2633 Description: Async OSC create to avoid the blocking unnecessarily.
2634 Details : If a OST has no remain object, system will block on the creating
2635 when need to create a new object on this OST. Now, ways use
2636 pre-created objects when available, instead of blocking on an
2637 empty osc while others are not empty. If we must block, we block
2638 for the shortest possible period of time.
2642 Description: improve handling recoverable errors
2643 Details : if request processig with error which can be recoverable on server
2644 request should be resend, otherwise page released from cache and
2647 Severity : enhancement
2649 Description: refine locking for avoid write wrong info into lov_objid file
2650 Details : fix possible races with add new target and write/update data in
2653 --------------------------------------------------------------------------------
2655 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2660 Frequency : on some architectures
2662 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2663 Details : Uninitialized flags sometimes cause configuration commands to
2667 Frequency : patchless clients only
2669 Description: missing __iget() symbol export
2670 Details : The __iget() symbol export is missing. To avoid the need for
2671 this on patchless clients the deathrow inode reaper is turned
2672 off, and we depend on the VM to clean up old inodes. This
2673 dependency was during via the fix for bug 12181.
2675 --------------------------------------------------------------------------------
2677 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2679 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2680 INTEROPERATE with older versions automatically. In many cases a
2681 special upgrade step is needed. Please read the
2682 user documentation before upgrading any part of a live system.
2683 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2684 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2685 * WARNING: Lustre configuration and startup changes are required with
2686 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2690 Severity : enhancement
2692 Description: MountConf
2693 Details : Lustre configuration is now managed via mkfs and mount
2694 commands instead of lmc and lconf. New obd types (MGS, MGC)
2695 are added for dynamic configuration management. See
2696 https://mail.clusterfs.com/wikis/lustre/MountConf for
2699 Severity : enhancement
2701 Description: dynamic OST addition
2702 Details : OSTs can now be added to a live filesystem
2704 Severity : enhancement
2706 Description: startup order invariance
2707 Details : MDTs and OSTs can be started in any order. Clients only
2708 require the MDT to complete startup.
2710 Severity : enhancement
2712 Description: parallel, asynchronous orphan cleanup
2713 Details : orphan cleanup is now performed in separate threads for each
2714 OST, allowing parallel non-blocking operation.
2716 Severity : enhancement
2718 Description: optimized stripe assignment
2719 Details : stripe assignments are now made based on ost space available,
2720 ost previous usage, and OSS previous usage, in order to try
2721 to optimize storage space and networking resources.
2723 Severity : enhancement
2725 Description: Permanently set tunables
2726 Details : All writable /proc/fs/lustre tunables can now be permanently
2727 set on a per-server basis, at mkfs time or on a live system.
2729 Severity : enhancement
2731 Description: Lustre message v2
2732 Details : Add lustre message format v2.
2734 Severity : enhancement
2736 Description: client OST exclusion list
2737 Details : Clients can be started with a list of OSTs that should be
2738 declared "inactive" for known non-responsive OSTs.
2742 Description: SPEC SFS validation failure on NFS v2 over lustre.
2743 Details : Changes the blocksize for regular files to be 2x RPC size,
2744 and not depend on stripe size.
2746 Severity : enhancement
2748 Description: Multiple MD RPCs in flight.
2749 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2750 To avoid overly-overloading MDS, introduce a limit on number of
2751 MDS RPCs in flight for a single client and add /proc controls
2752 to adjust this limit.
2754 Severity : enhancement
2756 Description: client read/write statistics
2757 Details : Add client read/write call usage stats for performance
2758 analysis of user processes.
2759 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2760 file access. extents_stats shows chunk size distribution.
2761 extents_stats_per_process show chunk size distribution per
2764 Severity : enhancement
2766 Description: mds statistics
2767 Details : Add detailed mds operations statistics in
2768 /proc/fs/lustre/mds/*/stats.
2772 Description: Failure of copying files with lustre special EAs.
2773 Details : Client side always return success for setxattr call for lustre
2774 special xattr (currently only "trusted.lov").
2779 Description: Refcount LNET uuids
2780 Details : The global LNET uuid list grew linearly with every startup;
2781 refcount repeated list entries instead of always adding to
2784 Severity : enhancement
2786 Description: Dynamic service threads
2787 Details : Within a small range, start extra service threads
2788 automatically when the request queue builds up.
2791 Frequency : mixed-endian client/server environments
2793 Description: mixed-endian crashes
2794 Details : The new msg_v2 system had some failures in mixed-endian
2797 Severity : enhancement
2799 Description: Easy OST removal
2800 Details : OSTs can be permanently deactivated with e.g. 'lctl
2801 conf_param lustre-OST0001.osc.active=0'
2803 Severity : enhancement
2805 Description: MGS proc entries
2806 Details : Added basic proc entries for the MGS showing what filesystems
2809 Severity : enhancement
2811 Description: provide MGS failover
2812 Details : Added config lock reacquisition after MGS server failover.
2814 Severity : enhancement
2816 Description: add Linux 2.4 support
2817 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2821 Description: a large application tries to do I/O to the same resource and dies
2822 in the middle of it.
2823 Details : Check the req->rq_arrival time after the call to
2824 ost_brw_lock_get(), but before we do anything about
2825 processing it & sending the BULK transfer request. This
2826 should help move old stale pending locks off the queue as
2827 quickly as obd_timeout.
2830 Frequency : when an incorrect nid is specified during startup
2832 Description: ptlrpc connect to non-existant node causes kernel crash
2833 Details : LNET can't be re-entered from an event callback, which
2834 happened when we expire a message after the export has been
2835 cleaned up. Instead, hand the zombie cleanup off to another
2838 Severity : enhancement
2840 Description: plain/inodebits lock performance improvement
2841 Details : Grouping plain/inodebits in granted list by their request modes
2842 and bits policy, thus improving the performance of search through
2846 Frequency : only if OST filesystem is corrupted
2848 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2849 Details : for a short time RPCs with bulk IO are in the replay list,
2850 but replay of bulk IOs is unimplemented. If the OST filesystem
2851 is corrupted due to disk cache incoherency and then replay is
2852 started it is possible to trip an assertion. Avoid putting
2853 committed RPCs into the replay list at all to avoid this issue.
2856 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2858 Description: System hang on startup
2859 Details : This bug allowed the liblustre (e.g. catamount) client to
2860 return to the app before handling all startup RPCs. This
2861 could leave the node unresponsive to lustre network traffic
2862 and manifested as a server ptllnd timeout.
2864 Severity : enhancement
2866 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2867 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2868 boolean to control whether to print peer debug info when a
2869 client's RPC times out.
2872 Frequency : only for kernels with patches from Lustre below 1.4.3
2874 Description: Remove old rdonly API
2875 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2878 Frequency : only for devices with external journals
2880 Description: Set external device read-only also
2881 Details : During a commanded failover stop, we set the disk device
2882 read-only while the server shuts down. We now also set any
2883 external journal device read-only at the same time.
2886 Frequency : when upgrading from 1.4 while trying to change parameters
2888 Description: The wrong (new) MDC name was used when setting parameters for
2889 upgraded MDT's. Also allows changing of OSC (and MDC)
2890 parameters if --writeconf is specified at tunefs upgrade time.
2893 Frequency : when setting specific ost indicies
2895 Description: QOS code breaks on skipped indicies
2896 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2897 created with --index need not be sequential.
2899 Severity : enhancement
2901 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2902 Details : The uninit_groups feature works in conjunction with the kernel
2903 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2904 up the pass1 processing of e2fsck. This is a read-only feature
2905 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2906 mount filesystems that have had this feature enabled.
2908 Severity : enhancement
2910 Description: Improve multi-block allocation algorithm to avoid fragmentation
2911 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2912 allocation locality and avoid filesystem fragmentation.
2914 ------------------------------------------------------------------------------
2916 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2918 * Support for kernels:
2919 2.4.21-47.0.1.EL (RHEL 3)
2920 2.6.5-7.283 (SLES 9)
2921 2.6.9-42.0.10.EL (RHEL 4)
2922 2.6.12.6 vanilla (kernel.org)
2923 2.6.16.27-0.9 (SLES 10)
2924 * Recommended e2fsprogs version: 1.39.cfs5
2926 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2930 Frequency : occasional, depends on client load and configuration
2931 Bugzilla : 12181, 12203
2932 Description: data loss for recently-modified files
2934 Details : In some cases it is possible that recently written or created
2935 files may not be written to disk in a timely manner (this should
2936 normally be within 30s unless client IO load is very high).
2937 The problem appears as zero-length files or files that are a
2938 multiple of 1MB in size after a client crash or client eviction
2939 that are missing data at the end of the file.
2941 This problem is more likely to be hit on clients where files are
2942 repeatedly created and unlinked in the same directory, clients
2943 have a large amount of RAM, have many CPUs, the filesystem has
2944 many OSTs, the clients are rebooted frequently, and/or the files
2945 are not accessed by other nodes after being written.
2947 The presence of the problem can be detected by looking at
2948 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2949 smaller than the second (nr_unused) then dirty files will not
2950 be flushed automatically to disk. "sync; sleep 10" should be
2951 run several times on the node before unmounting it to update
2952 Lustre (this is also safe to run on nodes without this problem).
2954 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2955 that can cause this same problem, so customers using that kernel
2956 also need to update the kernel in addition to Lustre. In order
2957 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2959 It is normal that files written just before a client crash (less
2960 than 30s) may not yet have been flushed to disk, even for local
2964 Frequency : frequent on thin XT3 nodes
2966 Description: UUID collision on thin XT3 Linux nodes
2967 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2968 randomly, since we relied on an insufficiently-seeded PRNG.
2973 Description: OSS hangs after "All ost request buffers busy"
2974 Details : A deadlock between quota and journal operations caused OSS
2975 hangs after printing "All ost request buffers busy."
2978 Frequency : always on liblustre builds
2980 Description: Cleanup compiler warnings on liblustre
2983 Frequency : always on liblustre builds on XT3
2985 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2989 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2990 > 32000 subdirectories
2991 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2992 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2993 subdirectory count crosses 32000. This will aid e2fsck to
2994 correctly handle more than 32000 subdirectories.
2997 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2999 Description: System hang on startup
3000 Details : This bug allowed the liblustre (e.g. catamount) client to
3001 return to the app before handling all startup RPCs. This
3002 could leave the node unresponsive to lustre network traffic
3003 and manifested as a server ptllnd timeout.
3005 Severity : enhancement
3007 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3008 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3009 boolean to control whether to print peer debug info when a
3010 client's RPC times out.
3015 Description: make O_SYNC working on 2.6 kernels
3016 Details : 2.6 kernels use different method for mark pages for write,
3017 so need add a code to lustre for O_SYNC work.
3022 Description: Failure to close file and release space on NFS
3023 Details : Put inode details into lock acquired in ll_intent_file_open.
3024 Use mdc_intent_lock in ll_intent_open to properly
3025 detect all kind of errors unhandled by mdc_enqueue.
3030 Description: proc file read during shutdown sometimes raced obd removal,
3032 Details : Add lock to prevent obd access after proc file removal.
3035 Frequency : Only for files larger than 4GB on 32-bit clients.
3037 Description: improperly doing page alignment of locks
3038 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3039 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3044 Description: RPCs being resent when they shouldn't be
3045 Details : Some RPCs that should not be resent are being resent. This
3046 can cause inconsistencies in the RPC state machine. Do not
3047 resend such requests.
3050 Frequency : rare, only with NFS export
3052 Description: Crash on NFS re-export node
3053 Details : under very unusual load conditions an assertion is hit in
3054 ll_intent_file_open()
3057 Frequency : only if OST filesystem is corrupted
3059 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3060 Details : for a short time RPCs with bulk IO are in the replay list,
3061 but replay of bulk IOs is unimplemented. If the OST filesystem
3062 is corrupted due to disk cache incoherency and then replay is
3063 started it is possible to trip an assertion. Avoid putting
3064 committed RPCs into the replay list at all to avoid this issue.
3069 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3070 Details : Large single O_DIRECT read and write calls can fail to allocate
3071 a sufficiently large buffer to process the request. In case of
3072 allocation failure the allocation is retried with a smaller
3073 buffer and broken into smaller requests.
3075 Severity : enhancement
3077 Description: Add -o localflock option to simulate old noflock behaviour.
3078 Details : This will achieve local-only flock/fcntl locks coherentness.
3083 Description: versioning check is incomplete
3084 Details : Checking the version difference of client vs. server, report
3085 error if the gap is too big.
3090 Description: add support PG_writeback bit
3091 Details : add support for PG_writeback bit for Lustre, for more carefull
3092 work with page cache in 2.6 kernel. This also fix some deadlocks
3093 and remove hack for work O_SYNC with 2.6 kernel.
3095 Severity : enhancement
3097 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3098 Details : The uninit_groups feature works in conjunction with the kernel
3099 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3100 up the pass1 processing of e2fsck. This is a read-only feature
3101 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3102 mount filesystems that have had this feature enabled.
3104 Severity : enhancement
3106 Description: Improve multi-block allocation algorithm to avoid fragmentation
3107 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3108 allocation locality and avoid filesystem fragmentation.
3110 ------------------------------------------------------------------------------
3112 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3114 * Support for kernels:
3115 2.6.9-42.0.3.EL (RHEL 4)
3116 2.6.5-7.276 (SLES 9)
3117 2.4.21-47.0.1.EL (RHEL 3)
3118 2.6.12.6 vanilla (kernel.org)
3119 2.6.16.21-0.8 (SLES10)
3120 * Recommended e2fsprogs version: 1.39.cfs2-0
3122 * The backwards-compatible /proc/sys/portals symlink has been removed
3123 in this release. Before upgrading, please ensure that you change
3124 any configuration scripts or /etc/sysctl.conf files that access
3125 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3126 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3127 in advance of the upgrade on any system running Lustre 1.4.6 or
3128 newer, since /proc/sys/lnet was added in that version.
3129 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3133 Frequency : only when quota is used
3135 Description: avoid scanning export list for quota master
3136 Details : Change the algorithms to avoid scanning export list in order
3137 to improve the efficiency.
3140 Frequency : MDS failover only, very rarely
3142 Description: "went back in time" messages on mds failover
3143 Details : The greatest transno may be lost when the current operation
3144 finishes with an error (transno==0) and the client's last_rcvd
3145 record is over-written. Save the greatest transno in the
3146 mds_last_transno for this case.
3149 Frequency : always for specific kernels and striping counts
3151 Description: client may get "Matching packet too big" without ACL support
3152 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3153 when trying to access files in certain configurations. The
3154 clients should in fact be denied when mounting because they do
3155 not understand ACLs.
3158 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3160 Description: many clients connecting with IO in progress causes connect timeouts
3161 Details : Avoid synchronous journal commits to avoid delays caused by many
3162 clients connecting/disconnecting when bulk IO is in progress.
3163 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3164 OST_IO_PORTAL to avoid delays behind potentially many pending
3168 Frequency : occasionally with multiple writers to a single file
3170 Description: shared writes to file may result in wrong size reported by stat()
3171 Details : Allow growing of kms when extent lock is cancelled
3174 Frequency : always with random mmap IO to multi-striped file
3176 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3177 Details : Only if the hole is at the end of OST object so that kms is too
3178 small. Fix is to increase kms accordingly in ll_nopage.
3181 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3183 Description: writes to a missing object would leak memory on the OST
3184 Details : If there is an inconsistency between the MDS and OST filesystems,
3185 such that the MDS references an object that doesn't exist, writes
3186 to that object will leak memory due to incorrect cleanup in the
3187 error handling path, eventually running out of memory on the OST.
3192 Description: Creating too long symlink causes lustre errors
3193 Details : Check symlink and name lengths before sending requests to MDS.
3196 Frequency : only if flock is enabled (not on by default)
3198 Description: posix locks not released on fd closure on 2.6.9+
3199 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3200 kernels, this caused such locks not to be released on fd close and
3201 then assertions on fs unmount about still used locks.
3204 Frequency : MDS failover only, very rarely
3206 Description: clients may get ASSERTION(granted_lock != NULL)
3207 Details : When request was taking a long time, and a client was resending
3208 a getattr by name lock request. The were multiple lock requests
3209 with the same client lock handle and
3210 mds_getattr_name->fixup_handle_for_resent_request found one of the
3211 lock handles but later failed with ASSERTION(granted_lock != NULL).
3216 Description: handle->h_buffer_credits > 0, assertion failure
3217 Details : h_buffer_credits is zero after truncate, causing assertion
3218 failure. This patch extends the transaction or creates a new
3222 Frequency : NFS re-export or patchless client
3223 Bugzilla : 11179, 10796
3224 Description: Crash on NFS re-export node (__d_move)
3225 Details : We do not want to hash the dentry if we don't have a lock.
3226 But if this dentry is later used in d_move, we'd hit uninitialised
3227 list head d_hash, so we just do this to init d_hash field but
3228 leave dentry unhashed.
3231 Frequency : NFS re-export or patchless client
3233 Description: NFS exports has problem with symbolic link
3234 Details : lustre client didn't properly install dentry when re-exported
3235 to NFS or running patchless client.
3238 Frequency : NFS re-export or patchless client
3240 Description: Various nfs/patchless fixes.
3241 Details : fixes reuse disconected alias for lookup process - this fixes
3242 warning "find_exported_dentry: npd != pd",
3243 fix permission error with open files at nfs.
3244 fix apply umask when do revalidate.
3247 Frequency : occasional
3249 Description: Crash on NFS re-export node
3250 Details : calling clear_page() on the wrong pointer triggered oops in
3251 generic_mapping_read().
3254 Frequency : rarely, using O_DIRECT IO
3256 Description: unaligned directio crashes client with LASSERT
3257 Details : check for unaligned buffers before trying any requests.
3260 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3262 Description: stale data returned from RAID cache
3263 Details : If only a small amount of IO is done to the RAID device before
3264 reading it again it is possible to get stale data from the RAID
3265 cache instead of reading it from disk.
3268 Frequency : always for sles10 kernel
3270 Description: sles10 support
3271 Details : ll_follow_link: compile fixes and using of nd_set_link
3272 under newer kernels.
3275 Frequency : depends on arch, kernel and compiler version, always on sles10
3278 Description: recursive or deep enough symlinks cause stack overflow
3279 Details : getting rid of large stack-allocated variable in
3283 Frequency : depends on hardware
3285 Description: lustre write performance loss in the SLES10 kernel
3286 Details : the performance loss is caused by using of write barriers in the
3287 ext3 code. The SLES10 kernel turns barrier support on by
3288 default. The fix is to undo that change for ldiskfs.
3290 ------------------------------------------------------------------------------
3292 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3294 * Support for kernels:
3295 2.6.9-42.0.3EL (RHEL 4)
3296 2.6.5-7.276 (SLES 9)
3297 2.4.21-47.0.1.EL (RHEL 3)
3298 2.6.12.6 vanilla (kernel.org)
3302 Frequency : quota enabled and large files being deleted
3304 Description: releasing more than 4GB of quota at once hangs OST
3305 Details : If a user deletes more than 4GB of files on a single OST it
3306 will cause the OST to spin in an infinite loop. Release
3307 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3312 Description: statfs data retrieved from /proc may be stale or zero
3313 Details : When reading per-device statfs data from /proc, in the
3314 {kbytes,files}_{total,free,avail} files, it may appear
3315 as zero or be out of date.
3318 Frequency : systems with MD RAID1 external journal devices
3320 Description: lconf's call to blkid is confused by RAID1 journal devices
3321 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3322 of returning all block devices that match the journal UUID.
3325 Frequency : always, for aggregate stripe size over 4GB
3327 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3328 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3329 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3332 Frequency : always if "lfs find" used on a local file/directory
3334 Description: "lfs find" segfaults if used on a local file/directory
3335 Details : The case where a directory component was not specified wasn't
3336 handled correctly. Handle this properly.
3339 Frequency : always on ppc64
3341 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3342 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3343 perform endianness conversion.
3346 Frequency : rarely (truncate to non-zero file size after write under load)
3347 Bugzilla : 10730, 10687
3348 Description: Files padded with zeros to next 4K multiple
3349 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3350 it is possible that files that are truncated to a non-zero size
3351 immediately after being written are filled with zero bytes beyond
3352 the truncated size. No file data is lost.
3354 Severity : enhancement
3356 Description: Allow recovery/failover for liblustre clients.
3357 Details : liblustre clients were unaware of failover configurations until
3360 Severity : enhancement
3362 Description: user file locks should fail when not mounting with flock option
3363 Details : Set up an error-returning stub in ll_file_operations.lock field
3364 to prevent incorrect behaviour when client is mounted without
3365 flock option. Also, set up properly f_op->flock field for
3369 Frequency : always on ia64
3371 Description: "lfs df" loops on printing out MDS statfs information
3372 Details : The obd_ioctl_data was not initialized and in some systems
3373 this caused a failure during the ioctl that did not return
3374 an error. Initialize the struct and return an error on failure.
3377 Frequency : SLES 9 only
3379 Description: Error of copying files with lustre special EAs as root
3380 Details : Client side always return success for setxattr call for lustre
3381 special xattr (currently only "trusted.lov").
3384 Frequency : rarely on clusters with both ia64+i386 clients
3386 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3387 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3388 clients are doing concurrent writes to the same file it is
3389 possible that the ia64 clients may LASSERT because the OST
3390 extent locks are not PAGE_SIZE aligned. Ensure that grown
3391 locks are always aligned on the request boundary.
3394 Frequency : specific use, occasional
3396 Description: Overwriting in use executable truncates on-disk binary image
3397 Details : If one node attempts to overwrite an executable in use by
3398 another node, we now correctly return ETXTBSY instead of
3399 truncating the file.
3401 Severity : enhancement
3403 Description: Async OSC create to avoid the blocking unnecessarily.
3404 Details : If a OST has no remain object, system will block on the creating
3405 when need to create a new object on this OST. Now, ways use
3406 pre-created objects when available, instead of blocking on an
3407 empty osc while others are not empty. If we must block, we block
3408 for the shortest possible period of time.
3413 Description: chmod on Lustre root is propagated to other clients
3414 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3415 invalid by the follow_mount time.
3418 Frequency : liblustre clients only
3420 Description: Race in 'instant cancel' lock handling could lead to such locks
3421 never to be granted in case of SMP MDS
3422 Details : Do not destroy not yet granted but cbpending locks in
3426 Frequency : replay/resend of open
3428 Description: non null lock assetion failure in mds_intent_policy
3429 Details : Trying to replay/resend lockless open requests resulted in
3430 mds_open() returning 0 with no lock. Now it sets a flag if
3431 a lock is going to be returned.
3433 Severity : enhancement
3435 Description: Checksum enhancements
3436 Details : New checksum enhancements allow for resending RPCs that failed
3439 Severity : enhancement
3441 Description: Tunables on number of dirty pages in cacche
3442 Details : Allow to set limit on number of dirty pages cached.
3447 Description: client crash on unmount - lock still has references
3448 Details : In some error handling cases it was possible to leak a lock
3449 reference on a client while accessing a file. This was not
3450 harmful to the client during operation, but would cause the
3451 client to crash when the filesystem is unmounted.
3454 Frequency : specific case, rare
3456 Description: ETXTBSY on mds though file not in use by client
3457 Details : ETXTBSY is no longer incorrectly returned when attempting to
3458 chmod or chown a directory that the user previously tried to
3459 execute or a currently-executing binary.
3462 Frequency : extremely rare except on liblustre-based clients
3464 Description: Lustre space not freed when files are deleted
3465 Details : Clean up open-unlinked files after client eviction. Previously
3466 the unlink was skipped and the files remained as orphans.
3471 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3472 Details : In some cases it was possible to send a blocking callback to
3473 a client doing a glimpse, even though that client didn't get
3474 a lock granted. When the glimpse lock is cancelled on the OST
3475 the freed lock is left on the waiting list and corrupted the list.
3478 Frequency : all core dumps
3480 Description: Broke core dumps to lustre
3481 Details : Negative dentry may be unhashed if parent does not have UPDATE
3482 lock, but some callers, e.g. do_coredump, expect dentry to be
3483 hashed after successful create, hash it in ll_create_it.
3485 ------------------------------------------------------------------------------
3487 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3489 * Support for kernels:
3490 2.6.9-42.0.2.EL (RHEL 4)
3491 2.6.5-7.276 (SLES 9)
3492 2.4.21-40.EL (RHEL 3)
3493 2.6.12.6 vanilla (kernel.org)
3497 Frequency : always on RHEL 3
3499 Description: Number of open files grows over time
3500 Details : The number of open files grows over time, whether or not
3501 Lustre is started. This was due to a filp leak introduced
3502 by one of our kernel patches.
3504 ------------------------------------------------------------------------------
3506 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3508 * Support for kernels:
3509 2.6.9-42.EL (RHEL 4)
3510 2.6.5-7.276 (SLES 9)
3511 2.4.21-40.EL (RHEL 3)
3512 2.6.12.6 vanilla (kernel.org)
3517 Bugzilla : 5719, 9635, 9792, 9684
3518 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3519 Details : If a server is under heavy load and cannot reply to new
3520 connection requests before the client resends the (re)connect,
3521 the connection handling code can behave badly if two service
3522 threads are concurrently handing separate (re)connections from
3523 the same client. Add better locking to the connection handling
3524 code, and ensure that only a single connection will be processed
3525 for a given client UUID, even if the lock is dropped.
3527 Severity : enhancement
3529 Description: add TCP zero-copy support to kernel
3530 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3531 sends if the hardware supports scatter-gather and checksumming.
3532 This allows socklnd to do client-write and server-read more
3533 efficiently and reduce CPU utilization from skbuf copying.
3536 Frequency : only if NFS exporting from client
3538 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3539 Details : If setting attributes on a file created under NFS that had
3540 never been opened it would be possible to oops the client
3541 if the file had no objects.
3544 Frequency : always for liblustre
3546 Description: liblustre client does MDS+OSTs setattr RPC for each write
3547 Details : When doing a write from a liblustre client, the client
3548 incorrectly issued an RPC to the MDS and each OST the file was
3549 striped over in order to update the timestamps. When writing
3550 with small chunks and many clients this could overwhelm the MDS
3551 with RPCs. In all cases it would slow down the write because
3552 these RPCs are unnecessary.
3554 Severity : enhancement
3556 Description: allow number of MDS service threads to be changed at module load
3557 Details : It is now possible to change the number of MDS service threads
3558 running. Adding "options mds mds_num_threads={N}" to the MDS's
3559 /etc/modprobe.conf will set the number of threads for the next
3560 time Lustre is restarted (assuming the "mds" module is also
3561 reloaded at that time). The default number of threads will
3562 stay the same, 32 for most systems.
3567 Description: OST crash if filesystem is unformatted or corrupt
3568 Details : If an OST is started on a device that has never been formatted
3569 or if the filesystem is corrupt and cannot even mount then the
3570 error handling cleanup routines would dereference a NULL pointer.
3575 Description: NULL pointer deref in llap_from_page.
3576 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3577 incorrect) mapping if the page was truncated/reclaimed while it was
3578 searched for. Check for this condition and skip such pages when
3579 doing readahead. Introduce extra check to llap_from_page() to
3580 verify page->mapping->host is non-NULL (so page is not anonymous).
3583 Frequency : Sometimes when using sys_sendfile
3585 Description: "page not covered by a lock" warnings from ll_readpage
3586 Details : sendfile called ll_readpage without right page locks present.
3587 Now we introduced ll_file_sendfile that does necessary locking
3588 around call to generic_file_sendfile() much like we do in
3592 Frequency : with certain MDS communication failures at client mount time
3594 Description: NULL pointer deref after failed client mount
3595 Details : a client connection request may delayed by the network layer
3596 and not be sent until after the PTLRPC layer has timed out the
3597 request. If the client fails the mount immediately it will try
3598 to clean up before the network times out the request. Add a
3599 reference from the request import to the obd device and delay
3600 the cleanup until the network drops the request.
3603 Frequency : occasionally during client (re)connect
3605 Description: assertion failure during client (re)connect
3606 Details : processing a client connection request may be delayed by the
3607 client or server longer than the client connect timeout. This
3608 causes the client to resend the connection request. If the
3609 original connection request is replied in this interval, the
3610 client may trip an assertion failure in ptlrpc_connect_interpret()
3611 which thought it would be the only running connect process.
3614 Frequency : only with obd_echo servers and clients that are rebooted
3616 Description: kernel BUG accessing uninitialized data structure
3617 Details : When running an obd_echo server it did not start the ping_evictor
3618 thread, and when a client was evicted an uninitialized data
3619 structure was accessed. Start the ping_evictor in the RPC
3620 service startup instead of the OBD startup.
3622 Severity : enhancement
3623 Bugzilla : 10193 (patchless)
3624 Description: Remove dependency on various unexported kernel interfaces.
3625 Details : No longer need reparent_to_init, exit_mm, exit_files,
3626 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3629 Frequency : rare (only users of deprecated and unsupported LDAP config)
3631 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3632 Details : LDAP apparently contains 'lustreName' attributes instead of
3633 'name'. A simple remapping of the name is sufficient.
3636 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3638 Description: waiting_locks_callback trips kernel BUG if client is evicted
3639 Details : Running with the dump_on_timeout debug flag turned on makes
3640 it possible that the waiting_locks_callback() can try to dump
3641 the Lustre kernel debug logs from an interrupt handler. Defer
3642 this log dumping to the expired_lock_main() thread.
3644 Severity : enhancement
3646 Description: Support NFS exporting on 2.6 kernels.
3647 Details : Implement non-rawops metadata methods for NFS server to use without
3648 changing NFS server code.
3651 Frequency : very rare (synthetic metadata workload only)
3653 Description: two racing renames might cause an MDS thread to deadlock
3654 Details : Running the "racer" program may cause one MDS thread to rename
3655 a file from being the source of a rename to being the target of
3656 a rename at exactly the same time that another thread is doing
3657 so, and the second thread has already enqueued these locks after
3658 doing a lookup of the target and is trying to relock them in
3659 order. Ensure that we don't try to re-lock the same resource.
3662 Frequency : only very large systems with liblustre clients
3664 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3665 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3666 all OSTs in parallel. This allows the eviction of stale liblustre
3667 clients to proceed much faster than if they were done in series,
3668 and also offers similar improvements for other set_info RPCs.
3673 Description: excessive CPU usage during initial read phase on client
3674 Details : During the initial read phase on a client, it would agressively
3675 retry readahead on the file, consuming too much CPU and impacting
3676 performance (since 1.4.5.8). Improve the readahead algorithm
3677 to avoid this, and also improve some other common cases (read
3678 of small files in particular, where "small" is files smaller than
3679 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3684 Description: MDS crash when receiving packet with unknown intent.
3685 Details : Do not LBUG in unknown intent case, just return -EFAULT
3687 Severity : enhancement
3688 Bugzilla : 9293, 9385
3689 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3690 Details : Do not serialize getattr (non-intent version) and statfs.
3693 Frequency : occasional, when OST network is overloaded/intermittent
3695 Description: client evicted by OST after bulk IO timeout
3696 Details : If a client sends a bulk IO request (read or write) the OST
3697 may evict the client if it is unresposive to its data GET/PUT
3698 request. This is incorrect if the network is overloaded (takes
3699 too long to transfer the RPC data) or dropped the OST GET/PUT
3700 request. There is no need to evict the client at all, since
3701 the pinger and/or lock callbacks will handle this, and the
3702 client can restart the bulk request.
3705 Frequency : Always when mmapping file with no objects
3707 Description: client crashes when mmapping file with no objects
3708 Details : Check that we actually have objects in a file before doing any
3709 operations on objects in ll_vm_open, ll_vm_close and
3715 Description: Request leak when working with deleted CWD
3716 Details : Introduce advanced request refcount tracking for requests
3717 referenced from lustre intent.
3719 Severity : Enhancement
3721 Description: Cache open file handles on client.
3722 Details : MDS now will return special lock along with openhandle, if
3723 requested and client is allowed to hold openhandle, even if unused,
3724 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3725 closing files for every read/write openration.
3727 Severity : Enhancement
3729 Description: Cache open negative dentries on client when possible.
3730 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3731 negative dentries on lock revocation.
3736 Description: Remounting a client read-only wasn't possible with a zconf mount
3737 Details : It wasn't possible to remount a client read-only with llmount.
3739 Severity : enhancement
3740 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3741 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3742 MPICH is included to add Lustre-specific ADIO interfaces.
3743 This is based closely on the UFS ADIO layer and only differs
3744 in file creation, in order to allow the OST striping to be set.
3745 This is user-contributed code and not supported by CFS.
3750 Description: extended inode attributes (immutable, append-only) work improperly
3751 when 2.4 and 2.6 kernels are used on client/server or vice versa
3752 Details : Introduce kernel-independent values for these flags.
3754 Severity : enhancement
3757 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3758 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3759 granularity. Now, Fractional MB granularity is be supported,
3760 this is very useful for low memory system.
3762 Severity : enhancement
3764 Description: Getattr by fid
3765 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3766 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3767 have locks and avoids getattr rpc requests for it on every stat.
3770 Frequency : Always, for filesystems larger than 2TB
3772 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3773 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3774 filesystems larger than 2TB to be mounted with mballoc enabled.
3777 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3779 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3780 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3781 filesystems larger than 2TB is mounted then the kernel will
3782 silently corrupt the start of the filesystem. CONFIG_LBD is
3783 enabled for all CFS-supported kernels, but the possibility of
3784 this happening with a modified kernel config exists.
3786 Severity : enhancement
3788 Description: add client O_DIRECT support for 2.6 kernels
3789 Details : It is now possible to do O_DIRECT reads and writes to files
3790 in the Lustre client mountpoint on 2.6 kernel clients.
3792 Severity : enhancement
3794 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3795 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3796 parallel, not waiting for response from every OST before sending
3797 a rpc to the next OST.
3802 Description: setattr vs write race when updating file timestamps
3803 Details : Client processes that update a file timestamp into the past
3804 right after writing to the file (e.g. tar) it is possible that
3805 the updated file modification time can be reset to the current
3806 time due to a race between processing the setattr and write RPC.
3808 Severity : enhancement
3810 Description: Bring 'lfs find' closer in line with regular Linux find.
3811 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3812 -print0 options and obtains all the needed info through the lustre
3815 Severity : enhancement
3817 Description: support up to 1024 configured devices on one node
3818 Details : change obd_dev array from statically allocated to dynamically
3819 allocated structs as they are first used to reduce memory usage
3824 Description: Flush dirty partially truncated pages during truncate
3825 Details : Immediatelly flush partially truncated pages in filter_setattr,
3826 this way we completely avoid having any pages in page cache on OST
3827 and can retire ugly workarounds during writes to flush such pages.
3832 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3833 Details : protect inode from truncation within vfs_unlink() context
3834 just take a reference before calling vfs_unlink() and release it
3835 when parent's i_sem is free.
3840 Description: last_id value checked outside lock on OST caused LASSERT failure
3841 Details : If there were multiple MDS->OST object precreate requests in
3842 flight, it was possible that the OST's last object id was checked
3843 outside a lock and incorrectly tripped an assertion. Move checks
3844 inside locks, and discard old precreate requests.
3847 Frequency : always, if extents are used on OSTs
3849 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3850 Details : OSTs using the extents format would not zero the high 16 bits of
3851 the index physical block number. This is not a problem for any
3852 OST filesystems smaller than 16TB, and no kernels support ext3
3853 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3854 new/modified files) and can be fixed for existing filesystems
3855 with e2fsprogs-1.39-cfs1.
3860 Description: import connection selection may be incorrect if timer wraps
3861 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3862 import connections to be ignored if the 32-bit jiffies counter
3863 wraps. Use a 64-bit jiffies counter.
3866 Frequency : very large clusters immediately after boot
3868 Description: LNET request buffers exhausted under heavy short-term load
3869 Details : If a large number of client requests are generated on a service
3870 that has previously never seen so many requests it is possible
3871 that the request buffer growth cannot keep up with the spike in
3872 demand. Instead of dropping incoming requests, they are held in
3873 the LND until the RPC service can accept more requests.
3876 Frequency : Sometimes during replay
3878 Description: Assertion failure in ll_local_open after replay.
3879 Details : If replay happened on an open request reply before we were able
3880 to set replay handler, reply will become not swabbed tripping the
3881 assertion in ll_local_open. Now we set the handler right after
3882 recognising of open request
3885 Frequency : very rare
3887 Description: kernel reports "badness in vsnprintf"
3888 Details : Reading from the "recovery_status" /proc file in small chunks
3889 may cause a negative length in lprocfs_obd_rd_recovery_status()
3890 call to vsnprintf() (which is otherwise harmless). Exit early
3891 if there is no more space in the output buffer.
3893 Severity : enhancement
3895 Description: clear OBD RPC statistics by writing to them
3896 Details : It is now possible to clear the OBD RPC statistics by writing
3897 to the "stats" file.
3902 Description: Client mtime is not the same on different clients after utimes
3903 Details : In some cases, the client was using the utimes() syscall on
3904 a file cached on another node. The clients now validate the
3905 ctime from the MDS + OSTs to determine which one is right.
3910 Description: Inability to activate failout mode
3911 Details : lconf script incorrectly assumed that in pythong string's numeric
3912 value is used in comparisons.
3915 Frequency : always with multiple stripes per file
3917 Description: Inefficient object allocation for mutli-stripe files
3918 Details : When selecting which OSTs to stripe files over, for files with
3919 a stripe count that divides evenly into the number of OSTs,
3920 the MDS is always picking the same starting OST for each file.
3921 Return the OST selection heuristic to the original design.
3926 Description: mount failures may take full timeout to return an error
3927 Details : Under some heavy load conditions it is possible that a
3928 failed mount can wait for the full obd_timeout interval,
3929 possibly several minutes, before reporting an error.
3930 Instead return an error as soon as the status is known.
3932 Frequency : quota enabled and large files being deleted
3934 Description: releasing more than 4GB of quota at once hangs OST
3935 Details : If a user deletes more than 4GB of files on a single OST it
3936 will cause the OST to spin in an infinite loop. Release
3937 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3942 Description: statfs data retrieved from /proc may be stale or zero
3943 Details : When reading per-device statfs data from /proc, in the
3944 {kbytes,files}_{total,free,avail} files, it may appear
3945 as zero or be out of date.
3948 Frequency : systems with MD RAID1 external journal devices
3950 Description: lconf's call to blkid is confused by RAID1 journal devices
3951 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3952 of returning all block devices that match the journal UUID.
3955 Frequency : always, for aggregate stripe size over 4GB
3957 Description: assertion fails when trying to use 4GB stripe size
3958 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3959 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3962 Frequency : always on ppc64
3964 Description: the first write on an ext3 filesystem with mballoc got stuck
3965 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3966 perform endianness conversion.
3968 ------------------------------------------------------------------------------
3970 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3972 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3973 INTEROPERATE with older versions automatically. Please read the
3974 user documentation before upgrading any part of a live system.
3975 * WARNING: Lustre networking configuration changes are required with
3976 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3979 * Support for kernels:
3980 2.6.9-22.0.2.EL (RHEL 4)
3981 2.6.5-7.244 (SLES 9)
3982 2.6.12.6 vanilla (kernel.org)
3985 Severity : enhancement
3986 Bugzilla : 7981/8208
3987 Description: Introduced Lustre Networking (LNET)
3988 Details : LNET is new networking infrastructure for Lustre, it includes
3989 a reorganized network configuration mode (see the user
3990 documentation for full details) as well as support for routing
3991 between different network fabrics. Lustre Networking Devices
3992 (LNDs) for the supported network fabrics have also been
3993 created for this new infrastructure.
3995 Severity : enhancement
3996 Description: Introduced Access control lists
3997 Details : clients can set ACLs on files and directories in order to have
3998 more fine-grained permissions than the standard Unix UGO+RWX.
3999 The MDS must be started with the "-o acl" mount option.
4001 Severity : enhancement
4002 Description: Introduced filesystem quotas
4003 Details : Administrators may now establish per-user quotas on the
4006 Severity : enhancement
4008 Description: Configuration change for the XT3
4009 The PTLLND is now used to run Lustre over Portals on the XT3
4010 The configure option(s) --with-cray-portals are no longer used.
4011 Rather --with-portals=<path-to-portals-includes> is used to
4012 enable building on the XT3. In addition to enable XT3 specific
4013 features the option --enable-cray-xt3 must be used.
4018 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4019 Details : A race between allocating a new llcd and re-getting the llcd_lock
4020 allowed another thread to grab newly-allocated llcd.
4022 Severity : enhancement
4024 Description: 2.6 OST async journal commit and locking fix to improve performance
4025 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4026 2.6 kernels are now async as they already were in 2.4 kernels so
4027 that they can commit concurrently with the network bulk transfer.
4028 For block-allocated files the filter allocation semaphore is held
4029 to avoid filesystem fragmentation during allocation. BKL lock
4030 removed for 2.6 xattr operations where it is no longer needed.
4035 Description: lconf incorrectly determined whether two IP networks could talk
4036 Details : In some more complicated routing and multiple-network
4037 configurations, lconf will avoid trying to make a network
4038 connection to a disjoint part of the IP space. It was doing the
4039 math incorrectly for one set of cases.
4044 Description: Fix for potential infinite loop processing records in an llog.
4045 Details : If an llog record is corrupted/zeroed, it is possible to loop
4046 forever in llog_process(). Validate the llog record length
4047 and skip the remainder of the block on error.
4050 Frequency : occasional (liblustre only)
4052 Description: liblustre could not open files whose last component is a symlink
4053 Details : sysio_path_walk() would incorrectly pass the open intent to
4054 intermediate path components.
4057 Frequency : rare (liblustre only with non-standard tuning)
4058 Bugzilla : 7201 (7350)
4059 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4060 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4061 after it has been released, only do this for referenced locks
4063 Severity : enhancement
4065 Description: specifying an (invalid) directory default stripe_size of -1
4066 would reset the directory default striping
4067 Details : stripe_size -1 was used internally to signal directory stripe
4068 removal, now use "all default" to signal dir stripe removal
4069 as a directory striping of "all default" is not useful
4072 Frequency : common for large clusters running liblustre clients
4074 Description: doing an ls when liblustre clients are running is slow
4075 Details : sending a glimpse AST to a liblustre client waits for every AST
4076 to time out, as liblustre clients will not respond. Since they
4077 cannot cache data we refresh the OST lock LVB from disk instead.
4079 Severity : enhancement
4081 Description: doing an ls at the same time as file IO can be slow
4082 Details : enqueue and other "small" requests can be blocked behind many
4083 large IO requests. Create a new OST IO portal for non-IO
4084 requests so they can be processed faster.
4087 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4089 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4090 created file with mode 000
4091 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4092 allow an ftruncate() as a non-root user to a file with mode 000.
4093 Lustre now respects this flag to disable mode checks when
4094 truncating a file owned by the user
4097 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4100 Description: Revoking locks from clients that went dead or catatonic might take
4102 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4103 cancellation of such locks instant on servers without waiting for
4104 any reply from clients. Clients drops these locks when cancel
4105 notification from server is received without replying.
4108 Frequency : liblustre-only, when liblustre client dies or becomes busy
4110 Description: Doing ls on Linux clients can take a long time with active
4112 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4113 granting such locks to it in the first place if possible. Locks
4114 are taken by proxy on the OST during the read or write and
4115 dropped immediately afterward. Add connect flags handling, do
4116 not grant locks to liblustre clients for glimpse ASTs.
4118 Severity : enhancement
4120 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4121 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4122 sequential reads and behaves badly with applications doing random
4123 reads. Improve it by reading ahead at least read region, and
4124 avoiding excessive large RPC for small reads.
4126 Severity : enhancement
4128 Description: Creating more than 1000 files for a single job may cause a load
4129 imbalance on the OSTs if there are also a large number of OSTs.
4130 Details : qos_prep_create() uses an OST index reseed value that is an
4131 even multiple of the number of available OSTs so that if the
4132 reseed happens in the middle of the object allocation it will
4133 still utilize the OSTs as uniformly as possible.
4138 Description: OST or MDS may oops in ping_evictor_main()
4139 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4140 but doesn't restart at beginning of obd_exports_timed list
4143 Severity : enhancement
4145 Description: improve by-nid export eviction on the MDS and OST
4146 Details : allow multiple exports with the same NID to be evicted at one
4147 time without re-searching the exports list.
4150 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4152 Description: MDS may oops in groups_free()
4153 Details : in rare race conditions a newly allocated group_info struct is
4154 freed again, and this can be NULL. The 2.4 compatibility code
4155 for groups_free() checked for a NULL pointer, but 2.6 did not.
4158 Frequency : common for liblustre clients doing little filesystem IO
4159 Bugzilla : 9352, 7313
4160 Description: server may evict liblustre clients accessing contended locks
4161 Details : if a client is granted a lock or receives a completion AST
4162 with a blocking AST already set it would not reply to the AST
4163 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4166 Frequency : lfs setstripe, only systems with more than 160 OSTs
4168 Description: unable to set striping with a starting offset beyond OST 160
4169 Details : llapi_create_file() incorrectly limited the starting stripe
4170 index to the maximum single-file stripe count.
4173 Frequency : LDAP users only
4175 Description: lconf did not handle in-kernel recovery with LDAP properly
4176 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4178 Severity : enhancement
4180 Description: bind OST threads to NUMA nodes to improve performance
4181 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4182 node and do their allocations there to localize memory access
4184 Severity : enhancement
4186 Description: llmount can determine client NID directly from Myrinet (GM)
4187 Details : the client NID code from gmnalnid was moved directly into
4188 llmount, removing the need to use this or specifying the
4189 client NID explicitly when mounting GM clients with zeroconf
4192 Frequency : if client is started with down MDS
4194 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4195 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4196 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4197 ptlrpc_import_delayed_req() because we didn't check if the
4198 request was interrupted, and we also didn't break out of the
4199 event loop if there was no timeout
4204 Description: data loss during non-page-aligned writes to a single file from
4205 both multiple nodes and multiple threads on one node at same time
4206 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4207 inconsistency led to false short-reads, that were cached and later
4208 used by ->prepare_write() to fill in partially written page,
4209 leading to data loss.
4212 Frequency : always, if lconf --abort_recovery used
4214 Description: lconf --abort_recovery fails with 'Operation not supported'
4215 Details : lconf was attempting to abort recovery on the MDT device and not
4218 Severity : enhancement
4220 Description: remove cleanup logs
4221 Details : replace lconf-generated cleanup logs with lustre internal
4222 cleanup routines. Eliminates the need for client-cleanup and
4225 Severity : enhancement
4227 Description: add support for EAs (user and system) on lustre filesystems
4228 Details : it is now possible to store extended attributes in the Lustre
4229 client filesystem, and with the user_xattr mount option it
4230 is possible to allow users to store EAs on their files also
4232 Severity : enhancement
4234 Description: Add possibility (config option) to show minimal available OST free
4236 Details : When compiled with --enable-mindf configure option, statfs(2)
4237 (and so, df) will return least minimal free space available from
4238 all OSTs as amount of free space on FS, instead of summary of
4239 free spaces of all OSTs.
4241 Severity : enhancement
4243 Description: do not expand extent locks acquired on OST-side
4244 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4245 by server: they are not cached anyway.
4248 Frequency : when mmap is used/binaries executed from Lustre
4250 Description: Unmmap pages before throwing them away from read cache.
4251 Details : llap_shrink cache now attempts to unmap pages before discarding
4252 them (if unmapping failed - do not discard). SLES9 kernel has
4253 extra checks that trigger if this unmapping is not done first.
4258 Description: lconf didn't resolve symlinks before checking to see whether a
4259 given mountpoint was already in use
4262 Frequency : when migrating failover services
4263 Bugzilla : 6395, 9514
4264 Description: When migrating a subset of services from a node (e.g. failback
4265 from a failover service node) the remaining services would
4266 time out and evict clients.
4267 Details : lconf --force (implied by --failover) sets the global obd_timeout
4268 to 5 seconds in order to quickly disconnect, but this caused
4269 other RPCs to time out too quickly. Do not change the global
4270 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4272 Severity : enhancement
4273 Frequency : if MDS is started with down OST
4274 Bugzilla : 9439,5706
4275 Description: Allow startup/shutdown of an MDS without depending on the
4276 availability of the OSTs.
4277 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4278 Add appropriate locking and lov-osc refcounts for safe
4279 cleaning. Add osc abort_inflight calls in case the
4280 synchronize never started.
4283 Frequency : occasional (Cray XT3 only)
4285 Description: root not authorized to access files in CRAY_PORTALS environment
4286 Details : The client process capabilities were not honoured on the MDS in
4287 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4288 been accessed by an authorized user then root was able to access
4289 the file on the local client also. The root user capabilities
4290 are now allowed on the MDS, as this environment has secure UID.
4293 Frequency : occasional
4295 Description: ldiskfs "too long searching" message happens too often
4296 Details : A debugging message (otherwise harmless) prints too often on
4297 the OST console. This has been reduced to only happen when
4298 there are fragmentation problems on the filesystem.
4303 Description: Division by zero in statfs when all OSCs are inactive
4304 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4305 lov_statfs divides by value returned by lov_get_stripecnt().
4309 Bugzilla : 9489, 3273
4310 Description: First write from each client to each OST was only 4kB in size,
4311 to initialize client writeback cache, which caused sub-optimal
4312 RPCs and poor layout on disk for the first writen file.
4313 Details : Clients now request an initial cache grant at (re)connect time
4314 and so that they can start streaming writes to the cache right
4315 away and always do full-sized RPCs if there is enough data.
4316 If the OST is rebooted the client also re-establishes its grant
4317 so that client cached writes will be honoured under the grant.
4322 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4323 Details : Now I/O RPCs go to different portal number and (presumably) fast
4324 lock requests (and glimses) and other RPCs get their own service
4325 threads pool that should be able to service those RPCs
4328 Severity : enhancement
4330 Description: Ability to exchange lustre version between client and servers and
4331 issue warnings at client side if client is too old. Also for
4332 liblustre clients there is ability to refuse connection of too old
4334 Details : New 'version' field is added to connect data structure that is
4335 filled with version info. That info is later checked by server and
4339 Frequency : rare, liblustre only.
4340 Bugzilla : 9296, 9581
4341 Description: Two simultaneous writes from liblustre at offset within same page
4342 might proceed at the same time overwriting eachother with stale
4344 Details : I/O lock withing llu_file_prwv was released too early, before data
4345 actually was hitting the wire. Extended lock-holding time until
4346 server acknowledges receiving data.
4349 Frequency : extremely rare. Never observed in practice.
4351 Description: avoid generating lustre_handle cookie of 0.
4352 Details : class_handle_hash() generates handle cookies by incrementing
4353 global counter, and can hit 0 occasionaly (this is unlikely, but
4354 not impossible, because initial value of cookie counter is
4355 selected randonly). Value of 0 is used as a sentinel meaning
4356 "unassigned handle" --- avoid it. Also coalesce two critical
4357 sections in this function into one.
4359 Severity : enhancement
4361 Description: allow liblustre clients to delegate truncate locking to OST
4362 Details : To avoid overhead of locking, liblustre client instructs OST to
4363 take extent lock in ost_punch() on client's behalf. New connection
4364 flag is added to handle backward compatibility.
4366 Severity : enhancement
4367 Bugzilla : 4928, 7341, 9758
4368 Description: allow number of OST service threads to be specified
4369 Details : a module parameter allows the number of OST service threads
4370 to be specified via "options ost ost_num_threads={N}" in the
4371 OSS's /etc/modules.conf or /etc/modprobe.conf.
4375 Bugzilla : 6146, 9635, 9895
4376 Description: servers crash with bad pointer in target_handle_connect()
4377 Details : In rare cases when a client is reconnecting it was possible that
4378 the connection request was the last reference for that export.
4379 We would temporarily drop the export reference and get a new
4380 one, but this may have been the last reference and the export
4381 was just destroyed. Get new reference before dropping old one.
4383 Severity : enhancement
4384 Frequency : if client is started with failover MDS
4386 Description: Allow multiple MDS hostnames in the mount command
4387 Details : Try to read the configuration from all specified MDS
4388 hostnames during a client mount in case the "primary"
4391 Severity : enhancement
4393 Description: Stop sending data to evicted clients as soon as possible.
4394 Details : Check if the client we are about to send or are sending data to
4395 was evicted already. (Check is done every second of waiting,
4396 for which l_wait_event interface was extended to allow checking
4397 of exit condition at specified intervals).
4400 Frequency : rare, normally only when NFS exporting is done from client
4402 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4404 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4407 Frequency : always, if srand() is called before liblustre initialization
4409 Description: Liblustre uses system PRNG disturbing its usage by user application
4410 Details : Introduce internal to lustre fast and high-quality PRNG for
4411 lustre usage and make liblustre and some other places in generic
4412 lustre code to use it.
4414 Severity : enhancement
4415 Bugzilla : 9477, 9557, 9870
4416 Description: Verify that the MDS configuration logs are updated when xml is
4417 Details : Check if the .xml configuration logs are newer than the config
4418 logs stored on the MDS and report an error if this is the case.
4419 Request --write-conf, or allow starting with --old_conf.
4421 Severity : enhancement
4423 Description: Handle symlinks in the path when checking if Lustre is mounted.
4424 Details : Resolve intermediate symlinks when checking if a client has
4425 mounted a filesystem to avoid duplicate client mounts.
4430 Description: lconf can hit an error exception but still return success.
4431 Details : The lconf command catches the Command error exception at the top
4432 level script context and will exit with the associated exit
4433 status, but doesn't ensure that this exit status is non-zero.
4438 Description: failure of ptlrpc thread startup can cause oops
4439 Details : Starting a ptlrpc service thread can fail if there are a large
4440 number of threads or the server memory is very fragmented.
4441 Handle this without oopsing.
4444 Frequency : always, only if liblustre and non-default acceptor port was used
4446 Description: liblustre cannot connect to servers with non-default acceptor port
4447 Details : tcpnal_set_default_params() was not called and was therefore
4448 ignoring the environment varaible TCPNAL_PORT, as well as other
4449 TCPNAL_ environment variables
4454 Description: two objects could be created on the same OST for a single file
4455 Details : If an OST is down, in some cases it was possible to create two
4456 objects on a single OST for a single file. No problems other
4457 than potential performance impact and spurious error messages.
4461 Bugzilla : 5681, 9562
4462 Description: Client may oops in ll_unhash_aliases
4463 Details : Client dcache may become inconsistent in race condition.
4464 In some cases "getcwd" can fail if the current directory is
4470 Description: Inode refcounting problems in NFS export code
4471 Details : link_raw functions used to call d_instantiate without obtaining
4472 extra inode reference first.
4476 Bugzilla : 9942, 9903
4477 Description: Referencing freed requests leading to crash, memleaks with NFS.
4478 Details : We used to require that call to ll_revalidate_it was always
4479 followed by ll_lookup_it. Also with revalidate_special() it is
4480 possible to call ll_revalidate_it() twice for the same dentry
4481 even if first occurence returned success. This fix changes semantic
4482 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4483 reference on a request referred from the intent.
4484 ll_intent_release() then releases such a request.
4487 Frequency : rare, normally benchmark loads only
4489 Description: unlinked inodes were kept in memory on the client
4490 Details : If a client is repeatedly creating and unlinking files it
4491 can accumulate a lot of stale inodes in the inode slab cache.
4492 If there is no other client load running this can cause the
4493 client node to run out of memory. Instead flush old inodes
4494 from client cache that have the same inode number as a new inode.
4497 Frequency : SLES9 2.6.5 kernel and long filenames only
4498 Bugzilla : 9969, 10379
4499 Description: utime reports stale NFS file handle
4500 Details : SLES9 uses out-of-dentry names in some cases, which confused
4501 the lustre dentry revalidation. Change it to always use the
4505 Frequency : rare, unless heavy write-truncate concurrency is continuous
4506 Bugzilla : 4180, 6984, 7171, 9963, 9331
4507 Description: OST becomes very slow and/or deadlocked during object unlink
4508 Details : filter_destroy() was holding onto the parent directory lock
4509 while truncating+unlinking objects. For very large objects this
4510 may block other threads for a long time and slow overall OST
4511 responsiveness. It may also be possible to get a lock ordering
4512 deadlock in this case, or run out of journal credits because of
4513 the combined truncate+unlink. Solution is to do object truncate
4514 first in one transaction without parent lock, and then do the
4515 final unlink in a new transaction with the parent lock. This
4516 reduces the lock hold time dramatically.
4519 Frequency : rare, 2.4 kernels only
4521 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4522 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4523 before unmounting their filesystem in order to avoid deadlock.
4524 The kernel_locked() function in 2.4 kernels only checks whether
4525 the kernel lock is held, not whether it is this process that is
4526 holding it as 2.6 kernels do.
4531 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4532 Details : The client ptlrpc code may be trying to reconnect to a down
4533 server before a previous connection attempt has timed out.
4534 Increase the reconnect interval to be longer than the connection
4535 timeout interval to avoid sending duplicate connections to
4539 Frequency : echo_client brw_test command
4541 Description: fix echo_client to work with OST preallocated code
4542 Details : OST preallocation code (5137) didn't take echo_client IO path
4543 into account: echo_client calls filter methods outside of any
4544 OST thread and, hence, there is no per-thread preallocated
4545 pages and buffers to use. Solution: hijack pga pages for IO. As
4546 a byproduct, this avoids unnecessary data copying.
4550 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4551 Description: Client can oops in mdc_commit_close() after open replay
4552 Details : It was possible for the MDS to return an open request with no
4553 transaction number in mds_finish_transno() if the client was
4554 evicted, but without actually returning an error. Clients
4555 would later try to replay that open and may trip an assertion
4556 Simplify the client close codepath, and always return an error
4557 from the MDS in case the open is not successful.
4560 Frequency : rare, 2.6 OSTs only
4562 Description: OST may deadlock under high load on fragmented files
4563 Details : If there was a heavy load and highly-fragmented OST filesystems
4564 it was possible to have all the OST threads deadlock waiting on
4565 allocation of biovecs, because the biovecs were not released
4566 until the entire RPC IO was completed. Instead, release biovecs
4567 as soon as they are complete to ensure forward IO progress.
4569 Severity : enhancement
4571 Description: Support for specifying external journal device at mount
4572 Details : If an OST or MDS device is formatted with an external journal
4573 device, this device major/minor is stored in the ext3 superblock
4574 and may not be valid for failover. Allow detecting and
4575 specifying the external journal at mount time.
4580 Description: Mounting an MDS with pending unlinked files may cause oops
4581 Details : target_finish_recovery() calls mds_postrecov() which returned
4582 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4583 considers this an error and immediately begins cleaning up the
4584 lov, just after starting the mds_lov process
4586 Severity : enhancement
4588 Description: Implement 'lfs df' to report actual free space on per-OST basis
4589 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4590 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4591 usage of objects; '-h' to report in human readable format.
4593 ------------------------------------------------------------------------------
4595 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4602 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4603 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4604 at startup during orphan unlinking. Instead, skip buddy update
4605 before setup, as it will be regenerated after recovery is complete.
4610 Description: If an OST is inactive, its locks might reference stale inodes.
4611 Details : lov_change_cbdata() must iterate over all namespaces, even if
4612 they are inactive to clear inode references from the lock.
4614 Severity : enhancement
4615 Frequency : occasional, if non-standard max_dirty_mb used
4617 Description: Client will block write RPCs if not enough grant
4618 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4619 then the client will block writes while waiting for another RPC
4620 to complete instead of consuming its dirty limit. With change
4621 we get improved performance when max_dirty_mb is small.
4623 Severity : enhancement
4624 Bugzilla : 3389, 6253
4625 Description: Add support for supplementary groups on the MDS.
4626 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4627 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4628 lookups for user supplementary groups into a cache.
4632 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4633 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4634 for liblustre O_EXCL is enforced by MDS.
4638 Description: readdir never returns NULL in liblustre.
4639 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4640 dentry in struct dirent.
4644 Frequency : liblustre only, depends on application IO pattern
4645 Description: liblustre clients evicted if not contacting servers
4646 Details : Don't put liblustre clients into the ping_evictor list, so
4647 they will not be evicted by the pinger ever.
4649 Severity : enhancement
4651 Description: Add ability to evict clients by NID from MDS.
4652 Details : By echoing "nid:$NID" string into
4653 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4654 $NID would be instantly evicted from this MDS and from all active
4655 OSTs connected to it.
4659 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4660 Details : lookup_it_finish() used to query file size from OSTs that was not
4665 Description: service threads change working directory to that of init
4666 Details : Starting lustre service threads may pin the working directory
4667 of the parent thread, making that filesystem busy. Threads
4668 now change to the working directory of init to avoid this.
4672 Frequency : during shutdown only
4673 Description: shutdown with a failed MDS or OST can cause unmount to hang
4674 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4679 Frequency : 2.6 only
4680 Description: chmod/chown may include an extra supplementary group
4681 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4682 supplementary group and if none is specified this is used.
4685 Bugzilla : 5479 (6816)
4687 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4688 Details : If lookup is in progress on a file that is unlinked we might try
4689 to revalidate the inode and fail in revalidate after lookup is
4690 complete and ll_file_open() enqueues the open again but
4691 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4694 Frequency : always, if lconf --abort_recovery used
4696 Description: lconf --abort_recovery fails with 'Operation not supported'
4697 Details : lconf was attempting to abort recovery on the MDT device and not
4700 ------------------------------------------------------------------------------
4702 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4707 Frequency : rare (only unsupported configurations with a node running as an
4709 Bugzilla : 6514, 5137
4710 Description: Mounting a Lustre file system on a node running as an OST could
4712 Details : OSTs now preallocates memory needed to write out data at
4713 startup, instead of when needed, to avoid having to
4714 allocate memory in possibly low memory situations.
4715 Specifically, if the file system is mounted on on OST,
4716 memory pressure could force it to try to write out data,
4717 which it needed to allocate memory to do. Due to the low
4718 memory, it would be unable to do so and the node would
4719 become unresponsive.
4721 Severity : enhancement
4723 Description: Addition of lconf --service command line option
4724 Details : lconf now accepts a '--service <arg>' option, which is
4725 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4727 Severity : enhancement
4729 Description: Failover mode is now the default for OSTs.
4730 Details : By default, OSTs will now run in failover mode. To return to
4731 the old behaviour, add '--failout' to the lmc line for OSTs.
4733 Severity : enhancement
4735 Description: Health checks are now provided for MDS and OSTs
4736 Details : Additional detailed health check information on MSD and OSTs
4737 is now provided through the procfs health_check value.
4740 Frequency : occasional, depends on IO load
4742 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4743 after numerous create/delete cycles
4744 Details : The ext3 inode allocation policy would not allocate new inodes
4745 very well on the OSTs because there are no new directories
4746 being created. Instead we look for groups with free space if
4747 the parent directories are nearly full.
4752 Description: Network or server problems during mount may cause partially
4753 mounted clients instead of returning an error.
4754 Details : The config llog parsing code may overwrite the error return
4755 code during mount error handling, returning success instead
4761 Description: MDS can fail to allocate large reply buffers
4762 Details : After long uptimes the MDS can fail to allocate large reply
4763 buffers (e.g. zconf client mount config records) due to memory
4764 fragmentation or consumption by the buffer cache. Preallocate
4765 some large reply buffers so that these replies can be sent even
4766 under memory pressure.
4770 Frequency : rare (liblustre)
4771 Description: fsx running with liblustre complained that using truncate() to
4772 extend the file doesn't work. This patch corrects that issue.
4773 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4774 ATTR_SIZE and lsm use in llu_setattr_raw.
4778 Frequency : rare, only 2.6 kernels
4779 Description: Unusual file access patterns on the MDS may result in inode
4780 data being lost in very rare circumstances.
4781 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4782 code in the 2.6 kernel caused the inode and/or EA data not to
4783 be read from disk, causing single-file corruption.
4787 Frequency : rare, only 2.6 filesystems using extents
4788 Description: Heavy concurrent write and delete load may cause data corruption.
4789 Details : It was possible under high-load situations to have an extent
4790 metadata block in the block device cache from a just-unlinked
4791 file overwrite a newly-allocated data block. We now unmap any
4792 metadata buffers that alias just-allocated data blocks.
4796 Frequency : filesystems with default stripe_count larger than 77
4797 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4798 Details : lconf specifies an inode size of 4096 bytes when the default
4799 stripe_count is larger than 77. This conflicts with the default
4800 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4801 this case to avoid pinning too much memory for large EAs.
4803 ------------------------------------------------------------------------------
4805 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4810 Frequency : rare (extremely heavy IO load with hundreds of clients)
4812 Description: Client is evicted, gets IO error writing to file
4813 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4814 caused clients to be evicted for AST timeouts. The fixes in
4815 bug 5192 mean we no longer need to have such short AST timeouts
4816 so ldlm_timeout has been increased.
4819 Frequency : occasional during --force or --failover shutdown under load
4820 Bugzilla : 5949, 4834
4821 Description: Server oops/LBUG if stopped with --force or --failover under load
4822 Details : a collection of import/export refcount and cleanup ordering
4823 issues fixed for safer force cleanup
4826 Frequency : only filesystems larger than 120 OSTs
4827 Bugzilla : 5990, 6223
4828 Description: lfs getstripe would oops on a very large filesystem
4829 Details : lov_getconfig used kfree on vmalloc'd memory
4832 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4833 Bugzilla : 6242, 6243
4834 Description: reading from files that had been truncated to a non-zero size
4835 but never opened returned no data
4836 Details : ll_file_read() reads zeros from no-object files to EOF
4841 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4842 Details : There's little harm in aborting MDS/OSS recovery and letting it
4843 try again, so I removed the LASSERT and return an error instead.
4845 Severity : enhancement
4847 Description: New debugging infrastructure for tracking down data corruption
4848 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4849 (b) cover more of the client-side code path, and (c) try to narrow
4850 down where problems occurred
4854 Bugzilla : 3819, 4364, 4397, 6313
4855 Description: Racing close and eviction MDS could cause assertion in mds_close
4856 Details : It was possible to get multiple mfd references during close and
4857 client eviction, leading to one thread referencing a freed mfd.
4859 Severity: : enhancement
4860 Bugzilla : 3262, 6359
4861 Description: Attempts to reconnect to servers are now more aggressive.
4862 Details : This builds on the enhanced upcall-less recovery that was added
4863 in 1.4.2. When trying to reconnect to servers, clients will
4864 now try each server in the failover group every 10 seconds. By
4865 default, clients would previously try one server every 25 seconds.
4870 Description: After recovery, certain operations trigger a failed
4871 assertion on a client.
4872 Details : Failing over an mds, using lconf -d --failover, while a
4873 client was doing a readdir() call would cause the client to
4874 LBUG after recovery completed and the readdir() was resent.
4876 Severity : enhancement
4878 Description: Default groups are now added by lconf
4879 Details : You can now run lconf --group <servicename> without having to
4880 manually add groups with lmc.
4883 Frequency : occasional
4885 Description: Nodes with an elan id of 0 trigger a failed assertion
4888 Frequency : always when accessing e.g. tty/console device nodes
4890 Description: tty and some other devices nodes cannot be used on lustre
4891 Details : file's private_data field is used by device data and lustre
4892 values in there got lost. New field was added to struct file to
4893 store fs-specific private data.
4896 Frequency : when exporting Lustre via NFS
4898 Description: NFSD failed occasionally when looking up a path component
4899 Details : NFSD is looking up ".." which was broken in ext3 directories
4900 that had grown large enough to become hashed.
4903 Frequency : Clusters with multiple interfaces not on the same subnet
4905 Description: Nodes will repeatedly try to reconnect to an interface which it
4906 cannot reach and report an error to the log.
4907 Details : Extra peer list entries will be created by lconf with some peers
4908 unreachable. lconf now validates the peer before adding it.
4911 Frequency : Only if a default stripe is set on the filesystem root.
4913 Description: Setting a default stripe on the filesystem root prevented the
4914 filesystem from being remounted.
4915 Details : The client was sending extra request flags in the root getattr
4916 request and did not allocate a reply buffer for the dir EA.
4919 Frequency : occasional, higher if lots of files are accessed by one client
4920 Bugzilla : 6159, 6097
4921 Description: Client trips assertion regarding lsm mismatch/magic
4922 Details : While revalidating inodes the VFS looks up inodes with ifind()
4923 and in rare cases can find an inode that is being freed.
4924 The ll_test_inode() code will free the lsm during ifind()
4925 when it finds an existing inode and then the VFS later attaches
4926 this free lsm to a new inode.
4930 Bugzilla : 6422, 7030
4931 Description: MDS deadlock between mkdir and client eviction
4932 Details : Creating a new file via mkdir or mknod (starting a transaction
4933 and getting the ns lock) can deadlock with client eviction
4934 (gets ns lock and trying to finish a synchronous transaction).
4937 Frequency : occasional
4938 Description: While starting a server, the fsfilt_ext3 module could not be
4940 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4941 kernels. Previously, lconf would still use the ext3 name
4942 when trying to load modules. Now, it will correctly use
4943 ext3 on 2.4 and ldiskfs on 2.6.
4945 Severity : enhancement
4946 Description: The default stripe count has been changed to 1
4947 Details : The interpretation of the default stripe count (0, to lfs
4948 or lmc) has been changed to mean striping across a single
4949 OST, rather than all available. For general usage we have
4950 found a stripe count of 1 or 2 works best.
4952 Severity : enhancement
4953 Description: Add support for compiling against Cray portals.
4954 Details : Conditional compiling for some areas that are different
4958 Frequency : occasional
4959 Bugzilla : 6409, 6834
4960 Description: Creating files with an explicit stripe count may lead to
4961 a failed assertion on the MDS
4962 Details : If some OSTs are full or unavailable, creating files may
4963 trigger a failed assertion on the MDS. Now, Lustre will
4964 try to use other servers or return an error to the
4968 Frequency : occasional
4970 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4971 caused lock timeout during readahead (since 1.4.2).
4972 Details : Processes doing readahead might match a lock that hasn't been
4973 granted yet if there are overlapping and conflicting lock
4974 requests. The readahead process waits on ungranted lock
4975 (original lock is CBPENDING), while OST waits for that process
4976 to cancel CBPENDING read lock and eventually evicts client.
4978 Severity : enhancement
4980 Description: Initial enabling of flock support for clients
4981 Details : Implements fcntl advisory locking and file status functions.
4982 This feature is provided as an optional mount flag (default
4983 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4984 locking are implemented yet, and those that are are not guaranteed
4985 to be completely correct in production environments.
4986 mount -t lustre -o [flock|noflock] ...
4989 Frequency : occasional
4991 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4992 assertion in the ext3 JBD (journaling) layer.
4993 Details : The b_committed_data struct is protected by the big kernel lock
4994 in 2.4 kernels, serializing journal_commit_transaction() and
4995 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4996 access to this struct. In 2.6 kernels there is finer grained
4997 locking to improve SMP performance of the JBD layer.
5001 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5006 Description: merge and cleanup kernel patches.
5009 -----------------------------------------------------------------------------
5011 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5013 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5014 versions of Lustre. Please update all servers and clients to
5015 version 1.4.2 or later at the same time. You must also run
5016 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5017 to update the configuration logs.
5019 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5020 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5021 - swab llog records of type '0' so we get proper header size/idx (5861)
5022 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5023 - fix rename of one directory over another leaking an inode (5953)
5024 - avoid SetPageDirty on 2.6 (5981)
5025 - don't re-add just-being-destroyed locks to the waiting list (5653)
5026 - when creating new directories, inherit the parent's custom
5027 striping settings if present parent (3048)
5028 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5029 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5030 - don't hold client locks on temporary worklist from l_lru (5666)
5031 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5032 - automatically evict dead clients (5921)
5033 - Update file size properly in create+truncate+fstat case (6196)
5034 - Do not unhash mountpoint dentries, do not allow removal of
5036 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5037 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5038 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5039 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5040 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5041 - Correctly handle reads of files with no objects (6243)
5042 - lctl recover will also mark a device active if deactivate used (5933)
5044 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5045 - allow --write-conf on an MDS with different nettype than client (5619)
5046 - don't write config llogs to MDS for mounts not from that MDS (5617)
5047 - lconf should create multiple TCP connections from a client (5201)
5048 - init scripts are now turned off by default; run chkconfig --on
5049 lustre and chkconfig --on lustrefs to use them
5050 - upcalls are no longer needed for clients to recover to failover
5052 - add --abort-recovery option to lconf to abort recovery on device
5054 - add support for an arbitrary number of OSTs (3026)
5055 - Quota support protocol changes.
5056 - forward compatibility changes to wire structs (6007)
5057 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5058 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5059 - improved "lustre status" script
5060 - initialize blocksize for non-regular files (6062)
5061 - added --disable-server and --disable-client configure options (5782)
5062 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5063 - Vanilla 2.4.29 support
5064 - increase maximum number of obd devices to 520 (6242)
5065 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5066 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5067 SLES 9 2.6.5 kernels
5068 - limit stripes per file to 160 (the maximum EA size) (6093)
5070 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5073 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5074 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5075 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5076 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5077 - small ext3 extents cleanups and fixes (5733)
5078 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5079 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5080 - avoid needless client->OST connect, fix handle mismatch (5317)
5081 - fix DLM error path that led to out-of-sync client, long delays (5779)
5082 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5083 - fix several locking issues related to i_size (5492,5624,5654,5672)
5084 - don't move pending lock onto export if it is already evicted (5683)
5085 - fix kernel oops when creating .foo in unlinked directory (5548)
5086 - fix deadlock in obdfilter statistics vs. object create (5811)
5087 - use time_{before,after} to avoid timer jiffies wrap (5882)
5088 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5089 - Do not leak request if server was not able to process it (5154)
5090 - If mds_open unable to find parent dir, make that negative lookup(5154)
5091 - don't create new directories with extent-mapping (5909, 5936)
5093 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5094 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5095 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5096 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5097 - support for RHEL4 kernel on i686 (5773)
5098 - provide error messages when incompatible logs are encountered (5898)
5100 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5101 * version 1.4.0.10 (1.4.1 release candidate 1)
5103 - don't keep a lock reference when lock is not granted (4238)
5104 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5105 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5106 - drop import inflight refcount on signal_completed_replay error (5255)
5107 - unlock page after async write error during send (3677)
5108 - handle missing objects in filter_preprw_read properly (5265)
5109 - no transno return for symlink open, don't save no-trasno open (3440)
5110 - don't try to complete elan receive that already failed (4012)
5111 - free RPC server reply state on error (5406)
5112 - clean up thread from ptlrpc_start_thread() on error (5160)
5113 - readahead could read extra page into cache that wasn't ejected (5388)
5114 - prevent races in class_attach/setup/cleanup/detach (5260)
5115 - don't dereference de->d_inode after l_dput of de (5458)
5116 - use "int" for stripe value returned from lock_to_stripe (5544)
5117 - mballoc allocation and error-checking fixes in 2.6 (5504)
5118 - block device patches to fix I/O request sizes in 2.6 (5482)
5119 - look up hostnames for IB nals (5602)
5120 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5121 - don't start multiple acceptors for the same port (5277)
5122 - fix incorrect LASSERT in mds_getattr_name (5635)
5123 - export a proc file for general "ping" checking (5628)
5124 - fix "lfs check" to not block when the MDS is down (5628)
5126 - service request history (4965)
5127 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5128 - create an "evict_client" /proc entry on OSTs, like the MDS has
5129 - fix mount usage message, return errors per mount(8) (5168)
5130 - change grep [] to grep "[]" in tests so they work in more UMLs
5131 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5132 - remove ancient LOV_MAGIC_V0 EA support (5047)
5133 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5134 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5135 - allow readahead allocations to fail when low on memory (5383)
5136 - mmap locking landed again, after considerable improvement (2828)
5137 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5139 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5142 - send OST transaction number in read/write reply to free req (4966)
5143 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5144 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5145 - the watchdog thread now runs as interruptible (5246)
5146 - flock/lockf fixes (but it's still disabled, pending 5135)
5147 - don't use EXT3 constants in llite code (5094)
5148 - memory shortage at startup could cause assertion (5176)
5150 - reorganization of lov code
5151 - single portals codebase
5153 - add extents/mballoc support (5025)
5154 - direct I/O reads in the obdfilter (4048)
5155 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5157 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5160 - send OST transaction number in read/write reply to free req (4966)
5161 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5162 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5163 - the watchdog thread now runs as interruptible (5246)
5164 - handle missing objects in filter_preprw_read properly (5265)
5165 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5166 - drop import inflight refcount on signal_completed_replay error (5255)
5167 - unlock page after async write error during send (3677)
5168 - return original error code on reconstructed replies (3761)
5169 - no transno return for symlink open, don't save no-trasno open (3440)
5171 - add pid to ldlm debugging output (4922)
5172 - bump the watchdog timeouts -- we can't handle 30sec yet
5173 - extra debugging for orphan dentry/inode bug (5259)
5175 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5178 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5179 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5180 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5181 - fix NULL dereference in /proc/sys/portals/routes (4827)
5182 - allow failed mdc_close() operations to be interrupted (4561)
5183 - stop precreate on OST before MDS would time out on it (4778)
5184 - don't send partial-page writes before EOF from client (4410)
5185 - discard client grant for sub-page writes on large-page clients (4520)
5186 - don't free dentries not owned by NFS code, check generation (4806)
5187 - fix lsm leak if mds_create_objects() fails (4801)
5188 - limit debug_daemon file size, always print CERROR messages (4789)
5189 - use transno after validating reply (3892)
5190 - process timed out requests if import state changes (3754)
5191 - update mtime on OST during writes, return in glimpse (4829)
5192 - add mkfsoptions to LDAP (4679)
5193 - use ->max_readahead method instead of zapping global ra (5039)
5194 - don't interrupt __l_wait_event() during strace
5196 - add software watchdogs to catch hung threads quickly (4941)
5197 - make lustrefs init script start after nfs is mounted
5198 - fix CWARN/ERROR duplication (4930)
5199 - return async write errors to application if possible (2248)
5200 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5201 - print NAL number in %x format (4645)
5202 - update barely-supported suse-2.4.21-171 series (4842)
5203 - support for sles 9 %post scripts
5204 - support for building 2.6 kernel-source packages
5205 - support for sles km_* packages
5207 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5210 - ignore -ENOENT errors in osc_destroy (3639)
5211 - notify osc create thread that OSC is being cleaned up (4600)
5212 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5213 - reconstruct ost_handle() like mds_handle() (4657)
5214 - create a new thread to do import eviction to avoid deadlock (3969)
5215 - let lconf resolve symlinked-to devices (4629)
5216 - don't unlink "objects" from directory with default EA (4554)
5217 - hold socknal file ref over connect in case target is down (4394)
5218 - allow more than 32000 subdirectories in a single directory (3244)
5219 - fix blocks count for O_DIRECT writes (3751)
5220 - OST returns ENOSPC from object create when no space left (4539)
5221 - don't send truncate RPC if file size isn't changing (4410)
5222 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5223 - bind to privileged port in socknal and tcpnal (3689)
5225 - rate limit CERROR/CWARN console message to avoid overload (4519)
5226 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5227 - basic mmap support (3918)
5228 - kernel patch series update from b1_4 (4711)
5230 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5233 - avoid crash during MDS cleanup with OST shut down (2775)
5234 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5235 - don't use bad inodes on the MDS (3744)
5236 - dynamic object preallocation to improve recovery speed (4236)
5237 - don't hold spinlock over lock dumping or change debug flags (4401)
5238 - don't zero obd_dev when it is force cleaned (3651)
5239 - print grants to console if they go negative (4431)
5240 - "lctl deactivate" will stop automatic recovery attempts (3406)
5241 - look for existing locks in ldlm_handle_enqueue() (3764)
5242 - don't resolve lock handle twice in recovery avoiding race (4401)
5243 - revalidate should check working dir is a directory (4134)
5245 - don't always mark "slow" obdfilter messages as errors (4418)
5247 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5250 - don't close LustreDB during write_conf until it is done (3860)
5251 - fix typo in lconf for_each_profile (3821)
5252 - allow dumping logs from multiple threads at one time (3820)
5253 - don't allow multiple threads in OSC recovery (3812)
5254 - fix debug_size parameters (3864)
5255 - fix mds_postrecov to initialize import for llog ctxt (3121)
5256 - replace config semaphore with spinlock (3306)
5257 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5258 - don't allow enqueue to complete on a destroyed export (3822)
5259 - down write_lock before checking llog header bitmap (3825)
5260 - recover from lock replay timeout (3764)
5261 - up llog sem before sending rpc (3652)
5262 - reduce ns lock hold times when setting kms (3267)
5263 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5264 - fix NULL deref and obd_dev leak on setup error (3312)
5265 - replace some LBUG about llog ops with error handling (3841)
5266 - don't match INVALID dentries from d_lookup and spin (3784)
5267 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5268 - fix invalid assertion in ptlrpc_set_wait (3880)
5270 - add libwrap support for the TCP acceptor (3996)
5271 - add /proc/sys/portals/routes for non-root route listing (3994)
5272 - allow setting MDS UUID in .xml (2580)
5273 - print the stack of a process that LBUGs (4228)
5275 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5278 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5279 - make sure to unset replay flag from failed open requests (3440)
5280 - if default stripe count is 0, use OST count for inode size (3636)
5281 - update parent mtime/ctime on client for create/unlink (2611)
5282 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5283 - free recovery state on server during a forced cleanup (3571)
5284 - unregister_reply for resent reqs (3063)
5285 - loop back devices mounting and status check on 2.6 (3563)
5286 - fix resource-creation race that can provoke i_size == 0 (3513)
5287 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5288 - more debugging for page-accounting assertion (3746)
5289 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5290 - avoid deadlock after precreation failure (3758)
5291 - fix race and lock order deadlock in orphan handling (3450, 3750)
5292 - add validity checks when grabbing inodes from l_ast_data (3599)
5294 - add /proc/.../recovery_status to obdfilter (3428)
5295 - lightweight CDEBUG infrastructure, debug daemon (3668)
5296 - change default OSC RPC parameters to be better on small clusters
5297 - turn off OST read cache for files smaller than 32MB
5298 - install man pages and include them in rpms (3100)
5299 - add new init script for (un)mounting lustre filesystems (2593)
5300 - run chkconfig in %post for init scripts (3701)
5301 - drop scimac NAL (unmaintained)
5303 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5306 - clean kiobufs before and after use (3485)
5307 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5308 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5309 - change init script to fail more clearly if not run as root (1528)
5310 - allow clients to reconnect during replay (1742)
5311 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5312 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5313 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5314 - fix cond_resched (3554)
5315 - only evict unfinished clients after recovery (3515)
5316 - allow bulk resend, prevent data loss (3570)
5317 - dynamic ptlrpc request buffer allocation (2102)
5318 - don't allow unlinking open directory if it isn't empty (2904)
5319 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5320 - remove extraneous obd dereference causing LASSERT failure (3334)
5321 - don't use get_cycles() when creating temp. files on the mds (3156)
5322 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5323 - handle EEXIST for set-stripe, set proper directory name (3336)
5325 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5326 - fix ksocknal_fmb_callback() error messages (2918)
5328 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5331 - don't copy lvb into (possibly NULL) reply on error (2983)
5332 - don't deref dentry after dput, don't free lvb on error (2922)
5333 - use the kms to determine writeback rpc length (2947)
5334 - increment oti_logcookies when osc is inactive (2948)
5335 - update client's i_blocks count via lvb messages (2543)
5336 - handle intent open/close of special files properly (1557)
5337 - mount MDS with errors=remount-ro, like obdfilter (2009)
5338 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5339 - don't use cancelling-locks' kms values (2947)
5340 - use highest lock extent for kms, not last one (2925)
5341 - don't dereference ERR_PTR() dentry in error handling path (3107)
5342 - fix thread race in portals_debug_dumplog() (3122)
5343 - create lprocfs device entries at setup instead of at attach (1519)
5344 - common AST error handler, don't evict client on completion race (3145)
5345 - zero nameidata in detach_mnt in 2.6 (3118)
5346 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5347 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5348 - initialize RPC timeout timer earlier for 2.6 (3219)
5349 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5350 - print nal/nid for unknown nid (3258)
5351 - additional checks for oscc recovery before doing precreate (3284)
5352 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5353 - don't crash in mdc_close for bad permissions on open (3285)
5354 - zero i_rdev for non-device files (3147)
5355 - clear page->private before handing to FS, better assertion (3119)
5356 - tune the read pipeline (3236)
5357 - fix incorrect decref of invalidated dentry (2350)
5358 - provide read-ahead stats and refine rpc in flight stats (3328)
5359 - don't hold journal transaction open across create RPC (3313)
5360 - update atime on MDS at close time (3265)
5361 - close LDAP connection when recovering to avoid server load (3315)
5362 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5363 - don't leak open file on MDS after open resend (3325)
5364 - serialize filter_precreate and filter_destroy_precreated (3329)
5365 - loop device shouldn't call sync_dev() for nul device (3092)
5366 - clear page cache after eviction (2766)
5367 - resynchronize MDS->OST in background (2824)
5368 - refuse to mount the same filesystem twice on same mountpoint (3394)
5369 - allow llmount to create routes for mounting behind routers (3320)
5370 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5371 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5372 - fix rare problem with rename on htree directories (3417)
5374 - allow default OST striping configuration per directory (1414)
5375 - fix compilation for qswnal for 2.6 kernels (3125)
5376 - increase maximum number of MDS request buffers for large systems
5377 - change liblustreapi to be useful for external progs like lfsck (3098)
5378 - increase local configuration timeout for slow disks (3353)
5379 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5381 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5384 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5385 - don't overwrite extent policy data in reply if lock was blocked (2901)
5386 - drop filter export grants atomically with removal from device (2663)
5387 - del obd_self_export from work_list in class_disconnect_exports (2908)
5388 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5389 - swab reply message in mdc_close, other PPC fixes (2464)
5390 - fix destroying of named logs (2325)
5391 - overwrite old logs when running lconf --write_conf (2264)
5392 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5393 - fix race in target_handle_connect (2898)
5394 - mds_reint_create() should take same inode create lock (2926)
5395 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5396 - don't close files for self_export to avoid uninitialized obd (2936)
5397 - allow MDS with the same name as client node (2939)
5398 - hold dentry reference for closed log files for unlink (2325)
5399 - reserve space for all logs during transactions (2059)
5400 - don't evict page beyond end of stripe extent (2925)
5401 - don't oops on a deleted current working directory (2399)
5402 - handle hard links to targets without a parent properly (2517)
5403 - don't dereference NULL lock when racing during eviction (2867)
5404 - don't grow lock extents when lots of conflicting locks (2919)
5406 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5409 - account for cache space usage on clients to avoid data loss (974)
5410 - lfsck support in lustre kernel code (2349)
5411 - reduce journal credits needed for BRW writes (2370)
5412 - orphan handling to avoid losing space on client/server crashes
5413 - ptlrpcd can be blocked, stopping ALL progress (2477)
5414 - use lock value blocks to assist in proper KMS, faster stat (1021)
5415 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5416 - recovery for initial connections (2355)
5417 - fixes for mds_cleanup_orphans (1934)
5418 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5419 - block all file creations until orphan recovery completes (1901)
5420 - client remove rq_connection from request struct (2423)
5421 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5422 - recovery timer race (2670)
5423 - mdc_close recovey bug (2532)
5424 - ptlrpc cleanup bug (2710)
5425 - mds timeout on local locks (2588)
5426 - namespace lock held during RPCs (2431)
5427 - handle interrupted sync write properly (2503)
5428 - don't try to handle a message that hasn't been replied to (2699)
5429 - client assert failure during cleanup after abort recovery (2701)
5430 - leak mdc device after failed mount (2712)
5431 - ptlrpc_check_set allows timedout requests to complete (2714)
5432 - wait for inflight reqs when ptlrpcd finishes (2710)
5433 - make sure unregistered services are removed from the srv_list
5434 - reset bulk XID's when resending them (caught by 1138 test)
5435 - unregister_bulk after timeout
5436 - fix lconf error (2694)
5437 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5438 - readahead locks pages, leaves pending causing memory pressure (2673)
5439 - increase OST request buffers to 4096 on large machines (2729)
5440 - fix up permission of existing directories in simple_mkdir (2661)
5441 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5442 - don't assign transno to errored transactions (2742)
5443 - don't delete objects on OST if given a bogus objid from MDS (2751)
5444 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5445 - if rq_no_resend, then timeout request after recovery (2432)
5446 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5447 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5448 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5449 - MDS recovery shouldn't delete objects that it has given out (2730)
5450 - if enqueue arrives after completion, don't clobber LVB (2819)
5451 - don't unlock pages twice when trigger_group_io returns error (2814)
5452 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5453 - don't write pages to disk if there was an error (1450)
5454 - don't ping imports that have recovery disabled (2676)
5455 - take buffered bytes into account when balancing socknal conn (2817)
5456 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5457 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5458 - remove little-endian swabbing of llog records (1987)
5459 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5460 - retry reposting request buffers if they fail (1191)
5461 - grow extent at grant time to avoid granting a revoked lock (2809)
5462 - lock revoke doesn't evict page if covered by a second lock (2765)
5463 - disable VM readahead to avoid reading outside lock extents (2805)
5465 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5466 - updated kernel patches for hp-2.4.20 kernel (2681)
5468 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5471 - fix truncated write corruption (2366)
5472 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5474 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5475 - ensure len(uuid) < 37 in lmc (1171)
5476 - fix ia64 OOPS in llog_test (2255)
5477 - zero end of page at obdfilter for partial page writes (2648)
5478 - don't leave stale dentries around after renames (bug 2428)
5479 - fix timeouts when evicting a client with a single lock held (2642)
5480 - set deadline for the initial HELLO message to drain (2634)
5481 - print out dotted-quad IP addresses in the socknal (2302)
5483 - additional debugging for MDS client eviction problem (2443)
5484 - fix mkfsoptions support for osts (2603, 2604)
5486 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5489 - add series for the vanilla 2.6.0 kernel
5490 - add series for the vanilla 2.4.24 kernel
5491 - add series for a cray x86/64 UL kernel drop
5492 - fix xattr patches for the vanilla 2.4.19 series
5494 - generate true UUIDs in lmc (1171)
5495 - have portals stack dumping break in UML (2466)
5496 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5497 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5498 - wake callback waiting threads on client eviction (2460)
5499 - Add --ptldebug and --subsystem to lmc (1719)
5500 - update assertion to allow safe interrupt allocation
5501 - set rq_no_resend for cancel requests (2432)
5502 - recalculate ptlrpcd timeout after resend (2494)
5503 - call vfs_rmdir when removing pending directories (2368)
5504 - fix renaming a file to itself (2429)
5505 - lmc creates a default one-stripe lov (2454)
5506 - expand procfs space to handle large clusters (2326)
5507 - increase UML stack to avoid overflow
5508 - update lconf's list of debug and subsystem masks
5509 - fix lfs find --obd (2510)
5510 - /proc tunable for disabling filter read caching (2591)
5511 - stop rpm packages from altering slapd.conf (2301)
5512 - disable nagle in the socknal under 0conf (2578)
5513 - choose mds inode size based on stripe count (2572)
5514 - fix kernel-source rpm problems (2516)
5516 - add --disable-doc to avoid pdf generation (2421)
5517 - update documentation, tests, type-os, comments
5518 - avoid format warnings on ia64
5519 - remove the TOE NAL
5520 - tiny code cleanups by removing unused fields
5522 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5525 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5526 - remove the most blatant lies from BUILDING file (2371)
5527 - change default debug level to reasonable production setting
5528 - reduce client side cache size to reduce cache flush time
5529 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5530 - make TCP zerocopy and pinger support enabled by default (2476)
5531 - sync writes completed after process exits caused crashes (2319)
5532 - maintain correct mount count on the MDS (2356)
5533 - backout 1557, because 2316 wasn't really fixed
5534 - better file I/O statistics gathering in /proc
5535 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5536 - another kernel patch to fix zero-copy TCP function export
5537 - don't take duplicate lock when processing re-sent getattr (2420)
5538 - lctl uses obd_self_export instead of creating new conn (2353)
5539 - MDS/OST recovery case which requires object creation asserted (2425)
5540 - move lfs from /usr/sbin to /usr/bin in packages
5541 - fix race between mds_client_add and mds_client_free (2417)
5542 - use kmalloc instead of slabs in portals (2430)
5543 - don't create duplicate records when a failover MDS is present (2442)
5544 - remove unnecessary mount age check (2332)
5545 - don't remove directory inodes from locks prematurely (2451)
5546 - don't break if MDS service name is the same as hostname (2103)
5547 - fix races in client write RPC generation when cache full (2482)
5549 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5552 - remove now-unused request->rq_obd (278)
5553 - if an allocation fails, print out how much memory we've used (1933)
5554 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5555 - add the "configurable stack size" patch to most series files (1256)
5556 - ability to write large log records, for 100+ OST configs (2306)
5557 - fix NULL deref when filter_prep fails (2314)
5558 - fix operator precedence error in filter_sync
5559 - dynamic allocation of socknal TX descriptors (2315)
5560 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5561 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5562 - ensure that CWARN messages in Portals always get to the syslog
5563 - __init/__exit are not for prototype decls (ldlm_init/exit)
5564 - x86-64 compile warning fixes
5565 - fix gateway LMC keyword conflict (2318)
5566 - fix MDS lock inversions in getattr/reint paths (1844)
5567 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5568 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5569 - DLM race condition prevented some lock evictions (2328)
5570 - ENOMEM detection and retry on socknal sends (2230)
5571 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5572 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5573 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5574 - better error messages when a client is rejected during recovery (1505)
5575 - avoid cancelling locks which were never granted, after failure (2330)
5576 - fix i_sem/journal inversion in mds_client_add (2333)
5577 - fix truncate/getattr lock cycle deadlock (2334)
5578 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5579 - fix two rare exit paths which could leak an l_lock() ref (2321)
5580 - fencepost error in MDS/OST orphan recovery (2226)
5581 - make log record alignment 8 bytes (1988)
5582 - lstripe now fails when requested offset > ost_count (2237)
5583 - ensure that all kernel series have a complete list.h (1607)
5584 - fix crashes in special-file operations (2316)
5585 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5587 - allow configurable automake binary, for testing new versions
5588 - small update to the lfs documentation
5590 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5592 * fix negative export reference count in fsfilt_sync (2312)
5594 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5595 * release candidate 0.9.1
5597 - orphans are moved into the PENDING directory for possible recovery
5598 - replayed opens now open by fid for orphan/rename safety (1042)
5599 - last close of an orphan inode generates a transno (683)
5600 - chdir() and mount() now pin the directory entry (1020)
5601 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5602 - discard very old requests without processing them (1502)
5603 - remove some common, well-understood CERRORs (1505)
5604 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5605 - clear "grant" flags in OST replies until OST grant code lands (1644)
5606 - fix read performance by not clobbering i_blksize on client (1598)
5607 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5608 - make LRU size a /proc tunable, clears locks when reduced (707)
5609 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5610 - server locks take references on exports now (1558)
5611 - build fixes for 2.4.20-rh trees (1663)
5612 - return an error from lov_create if all OSCs are inactive (1751)
5613 - fix import levels when a reconnect happens without a timeout (1597)
5614 - exit early from mds_open if we get a lookup error (1749)
5615 - partial page read at EOF wouldn't wait for disk before sending (1642)
5616 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5617 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5618 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5619 - write/truncate lock inversion (1639)
5620 - Don't auto-load obdclass, portals modules during cleanup (1495)
5621 - fix timestamps from jumping to "now" (1763)
5622 - extra journal assertions (1648)
5623 - add an extra multiunlink test (1771)
5624 - fix read_record/write_record API (1776)
5625 - fix leak of offset_extent, possible incorrect i_size later (1772)
5626 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5627 - Debugging for the kqswnal_get_idle_tx problems (1820)
5628 - Allow recovery to be attempted multiple times (1536)
5629 - Write out MDS last_rcvd file after it is first created (1600)
5630 - Fix tx_descriptor leak in failed transmit situations (1827)
5631 - ext3 journaling fixes for assertion failure after IO error (1871)
5632 - class_export_put() on freed export after completion AST error (1896)
5633 - Fix revalidate looping in VFS (1322)
5634 - Don't access a freed export during MDS_REINT timeout (1521)
5635 - Add open-unlink recovery support on the MDS (1673,1764)
5636 - Return an error if no MDS data was read from last_rcvd (1946)
5637 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5638 - Don't setup a disk device that doesn't match exported UUID (317)
5639 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5640 - avoid committing NULL handle in force close
5641 - local.sh is now a one-stripe LOV configuration
5642 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5643 - fix ext3 htree duplicate directory entry corruption (1516)
5644 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5645 - update to new LOV EA format (2097)
5646 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5648 2003-06-15 Phil Schwan <phil@clusterfs.com>
5651 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5652 - per-import/export recovery handling (958, 931, 959)
5653 - multiple last-rcvd slots, for serving multiple FSes (949)
5654 - connections are again shared between multiple imp/exports (963, 964)
5655 - "umount -f" would hang if any requests needed to be sent (393, 978)
5656 - avoid pinning large req buffer by copying for queued messages (989)
5657 - add "uuid" to "lctl device" command to help upcalls (991)
5658 - "open" RPCs with transnos would confuse recovery counters (1037)
5659 - do proper endian conversion of all wire messages (288, 340, 891)
5660 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5661 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5662 - fix ptlrpc_connection leak in target_handle_connect (1174)
5663 - fix import refcounting bug in OST and MDS cleanup (1134)
5664 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5665 - fix very unlikely obd_types race condition (501)
5666 - remove osc_open hack for echo_client (1187)
5667 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5668 - a failure in read_inode2 leads to deadlock (1139)
5669 - cancel ack-locks as soon as transaction is committed (1072)
5670 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5671 - make sure to commitrw after any preprw to avoid deadlock (1162)
5672 - failing to execute a file in a lustre FS would lock inode (1203)
5673 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5674 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5675 - obd_uuid2tgt would walk off the end of the list (1255)
5676 - on IA64 the osc would give portals incorrect bulk size (1258)
5677 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5678 - fix lock inversion caused by new llite matching code (1282)
5679 - limit the number of dirty pages on a client to 10MB (1286)
5680 - timed out locks were not being corrected cancelled (1289)
5681 - fix O_DIRECT above 4GB on IA-32 (1292)
5682 * major user-visible changes
5683 - fail out/fail over policy now controlled by the upcall (993)
5685 - add OBD_PING to check server availability and failure (954)
5686 - lustre messages are now sent in sending host order (288, 340, 891)
5687 - add eadatalen to MDS getattr reply (340)
5688 - OST read replies may contain second buffer, with per-page status (593)
5690 2003-03-11 Phil Schwan <phil@clusterfs.com>
5693 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5694 - failed open()s could cause deadlock; fixed (867, 869)
5695 - stop cancelling OST locks when files are closed (481)
5696 - overlapping XID spaces caused network corruption (851, 853)
5697 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5698 - setattr_raw added, to do single-RPC, server-side setattrs
5699 - lmc/lconf syntax change for OST UUIDs
5700 - fix crashy race condition between ptlrpc_free_req and osc_close
5701 - don't use request in mdc_enqueue if we hit a timeout (889)
5702 - don't set the inode i_size for regular files from the MDS (896)
5703 - handle out of order completion AST (842)
5704 - don't LBUG if a lock request times out after receiving AST (913)
5705 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5706 - if a bad lock AST arrives, send an error instead of dropping entirely
5707 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5708 - fix leak in bulk IO when only partially completed (899, 900, 926)
5709 - fix O_DIRECT for ia64 (55)
5710 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5711 - C-z after timeout could hang a process forever; fixed (977)
5713 - client-side I/O cache (678, 924, 929, 941, 970)
5715 - READPAGE and SETATTRs which don't take server-side locks get
5718 2003-02-11 Phil Schwan <phil@clusterfs.com>
5721 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5722 - Fix fsfilt_extN_readpage() to read a full page of directory
5723 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5724 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5725 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5726 - Most intent operations were removed, in favour of a new RPC mode
5727 that does a single RPC to the server and bypasses most of the VFS
5728 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5729 - Aggressively cancel local locks on DLM servers
5730 - mds_reint_unlink sends EA to the client if it's the last nlink.
5731 client uses that EA to unlink OST objects.
5732 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5733 - recursive symlinks were fixed (439)
5734 - fixed NULL deref in DEBUG_REQ
5735 - filter_update_lastobjid no longer calls sync, which annoyed extN
5736 - fixed multi-client small-writes to a single file problem (445)
5737 - fixed mtime updates during file writes (607)
5738 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5739 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5740 - fixed Linux OST size reporting problem (444, 656)
5741 - OST now updates object mtime with writes or setattr (607, 619)
5742 - client verifies file size before zeroing page past EOF (445)
5743 - OST now writes last allocated objid to disk with allocation (108)
5744 - LOV on echo now works (409)
5746 - mds_reint_unlink sends a new buffer, with the EA included. this
5747 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5748 set if a regular file was being unlinked, and it was the last link
5749 - use PtlGet from the target for bulk writes (315)
5750 - OST now updates object mtime with writes or setattr (607, 619)
5751 - LDLM now has a grant-time callback to revalidate locked items, if
5753 - Many MDS operations were reorganized to combat race conditions
5755 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5756 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5758 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5761 - Fully reactivate OST imports after reconnection (512, others)
5762 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5763 - More graceful error handling for truncating on dead OST (515)
5764 - Don't error out unless we're actually accessing dead stripes (474)
5765 - Fix garbage sizes when stripes are missing (410)
5766 - LRU counters were broken, causing constant lock purge (433, 432)
5767 - garbage on read from stripes with failed OSTs (441)
5768 - mark OSCs as active before reconnecting during recovery (438)
5769 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5770 - lfind did not preserve OST order in output (443)
5771 - symlinks cause hung clients, incorrect data (439)
5772 - stop dereferencing request after dropping refcount (457)
5773 - don't LASSERT(spin_is_locked) on non-SMP (455)
5774 - fixes for many rename() bugs
5775 - fstat didn't correctly synchronize attributes (399)
5776 - server must handle lock cancellation during blocking AST prep (487)
5777 - bulk descriptors were free()d too soon (511)
5778 - fix paths in lconf, which would load incorrect modules (451, 507)
5779 - fix confusing lconf 'host not found' error message (386)
5780 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5781 - fix race condition in mdc_blocking_ast() for inode access (526)
5782 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5783 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5784 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5785 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5786 - fix lstripe in several ways (kernel side) (527)
5787 - fix request leak in ldlm_cli_enqueue (262)
5788 - incorrect OSC was marked inactive after OST failure
5789 - call mds_fs_cleanup before unmounting filesystem (524)
5790 - fix races between taking ns_lock and ldlm_lock_change_resource
5791 - fix races updating LOV export open file list
5792 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5793 - fix recovery NULL deref in ldlm_cli_cancel_unused
5794 - fix some DLM races by using new hash table for lock handles (419)
5795 - permit the client to specify desired inodes, at replay
5796 - duplicate requests when we queue them for replay reintegration
5797 - fix last_rcvd offset calculation
5798 - sync after each recovered transaction, so we always make progress
5799 - never, not always, ERESTART requests without transnos
5800 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5802 - skip replay if the MDS says that the client is already connected
5803 - don't check for a recovery-enabled export to match lctl's UUID
5804 - don't INC_USE_COUNT for phantom exports
5805 - don't crash when cleaning up phantom exports (567)
5806 - don't double-finish or set replay data for errored mdc_open requests
5807 - abort requests when they time out, so we don't get old replies
5808 - send/receive replies for AST messages again
5809 - if the client says that it doesn't have the lock, cancel it on the
5811 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5812 mark it as destroyed, it will all work out when decref is called
5813 - fix module use counts (22, 581)
5815 - ASTs now expect a reply (server cancels lock on error reply)
5817 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5820 - fix many simultaneous client startup (392)
5821 - fix dentry->d_it clobbering
5822 - credentials weren't being shipped for readdir/getattr operations
5823 - remove invalid assertions triggered during some concurrent MD
5825 - proper Lustre versions added (336, 389)
5826 - fix memory leak for create error case (398)
5827 - fix LOV locking bug that would get cli/srv out of sync
5828 - fix echo client over LOV (409)
5829 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5830 - fix double-O_EXCL intent crash (424)
5831 - avoid sending multiple lock CANCELs (352)
5833 - MDS can do multi-client recovery (modulo bugs in new code)
5835 - many updates, edits, cleanups
5837 2002-11-18 Phil Schwan <phil@clusterfs.com>
5840 - fix null d_it dereference (346)
5841 - fix full OST/dbench hang (333)
5842 - fix permission problem with file removal (286)
5843 - fix removal of OSCs from LOV when they fail
5844 - fix NULL deref during bulk timeout (214)
5845 - fix problems related to multiple filesystems on one MDS (241)
5846 - fixed serious subtle metadata locking bugs
5847 - free locks on clients when inodes are removed due to memory
5849 - fix inode pointer in lock data (285)
5850 - partial support for multiple MDS on a single host (241)
5851 - data locks weren't cancelled at clear_inode time (290, 311)
5852 - intent locks could lead to unbounded lock growth (205)
5853 - added a maximum lock count, an LRU list, and a flusher
5854 - fix multiple rename (365)
5855 - properly abstracted the echo client
5856 - OSC locked 1 byte too many; fixed
5857 - rewrote brw callback code:
5858 - fixed recovery bugs related to LOVs (306)
5859 - fixed too-many-pages-in-one-write crash (191)
5860 - fixed (again) crash in sync_io_timeout (214)
5861 - probably fixed callback-related race (385)
5863 - Add capability to MDS protocol
5864 - LDLM cancellations and callbacks on different portals
5866 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5869 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5870 - defer open object destruction to close time (601981, 138)
5871 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5872 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5873 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5874 - don't call obd_setattr in ll_file_release for destroyed objects
5875 * protocol change to lustre_msg: move |version| and add |flags|
5876 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5877 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5878 * added replay of create, unlink, link and rename operations during
5879 MDS failover; recovery should be much more robust now
5880 * remove failed OSCs from LOVs (only lov_create uses this so far)
5881 * the lustre-HOWTO was brought (more) up to date (582544)
5883 2002-10-23 Phil Schwan <phil@clusterfs.com>
5886 - in-use dentries weren't being reused properly (617851)
5887 - prevent multiple LDLM setup (599178)
5888 - fix LOV size calculations for truncate (617853)
5889 - fix client handling of MDS intent errors (POSIX)
5890 - fix permission bug in lovstripe.c test (624321)
5891 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5892 - truncate past end of file could corrupt data
5893 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5894 - a race in recovery could return ETIMEDOUT to apps (623947)
5895 - building outside the source directory was fixed
5896 * the lustre-HOWTO was brought (more) up to date (582544)
5897 * major progress was made on recovery functionality
5899 2002-10-10 Phil Schwan <phil@clusterfs.com>
5902 - recovery deadlock fix
5903 - rm -rf causes LBUG fix (617817)
5904 - file open by multiple tasks fix (618962)
5905 - directory permissions bugs (602707 and 620007)
5906 - journal_stop fixed with locking (611313)
5907 - O_APPEND failures resolved (618273, perhaps 614459)
5908 - lconf PATH fix (619770)
5909 - IA64 build fix (621450)
5910 - RPC buffer sizes scale with amount of memory
5912 2002-10-01 Phil Schwan <phil@clusterfs.com>
5915 - locks would be cancelled without throwing away data pages,
5916 resulting in inconsistent data (605627)
5917 - inode attributes were not always being refreshed (605627, 612449)
5918 - lconf now continues to cleanup after lctl reports an error
5919 - MDS now enforces user permissions (602707)
5920 - lprocfs cleanup fixed, but not yet enabled (614157)
5921 - fixed infinite server hang, should a client not respond to an AST
5922 - avoid going into recovery if user calls readlink() with a buffer
5923 that's too small (613941)
5924 - AST RPCs no longer require replies (614867) -- this may be changed
5925 - don't crash server if client sends an IOV that's too big (611336)
5926 - fixed lock conversion deadlock (611892)
5927 - fixed the following of symlinks (614622)
5928 * recovery: the server can remove locks from a client that dies, other
5929 clients can make progress
5930 * more extN patch fixes
5931 * compile-time configurable ptlrpc buffer allocations
5933 - collaborative read cache document
5934 - Lustre Lite Performance CDR document-in-progress
5936 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5939 - fix typo in patch-2.4.18
5941 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5944 - clear ptlrpc request each time in handle_incoming_request()
5945 - unlink of files now destroys the object on the OST
5947 2002-09-19 Peter Braam <braam@clusterfs.com>
5949 * add hard link support
5950 * change obdfile creation method
5951 * kernel patch changed
5953 2002-09-19 Peter Braam <braam@clusterfs.com>
5956 - stack overflow bug in extN fixed
5958 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5960 * documentation updates
5961 - add man pages for config tools
5962 - update tests/README to describe testing with new config tools
5963 - finish metadata API descriptions
5964 * bug fixes and cleanups
5965 - statfs workaround for 16TB limit
5966 - LOV stripe allocation improved, can stripe on subset of OSTs
5967 - LOV file size/IO offset was wrong for files > 4GB in size
5968 - object EA data was being dropped, caused files to be unreadable
5969 - memory overflow with non-LOV OST caused memory corruption
5970 - fixed regression tests to work with new config tools, obdfilter
5971 - fixed bug when directory size became larger than 1 block
5972 - fixed bug (for single client case) when PWD was deleted
5973 - invalidate local directory pages when doing intent-based ops
5974 - avoid LDLM oops when lock callback contained bad data
5976 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5978 * documentation updates
5979 * bug fixes and cleanups
5980 - configuration tools
5983 - 64-bit compile warnings
5984 - 64-bit internal statfs data
5986 * test_brw on persistent OST devices
5988 * lprocfs (disabled)
5990 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5992 * documentation updates
5993 * bug fixes and cleanups
5994 * configuration tools
5996 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5999 * many small fixes to 0_5_4
6000 * io/network handling
6001 * thinkos in MDS operations
6003 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6006 * crucial basic fixes to 0.5.3
6007 * IOR, Iozone work over Elan
6010 2002-08-07 Phil Schwan <phil@clusterfs.com>
6011 * version 0_5_3, our first alpha
6012 * we use the new Portals iovs
6013 * documentation updates
6014 * bug fixes and cleanups
6015 * small changes in the DLM wire protocol
6017 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6018 * version 0_5_1 with some initial stability,
6019 * locking on MD and file I/O.
6020 * documentation updates
6021 * several bug fixes since 0.5.0
6022 * small changes in wire protocol
6024 2002-07-18 Phil Schwan <phil@clusterfs.com>
6026 * delivered as Lustre Light Alpha
6027 * fixed a crash after handling invalid MDS requests
6028 * fixed directory pages for architectures with non-4k pages sizes
6030 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6031 * release version v0_4_4
6032 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6034 * Fixes a number of interruption problems with OST operations.
6035 * Update documentation for portals header changes
6036 * Move all wire protocol structs/defines to lustre_idl.h
6037 * Fixes symlink length bug.
6038 * Add tcpdump to repository.
6040 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6041 * release version v0_4_3
6042 * Fixes statfs for inodes on extN.
6043 * Fixes bug in runtests which would delete /etc/hosts.
6044 * Use 64-bit object IDs wherever possible (not into VFS though)
6045 Remove ost_get_info, which is unused by lustre, and out of date.
6047 2002-07-03 Peter Braam <braam@clusterfs.com>
6048 * release version v0_4_2 Fixes a lookup error (type not passed)
6049 * move forward to head of Portals
6050 * move forward to latest Lustre kernel
6052 2002-06-25 Peter Braam <braam@clusterfs.com>
6053 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob