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.
17 Severity : enhancement
19 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
21 Severity : enhancement
23 Description: correctly handle big reply message.
24 Details : send LNet event if reply is bigger then buffer and adjust this buffer
29 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
31 Severity : enhancement
33 Description: Add support for OEL5.
37 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
41 Description: client could not reconnect to OST because of an active request.
42 Details : The client now retries to reconnect to the same server, if a
43 connect request failed with EBUSY or -EAGAIN. Test case 35b
44 conf-sanity.sh has been created.
48 Description: Don't allow make backward step on assignin osc next id.
49 Details : race between allocation next id and ll_sync thread can be cause
50 of set wrong osc next id and can be kill valid ost objects.
54 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
58 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
62 Description: Error handling in osc_statfs_interpret() has been improved.
63 Details : Check in osc_statfs_interpret() for EBADR.
67 Description: Do not send statfs() requests to OSTs disabled by administrator.
68 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
70 Severity : enhancement
72 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
74 Severity : enhancement
76 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
78 Severity : enhancement
80 Description: Enhancement for lfs(1) command to use numeric uid/gid.
84 Description: Improved cleanup procedure for conf-sanity test_46a.
88 Description: Do not start bulk transfer, if client tried to reconnect.
89 Details : Do not start bulk transfer, if client tried to reconnect,
90 but failed due to active rpcs. Check for exp_abort_active_req
95 Description: Prevent inconsistences between linux and lustre mount structures.
96 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
97 Make the sleep interruptible.
99 Severity : enhancement
101 Description: provide server to client comms path
102 Details : server to client communications path via new LDLM_SET_INFO rpc
106 Description: truncate starts GFP_FS allocation under transaction causing deadlock
107 Details : ldiskfs_truncate calls grab_cache_page which may start page
108 allocation under an open transaction. This may lead to
109 calling prune_icache with consequent lustre reentrance.
113 Description: send correctly lsm on open replay
114 Details : MDS is trust to LSM size on replay open, but client can set wrong size
119 Description: enable client interface failover
120 Details : When a child reconnects from another NID, properly update export
121 nid hash position and ldlm reverse import.
123 Severity : enhancement
125 Description: Communicate OST degraded/readonly state via statfs to MDS
126 Details : Flags in the statfs returned from OSTs indicate whether the
127 OST is in a degraded RAID state, or if the filesystem has
128 turned read-only after a filesystem error is detected.
133 Description: don't panic if EPROTO was hit when reading symlink
134 Details : correctly handling request reference in error cases.
136 Severity : enhancement
138 Description: Add LustreNetLink, a kernel-userspace communcation path.
140 Severity : enhancement
142 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
147 Description: access to llog context before init.
148 Details : move handling CATALOGS file at osc layer and forbid access to llog
152 Frequency : rare, only when client is writing to OST on local node
154 Description: Avoid deadlock for local client writes
155 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
156 memory freeing context. This allows OST threads to set the
157 PF_MEMALLOC flag on task structures in order to allocate memory
158 from reserved pools and complete IO.
159 Use GFP_HIGHUSER for OST allocations for non-local client writes,
160 so that the OST threads generate memory pressure and allow
161 inactive pages to be reclaimed.
163 Severity : enhancement
165 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
170 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
171 Details : move ldlm namespace creation in setup phase to avoid grab
172 _lprocfs_lock with cli_sem held.
176 Description: Temporarily disable grant shrink.
177 Details : Disable the feature for debugging.
180 Frequency : only when formatting test filesystems
182 Description: Unable to run several mkfs.lustre on loop devices at the same time
183 Details : mkfs.lustre returns error 256 on the concurrent loop devices
184 formatting. The solution is to proper handle the error.
186 Severity : enhancement
188 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
190 Severity : enhancement
192 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
197 Description: resolve race between obd_disconnect and class_disconnect_exports
198 Details : if obd_disconnect will be called to already disconnected export he
199 forget release one reference and osc module can't unloaded.
201 Severity : enhancement
203 Description: Allow tuning service thread via /proc
204 Details : For each service a new
205 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
206 created that can be used to set min/max thread counts, and get the
207 current number of running threads.
212 Descriptoin: don't return error if have partially created objects for file.
213 Details : lov_update_create_set uses set->set_success as index for created
214 objects, so if some requests will be failed, they will have hole
215 at end of array and qos_shrink_lsm for allocate correct lsm.
217 Severity : enhancement
219 Description: Update OFED support to 1.4.1
224 Description: Reduce small size read RPC
225 Details : Set read-ahead limit for every file and only do read-ahead when
226 available read-ahead pages are bigger than 1M to avoid small size
229 Severity : enhancement
231 Description: move AT tunable parameters for more consistent usage
232 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
234 Severity : enhancement
236 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
237 Details : allow skip disconnected ost for send statfs request and hide error
243 Description: fix racy locking of mballoc block bitmaps causing BUG
244 Details : The locking of the mballoc buddy bitmap and the in-memory
245 block bitmap was using two different spin locks in some
246 cases. This made it possible to incorrectly access the
247 mballoc bitmap while another process was modifying it,
248 causing a sanity assertion to fail. While no on-disk corruption
249 was reported, there was some risk of this happening.
252 Frequency : rare, on llog test 6
254 Descriptoin: don't allow connect to already connected import
255 Details : allowing connect to already connected import is hide connecting problem.
260 Descriptoin: allow kill process which wait statahead result
261 Details : for some reasons 'ls' can stick in waiting result from statahead,
262 in this case need way for kill this process.
264 Severity : enhancement
266 Description: Add state history info file, enhance import info file
267 Details : Track import connection state changes in a new osc/mdc proc file;
268 add overview-type data to the osc/mdc import proc file.
270 Severity : enhancement
272 Description: MDS create should not wait for statfs RPC while holding DLM lock.
274 Severity : enhancement
276 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
281 Descriptoin: OST grant too much space to client even there are not enough space.
282 Details : Client will shrink its grant cache to OST if there are no write
283 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
284 this grant cache if there are already not enough avaible space
285 (left_space < total_clients * 32M).
288 Frequency : start MDS on uncleanly shutdowned MDS device
290 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
291 Details : stay in waiting mds<>ost recovery finished produce random bugs
292 due race between two ll_sync thread for one lov target. send
293 ACTIVATE event only if connect realy finished and import have
297 Frequency : rare, connect and disconnect target at same time
299 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
300 Details : don't call obd_disconnect under lov_lock. this long time
301 operation and can block ptlrpcd which answer to connect request.
306 Descriptoin: don't lose wakeup for imp_recovery_waitq
307 Details : recover_import_no_retry or invalidate_import and import_close can
308 both sleep on imp_recovery_waitq, but we was send only one wakeup
312 Frequency : always with long access acl
314 Descriptoin: mds can't pack reply with long acl.
315 Details : mds don't control size of acl but they limited by reint/getattr
318 Severity : enhancement
320 Description: Update to SLES10 kernel-2.6.16.60-0.33.
322 Severity : enhancement
324 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
327 Frequency : start MDS on uncleanly shutdowned MDS device
329 Descriptoin: aborting recovery hang on MDS
330 Details : don't throttle destroy RPCs for the MDT.
333 Frequency : on remount
335 Description: external journal device not working after the remount
336 Details : clear dev_rdonly flag for external journal devices in
342 Description: shutdown vs evict race
343 Details : client_disconnect_export vs connect request race.
344 if client will evicted at this time - we start invalidate
345 thread without referece to import and import can be freed
349 Frequency : rare, need acl's on inode.
351 Description: client can't handle ost additional correctly
352 Details : if ost was added after client connected to mds client can have
353 hit lnet_try_match_md ... to big messages to wide striped files.
354 in this case need teach client to handle config events about add
355 lov target and update client max ea size at that event.
357 Severity : enhancement
359 Description: Changelogs
360 Details : Changelogs are a lightweight mechanism to track filesystem
361 metadata and namespace changes. The changelog is recorded
362 permanently on the MDTs, and is periodically "consumed" / purged
363 when records are no longer needed.
365 Severity : enhancement
367 Description: compact fld format with extents
368 Details : Store range of seq rather than every seq in FLD. Seq
369 controller update FLD rather than clients. In Case of CMD, mdt0
370 has FLD, all other metadata server act as non persistent proxy
371 for FLD queries and cache fld entries in fld cache.
376 Description: don't skip ost target if they assigned to file
377 Details : Drop slow OSCs if we can, but not for requested start idx.
378 This means "if OSC is slow and it is not the requested
379 start OST, then it can be skipped, otherwise skip it only
380 if it is inactive/recovering/out-of-space.
384 Description: more cleanup in mds_lov
385 Details : not send LOV EA under replay, we can't know about they size at this
386 time. Don't allow client connect to mds before any ost connected,
387 for avoid problems with LOV EA size and returning EIO to client.
389 Severity : enhancement
391 Description: Interoperability at server side (Disk interoperability)
393 Severity : enhancement
395 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
397 Severity : enhancement
399 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
401 Severity : enhancement
403 Description: New client IO stack (CLIO).
405 Severity : enhancement
407 Description: Commit on sharing. Eliminate inter-client dependencies between
408 uncommitted transactions by doing transaction commits.
409 Thereby clients may recovery independently.
412 Frequency : Create a symlink file with a very long name
414 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
415 Details : If there is no extra space in the request for early cancels,
416 ldlm_req_handles_avail() returns 0 instead of a negative value.
418 Severity : enhancement
420 Description: Add /proc entry for import status
421 Details : The mdc, osc, and mgc import directories now have
422 an import directory that contains useful import data for debugging
425 Severity : enhancement
427 Description: Re-disable certain /proc logging
428 Details : Enable and disable client's offset_stats, extents_stats and
429 extents_stats_per_process stats logging on the fly.
432 Frequency : Only on FC kernels 2.6.22+
434 Description: oops in statahead
435 Details : Do not drop reference count for the dentry from VFS when lookup,
436 VFS will do that by itself.
438 Severity : enhancement
440 Description: Generic /proc file permissions
441 Details : Set /Proc file permissions in a more generic way to enable non-
442 root users operate on some /proc files.
446 Description: Hitting mdc_commit_close() ASSERTION
447 Details : Properly handle request reference release in
448 ll_release_openhandle().
452 Description: quota recovery deadlock during mds failover
453 Details : This patch includes att18982, att18236, att18237 in bz14840.
455 1. fix osts hang when mds does failover with quotaon
456 2. prevent watchdog storm when osts threads wait for the
461 Frequency : only patchless client
462 Description: add workaround for race between add/remove dentry from hash
464 Severity : enhancement
466 Description: Allow OST glimpses to return PW locks
470 Description: LBUG when llog conf file is full
471 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
475 Description: Prevent import from entering FULL state when server in recovery
479 Description: service mount cannot take device name with ":"
480 Details : Only when device name contains ":/" will mount treat it as
486 Description: replace ptlrpcd with the statahead thread to interpret the async
487 statahead RPC callback
491 Frequency : on recovery
492 Description: I/O failures after umount during fail back
493 Details : if client reconnected to restarted server we need join to recovery
494 instead of find server handler is changed and process self
495 eviction with cancel all locks.
497 Severity : enhancement
499 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
501 Severity : enhancement
503 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
505 Severity : enhancement
507 Description: Upcall on Lustre log has been dumped
508 Details : Allow for a user mode script to be called once a Lustre log has
509 been dumped. It passes the filename of the dumped log to the
510 script, the location of the script can be specified via
511 /proc/sys/lnet/debug_log_upcall.
516 Description: avoid idr_remove called for id which is not allocated.
517 Details : Move assigment s_dev for clustered nfs to end of initialization,
518 to avoid problem with error handling.
523 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
524 Details : When connection is reused this not moved from CONN_UNUSED_HASH
525 into CONN_USED_HASH and this prodice warning when put connection
526 again in unused hash.
529 Severity : enhancement
531 Description: File striping can now be set to use an arbitrary pool of OSTs.
533 Severity : enhancement
535 Description: Export bytes_read/bytes_write count on OSC/OST.
539 Description: Early reply size mismatch, MGC loses connection
540 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
541 the connect flags are properly negotiated.
546 Description: quotas are not honored with O_DIRECT
547 Details : all writes with the flag O_DIRECT will use grants which leads to
548 this problem. Now using OBD_BRW_SYNC to guard this.
552 Description: add quota statistics
553 Details : 1. sort out quota proc entries and proc code.
554 2. add quota statistics
556 Severity : enhancement
558 Description: enable quota support for HEAD.
562 Description: Properly propagate oinfo flags from lov to osc for statfs
563 Details : restore missing copy oi_flags to lov requests.
565 Severity : enhancement
567 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
568 : l_getgroups(8), lst(8), routerstat(8)
569 Details : included man pages for llobdstat(8), llstat(8),
570 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
572 Severity : enhancement
574 Description: configure's --enable-quota should check the
575 : kernel .config for CONFIG_QUOTA
576 Details : configure is terminated if --enable-quota is passed but
577 : no quota support is in kernel
581 Description: Remove portals compatibility
582 Details : Remove portals compatibility, not interoperable with releases
587 Description: Resolve device initialization race
588 Details : Prevent proc handler from accessing devices added to the
589 obd_devs array but yet be intialized.
591 Severity : enhancement
593 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
595 Severity : enhancement
597 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
602 Description: Using wrong pointer in osc_brw_prep_request
603 Details : Access to array[-1] can produce panic if kernel compiled with
604 CONFIG_PAGE_ALLOC enabled
608 Description: Client runs out of low memory
609 Details : Consider only lowmem when counting initial number of llap pages
613 Description: *optional* service tags registration
614 Details : if the "service tags" package is installed on a Lustre node
615 When the filesystem is mounted, a local-node service tag will
616 be created. See http://inventory.sun.com/ for more information
617 about the Service Tags asset management system.
621 Description: Kernel BUG tries to release flock
622 Details : Lustre does not destroy flock lock before last reference goes
623 away. So always drop flock locks when client is evicted and
624 perform unlock regardless of successfulness of speaking to MDS.
628 Description: add refcount protection for osc callbacks, avoid panic on shutdown
632 Description: sanity test 65a fails if stripecount of -1 is set
633 Details : handle -1 striping on filesystem in ll_dirstripe_verify
638 Description: ASSERTION(CheckWriteback(page,cmd)) failed
639 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
642 Severity : enhancement
644 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
648 Description: do not process already freed flock
649 Details : flock can possibly be freed by another thread before it reaches
650 to ldlm_flock_completion_ast.
654 Description: LBUG during stress test
655 Details : Need properly lock accesses the flock deadlock detection list.
659 Description: oops in page fault handler
660 Details : kernel page fault handler can return two special 'pages' in error
661 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
665 Description: timeout with invalidate import.
666 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
667 should be handled by ptlrpcd. This produce long age waiting and
668 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
670 Severity : enhancement
672 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
676 Description: enable MGS and MDT services start separately
677 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
678 a co-located MGS without starting the MGS, which is a complement
679 to 'nosvc' mount option.
683 Frequency : after recovery
684 Description: precreate to many object's after del orphan.
685 Details : del orphan st in oscc last_id == next_id and this triger growing
686 count of precreated objects. Set flag LOW to skip increase count
687 of precreated objects.
691 Frequency : rare, on clear nid stats
692 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
693 Details : when clean nid stats sometimes try destroy live entry,
694 and this produce panic in free.
698 Description: Stack overflow during MDS log replay
699 ease stack pressure by using a thread dealing llog_process.
703 Description: wait until IO finished before start new when do lock cancel.
704 Details : VM protocol want old IO finished before start new, in this case
705 need wait until PG_writeback is cleared until check dirty flag and
706 call writepages in lock cancel callback.
708 Severity : enhancement
710 Description: using special macro for print time and cleanup in includes.
714 Description: mds_mfd_close() ASSERTION(rc == 0)
715 Details : In mds_mfd_close(), we need protect inode's writecount change
716 within its orphan write semaphore to prevent possible races.
720 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
724 Frequency : rare, on shutdown ost
725 Description: don't hit live lock with umount ost.
726 Details : shrink_dcache_parent can be in long loop with destroy dentries,
727 use shrink_dcache_sb instead.
731 Description: don't panic with use echo client
732 Details : echo client pass NULL as client nid pointer and this produce null
737 Description: fix build on ppc32
738 Details : compile code with -m64 flag produce wrong object file for ppc32.
742 Description: add message levels for liblustreapi
746 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
747 Details : Move the 'good_osts' check before the 'total_bavail' check. This
748 will result in an -EAGAIN and in the exit call path we call
749 alloc_rr() which will with increasing aggressiveness attempt to
750 aquire precreated objects on the minimum number of required OSCs.
754 Description: Use old size assignment to avoid deadlock
755 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
756 the scheduling while holding a spinlock. We do not need locking
757 for size in ll_update_inode() because size is only updated from
758 the MDS for directories or files without objects, so there is no
759 other place to do the update, and concurrent access to such inodes
760 are protected by the inode lock.
764 Description: resolve "_IOWR redefined" build error on SLES10
768 Description: dump the memory debugging after all modules are unloaded to
769 suppress false negative in conf_sanity test 39
771 Severity : enhancement
773 Description: build kernel-ib packages for OFED 1.3 in our release cycle
778 Description: fix SLES kernel versioning
779 Details : the kernel version for our SLES 10 kernel did not include a "-"
780 before the "smp" at the end. while this was not a problem in
781 general, it did mean that software trying to use the kernel
782 version to try to detect a vendor specific kernel would fail.
783 this was most evident by the OFED build scripts.
787 Description: Don't update lov_desc members until making sure they are valid
788 Details : When updating lov_desc members via proc fs, need fix their
789 validities before doing the real update.
793 Description: don't put request into delay list while invalidate in flight.
794 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
795 import in flight. this produce timeout for invalidate and sometimes
796 can cause stale data.
800 Frequency : on ppc only
801 Description: not convert ost objects for directory because it's not exist.
802 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
806 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
808 Severity : enhancement
810 Description: Update to support 2.6.22.14 vanilla kernel.
814 Frequency : rare, on recovery
815 Description: read procfs can produce deadlock in some situation
816 Details : Holding lprocfs lock which send rpc can produce block for destroy
817 obd objects and this also block reconnect with -EALREADY. This
818 isn't fix all lprocfs bugs - but make it rare.
820 Severity : enhancement
822 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
825 Frequency : frequent on X2 node
827 Description: mdc_set_open_replay_data LBUG
828 Details : Set replay data for requests that are eligible for replay.
832 Description: lustre_mgs: operation 101 on unconnected MGS
833 Details : When MGC is disconnected from MGS long enough, MGS will evict the
834 MGC, and late on MGC cannot successfully connect to MGS and a lot
835 of the error messages complaining that MGS is not connected.
839 Frequency : on network error
840 Description: panic with double free request if network error
841 Details : mdc_finish_enqueue is finish request if any network error ocuring,
842 but it's true only for synchronus enqueue, for async enqueue
843 (via ptlrpcd) this incorrect and ptlrpcd want finish request
846 Severity : enhancement
848 Description: client-side metadata stat-ahead during readdir(directory readahead)
849 Details : perform client-side metadata stat-ahead when the client detects
850 readdir and sequential stat of dir entries therein
853 Frequency : on start mds
855 Description: Implement get_info(last_id) in obdfilter.
858 Frequency : occasional
860 Description: Correctly check stale fid, not start epoch if ost not support SOM
861 Details : open with flag O_CREATE need set old fid in op_fid3 because
862 op_fid2 was overwritten with new generated fid, but mds can answer
863 with one of these two fids and both is not stale. Setattr
864 incorrectly started an epoch and assume will be called
865 done_writing, but without SOM done_writing ever being called.
868 Frequency : rare, depends on device drivers and load
870 Description: MDS or OSS nodes crash due to stack overflow
871 Details : Code changes in 1.8.0 increased the stack usage of some functions.
872 In some cases, in conjunction with device drivers that use a lot
873 of stack the MDS (or possibly OSS) service threads could overflow
874 the stack. One change which was identified to consume additional
875 stack has been reworked to avoid the extra stack usage.
878 Frequency : occasional
880 Description: Do not fail import if osc_interpret_create gets -EAGAIN
881 Details : If osc_interpret_create got -EAGAIN it immediately exits and
882 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
883 oscc_has_objects and see OSC has no objests and call
884 oscc_internal_create to resend create request.
886 Severity : enhancement
888 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
890 Severity : enhancement
892 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
895 Frequency : very rare
897 Description: Fix replay if there is an un-replied request and open
898 Details : In some cases, older replay request will revert the
899 mcd->mcd_last_xid on MDS which is used to record the client's
902 Severity : enhancement
904 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
906 Severity : enhancement
908 Description: Add rhel5 support to HEAD.
910 Serverity : enhancement
912 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
917 Description: Don't allow skipping OSTs if index has been specified.
918 Details : Don't allow skipping OSTs if index has been specified, make
919 locking in internal create lots better.
923 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
924 Details : access to bitfield in structure is always rounded to long
925 and this produce problem with not atomic change any bit.
929 Description: Lustre make rpms failed.
930 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
931 builds Lustre rpms from tarball.
934 Frequency : rare on shutdown ost
936 Description: If llog cancel was not send before clean_exports phase, this can
937 produce deadlock in llog code.
938 Details : If llog thread has last reference to obd and call class_import_put
939 this produce deadlock because llog_cleanup_commit_master wait when
940 last llog_commit_thread exited, but this never success because was
941 called from llog_commit_thread.
945 Description: allow userland application know is lost one of stripes.
946 Details : fill lvb_blocks with error code on ost and return it to
947 application if error flag found.
951 Description: NULL lov_tgts causing MDS oops
952 Details : more safe checks for NULL lov_tgts for avoid oops.
954 Severity : enhancement
956 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
960 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
961 Details : If an OST is down the MDS will hang indefinitely in
962 obd_statfs_rqset() waiting for the statfs data. While for
963 MDS QOS usage of statfs, it should not stuck in waiting.
965 Severity : enhancement
967 Description: remote_acl support
968 Details : Support ACL-based permission check for remote user.
969 Support setfacl/getfacl for remote user with the utils
970 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
971 the system "{s,g}etfacl" utils.
973 Severity : enhancement
975 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
977 Severity : enhancement
979 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
983 Description: make mgs_setparam() handle fsname containing dash
984 Details : fsname containing a dash does not work with lctl conf_param
986 Severity : enhancement
988 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
990 Severity : enhancement
992 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
996 Frequency : rare, at shutdown
997 Description: access already free / zero obd_namespace.
998 Details : if client_disconnect_export was called without force flag set,
999 and exist connect request in flight, this can produce access to
1000 NULL pointer (or already free pointer) when connect_interpret
1001 store ocd flags in obd_namespace.
1005 Frequency : only at startup
1006 Description: not alloc memory with spinlock held.
1007 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1008 if any spinlock held.
1010 Severity : enhancement
1012 Description: make lustre randomly fail allocating memory
1013 Details : Make lustre randomly failed allocating memory for testing purpose.
1015 Severity : enhancement
1017 Description: lost problems with lov objid file
1018 Details : Fixes some scability and access to not inited memory problems
1019 in work with lov objdid file.
1024 Description: lfs find does not continue on file error
1025 Details : Continue other files processing when a file/dir is absent.
1029 Description: Inconsistent usage of lustre_pack_reply()
1030 Details : Standardize the usage of lustre_pack_reply() such that it
1031 always generate a CERROR on failure.
1034 Frequency : occasional
1036 Description: MDS hang or stay in waiting lock
1037 Details : If client receive lock with CBPENDING flag ldlm need send lock
1038 cancel as separate rpc, to avoid situation when cancel request
1039 can't processed due all i/o threads stay in wait lock.
1043 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1047 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1051 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1055 Description: Update to RHEL4 latest kernel.
1057 Severity : enhancement
1059 Description: Build SLES10 patchless client fails
1060 Details : The configure was broken by run ./configure with
1061 --with-linux-obj=.... argument for patchless client. When the
1062 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1063 can't search header adequately. Use absolute path such as
1064 -I($LINUX)/include instead.
1068 Description: interrupt oig_wait produce painc on resend.
1069 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1070 private set, and this produce situation when rq_ptlrpcd_data not
1071 copyed to new allocated request and triggered LBUG on assert
1072 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1073 ptlrpcd set if request is changed and can be safety changed to use
1078 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1079 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1080 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1081 these macros and replace them with their difinitions instead.
1085 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1086 Details : Update which_patch & target file for SLES10 latest kernel.
1089 Description: Remove client patches from SLES 10 kernel.
1090 Details : This causes SLES 10 clients to behave as patchless clients
1091 even on a Lustre-patched (server) kernel.
1093 Severity : enhancement
1095 Description: self-adjustable client's lru lists
1096 Details : use adaptive algorithm for managing client cached locks lru
1097 lists according to current server load, other client's work
1098 pattern, memory activities, etc. Both, server and client
1099 side namespaces provide number of proc tunables for controlling
1102 Severity : enhancement
1104 Description: light-weight GSS support
1105 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1106 and reduce performance overhead.
1108 Severity : enhancement
1110 Description: Linux keyring support
1111 Details : Support using service of Linux keyring for Lustre GSS internal
1112 context refresh/cache mechanism.
1116 Description: Fix errors in lfs documentation
1117 Details : Fixes man pages
1121 Description: don't use GFP_* in generic Lustre code.
1122 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1126 Description: obdclass is limited by single OBD_ALLOC(idarray)
1127 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1131 Description: warnings with build patchless client with vanila 2.6.19 and up
1132 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1133 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1137 Frequency : Only for SLES
1138 Description: sanity_quota fail test_1
1139 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1140 file, which makes TSTID[2] inunique.
1144 Frequency : Only for SLES9
1145 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1146 Details : Update target/ChangeLog/which_patch .
1150 Frequency : Only for test_10 in sanity-quota.sh
1151 Description: a bug in quota test code
1152 Details : checking a test flag in a wrong place
1156 Frequency : Only for test_13 in sanity-quota.sh
1157 Description: a bug in quota test script
1158 Details : shouldn't check overfull allocated quota
1162 Description: jbd statistics
1163 Details : Port older jbd statistics patch for sles10
1165 Severity : enhancement
1167 Description: Add printing inode info into message about error in writepage.
1172 Description: reply_lock_interpret crash due to race with it and lock cancel.
1173 Details : Do not replay locks that are being cancelled. Do not reference
1174 locks by their address during replay, just by their handle.
1178 Frequency : When flocks are used.
1179 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1180 Details : Flock locks might destroy just granted lock if it could be merged
1181 with another existing flock, this is done in completion handler,
1182 so teach ldlm_cli_enquque_fini that this is a valid case for
1188 Description: Oops in read and write path when failing to allocate lock.
1189 Details : Check if lock allocation failed and return error back.
1193 Description: lstripe command fails for valid OST index
1194 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1195 instead of lov->desc.ld_active_tgt_count.
1199 Description: sanity.sh failed test 103
1200 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1205 Description: df doesn't work properly if diskfs blocksize != 4K
1206 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1210 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1211 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1213 Severity : enhancement
1215 Description: Adaptive timeouts
1216 Details : RPC timeouts adapt to changing server load and network
1217 conditions to reduce resend attempts and improve recovery time.
1221 Description: llapi_file_create() does not allow some changes
1222 Details : add llapi_file_open() that allows specifying the mode and
1223 open flags, and also returns an open file handle.
1227 Description: merge and cleanup kernel patches.
1228 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1232 Description: Add journal checksum support.(Kernel part)
1233 Details : The journal checksum feature adds two new flags i.e
1234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1235 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1236 indicates that the commit block contains the checksum for
1237 the blocks described by the descriptor blocks. Now commit
1238 record can be sent to disk without waiting for descriptor
1239 blocks to be written to disk. This behavior is controlled
1240 using JBD2_FEATURE_ASYNC_COMMIT flag.
1244 Description: OSS needs mutliple precreate threads
1245 Details : Add ability to start more than one create thread per OSS.
1249 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1250 Details : execute lfs setstripe on client
1254 Description: mds_obd_create error creating tmp object
1255 Details : When the user sets quota on root, llog will be affected and can't
1256 create files and write files.
1259 Frequency : Always on ia64 patchless client, and possibly others.
1261 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1262 Details : This allows the patchless client to be loaded on architectures
1263 without this export.
1267 Description: RedHat Update kernel for RHEL5
1268 Details : Add the kernel config file for RHEL5.
1272 Description: RedHat Update kernel for RHEL5
1273 Details : Modify the kernel config file more closer RHEL5.
1277 Description: Build failure against Centos5 (RHEL5)
1278 Details : Define PAGE_SIZE when it isn't present.
1282 Description: test_7 on sanity-quota.sh can't be used on separate machines
1283 Details : add facet to handle it
1287 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1288 Details : with NFS, the anon dentry's parent was set to itself in
1289 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1290 corresponding dentry other than use rec->ur_name.
1292 Severity : enhancement
1294 Description: lfs setstripe enhancement
1295 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1299 Description: enable data checksumming by default
1300 Details : enable checksum by default, allow --disable-checksum
1301 configure option and "-o nochecksum" mount option. Checksums
1302 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1306 Description: lustre support for RHEL5
1307 Details : Add support for RHEL5.
1311 Description: Client eviction due to failover config
1312 Details : after a connection loss, the lustre client should attempt to
1313 reconnect to the last active server first before trying the
1314 other potential connections.
1318 Description: when mds and osts use different quota unit(32bit and 64bit),
1319 quota will be released repeatly.
1320 Details : void sending multiple quota reqs to mds, which will keep the
1321 status between the reqs.
1325 Description: osts not allocated evenly to files
1326 Details : change the condition to increase offset_idx
1329 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1331 Description: Data corruption for OSTs that are formatted larger than 2TB
1333 Details : When generating the bio request for lustre file writes the
1334 sector number would overflow a temporary variable before being
1335 used for the IO. The data reads correctly from Lustre (which
1336 will overflow in a similar manner) but other file data or
1337 filesystem metadata may be corrupted in some cases.
1341 Description: Tune the kernel for good SCSI performance.
1342 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1343 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1348 Description: rewrite ext2-derived code in obdclass/uuid.c
1349 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1350 from scratch preserving functionality.
1354 Description: "lfs find -obd UUID" prints directories
1355 Details : "lfs find -obd UUID" will return all directory names instead
1356 of just file names. It is incorrect because the directories
1357 do not reside on the OSTs.
1361 Description: use i_size_read and i_size_write in 2.6 port
1362 Details : replace inode->i_size access with i_size_read/write()
1365 Frequency : when removing large files
1367 Description: scheduling issue during removal of large Lustre files
1368 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1369 It causes scheduling issues when removing large files (17TB in the
1373 Frequency : only with liblustre clients on XT3
1375 Description: evictions taking too long
1376 Details : allow llrd to evict clients directly on OSTs
1379 Frequency : only on ppc
1381 Description: /proc/fs/lustre/devices broken on ppc
1382 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1383 We should make sure the type of 'index' is loff_t and then cast
1384 explicitly as needed below. Do not assign an explicitly cast
1388 Frequency : only for rhel5
1390 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1391 Details : Modify the target file & which_kernel.
1393 Severity : enhancement
1395 Description: omit set fsid for export NFS
1396 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1397 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1401 Frequency : Always, for kernels after 2.6.16
1402 Description: Fix warning idr_remove called for id=.. which is not allocated.
1403 Details : Last kernels save old s_dev before kill super and not allow
1404 to restore from callback - restore it before call kill_anon_super.
1408 Description: buffer overruns could theoretically occur
1409 Details : llapi_semantic_traverse() modifies the "path" argument by
1410 appending values to the end of the origin string, and a
1411 overrun may occur. Adding buffer overrun check in liblustreapi.
1415 Description: Fix error on 'ls .' at the top of the Lustre mount.
1416 Details : Don't revalidate dentry if it is a root dentry.
1420 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1421 Details : Modify vm-tunables-rhel4.patch.
1425 Description: Kernel config for 2.6.18-vanilla.
1426 Details : Modify targets/2.6-vanilla.target.in.
1427 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1428 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1429 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1430 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1434 Description: improve handling recoverable errors
1435 Details : if request processig with error which can be recoverable on server
1436 request should be resend, otherwise page released from cache and
1441 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1442 Details : Modify target file & which_patch.
1443 A flaw was found in the IA32 system call emulation provided
1444 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1445 value could be stored in the %RAX register, which could trigger an
1446 out-of-bounds system call table access. An untrusted local user
1447 could exploit this flaw to run code in the kernel
1448 (ie a root privilege escalation). (CVE-2007-4573).
1452 Description: O_DIRECT bypasses client statistics.
1453 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1454 read_ahead_stats were updated. copied the stats section from
1455 osc_send_oap_rpc() into async_internal().
1459 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1463 Description: change order of libsysio includes
1464 Details : '#include sysio.h' should always come before '#include xtio.h'
1468 Description: handle "serious error: objid * already exists" more gracefully
1469 Details : If LAST_ID value on disk is smaller than the objects existing in
1470 the O/0/d* directories, it indicates disk corruption and causes an
1471 LBUG(). If the object is 0-length, then we should use the existing
1472 object. This will help to avoid a full fsck in most cases.
1474 Severity : enhancement
1476 Description: adapt the lustre_config script to support the upgrade case
1477 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1478 server targets to 1.6 in parallel.
1482 Description: To avoid grant space > avaible space when the disk is almost
1483 full. Without this patch you might see the error "grant XXXX >
1484 available" or some LBUG about grant, when the disk is almost
1486 Details : In filter_check_grant, for non_grant cache write, we should
1487 check the left space by if (*left > ungranted + bytes), instead
1488 of (*left > ungranted), because only we are sure the left space
1489 is enough for another "bytes", then the ungrant space should be
1490 increase. In client, we should update cl_avail_grant only there
1491 is OBD_MD_FLGRANT in the reply.
1495 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1496 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1497 described in CVE-2007-4573. This problem could allow untrusted
1498 local users to gain root access.
1501 Frequency : when using O_DIRECT and quotas
1503 Description: Incorrect file ownership on O_DIRECT output files
1504 Details : block usage reported by 'lfs quota' does not take into account
1505 files that have been written with O_DIRECT.
1510 Description: touch file failed when fs is not full
1511 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1512 otherwise we can get ENOSP while fs is not full.
1516 Description: parallel lock callbacks
1517 Details : Instead of sending blocking and completion callbacks as separated
1518 requests, adding them to a set and sending in parallel.
1521 Frequency : only for Cray XT3
1522 Bugzilla : 12829/13455
1523 Description: Changing primary group doesn't change the group lustre assigns to
1525 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1526 overridden with the primary group provided by the group upcall,
1527 whereas the supplied fsgid can be trusted if it is in the list of
1528 supplementary groups returned by the group upcall.
1530 Severity : enhancement
1532 Description: Allow masking D_WARNING, D_ERROR messages from console
1533 Details : Console messages can now be disabled via lnet.printk.
1537 Description: User code with malformed file open parameter crashes client node
1538 Details : Before packing join_file req, all the related reference should be
1539 checked carefully in case some malformed flags cause fake
1540 join_file req on client.
1544 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1545 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1546 mode is equal to the granted mode. However, it should also check
1547 that LDLM_FL_CP_REQD is not set.
1551 Description: Heavy nfs access might result in deadlocks
1552 Details : After ELC code landed, it is now improper to enqueue any mds
1553 locks under och_sem, because enqueue might want to decide to
1554 cancel open locks for same inode we are holding och_sem for.
1558 Description: Client eviction while running blogbench
1559 Details : A lot of unlink operations with concurrent I/O can lead to a
1560 deadlock causing evictions. To address the problem, the number of
1561 oustanding OST_DESTROY requests is now throttled to
1562 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1563 callbacks are processed in priority.
1567 Description: enable ACLs on MDS by default
1568 Details : ACLs must be enabled on MDS by default.
1571 Frequency : PPC/PPC64 only
1573 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1574 Details : fix duplicated definitions between asm-ppc64/types.h and
1575 lustre_types.h on PPC.
1578 Frequency : PPC/PPC64 only
1580 Description: asm-ppc/segment.h does not exist
1581 Details : fix compile issue on PPC.
1585 Description: better handle error messages in extents code
1588 Frequency : RHEL4 only
1590 Description: mkfs is very slow on IA64/RHEL4
1591 Details : A performance regression has been discovered in the MPT Fusion
1592 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1593 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1594 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1596 Severity : enhancement
1598 Description: SNMP support enhancement
1599 Details : Adding total number of sampled request for an MDS node in snmp
1602 Severity : enhancement
1604 Description: Optimize ldlm waiting list processing for PR extent locks
1605 Details : When processing waiting list for read extent lock and meeting read
1606 lock that is same or wider to it that is not contended, skip
1607 processing rest of the list and immediatelly return current
1608 status of conflictness, since we are guaranteed there are no
1609 conflicting locks in the rest of the list.
1613 Description: Time out and refuse to reconnect
1614 Details : When the failover node is the primary node, it is possible
1615 to have two identical connections in imp_conn_list. We must
1616 compare not conn's pointers but NIDs, otherwise we can defeat
1617 connection throttling.
1621 Description: port llog fixes from b1_6 into HEAD
1622 Details : Port llog reference couting and some llog cleanups from b1_6
1623 (bug 10800) into HEAD, for protect from panic and access to already
1624 free llog structures.
1628 Description: Detect stride IO mode in read-ahead
1629 Details : When a client does stride read, read-ahead should detect that and
1630 read-ahead pages according to the detected stride pattern.
1634 Description: data checksumming impacts single node performance
1635 Details : add support for several checksum algorithm. Currently, only CRC32
1636 and Adler-32 are supported. The checksum type can be changed on
1637 the fly via /proc/fs/lustre/osc/*/checksum_type.
1641 Description: use adler32 for page checksums
1642 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1647 Description: build for x2 fails
1648 Details : fix compile issue on Cray systems.
1652 Description: Properly match for duplicate locks
1653 Details : Due to different lock order from skiplists code, we need to
1654 traverse entire list for now
1657 Frequency : only on PPC/SLES10
1659 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1660 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1661 be defined. Add a hack in mkfs_lustre.c to work around this compile
1666 Description: LASSERT on MDS when client holding flock lock dies
1667 Details : ldlm pool logic depends on number of granted locks equal to
1668 number of released locks which is not true for flock locks, so
1669 just exclude such locks from consideration.
1673 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1674 Details : Use fsfilt_sync() for both the whole filesystem sync and
1675 individual file sync to eliminate dangerous inode locking
1676 with I_LOCK that can lead to a deadlock.
1680 Description: performance in 1.6.3
1681 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1682 sure that 1MB requests are not fragmented by the block layer.
1684 Severity : enhancement
1686 Description: organize the server-side client stats on per-nid basis
1687 Details : Change the structure of stats under obdfilter and mds to
1695 The "uuid"s file would list the uuids of _active_ exports.
1696 And the clear entry is to clear all stats and stale nids.
1698 Severity : enhancement
1700 Description: eliminate client locks in face of contention
1701 Details : file contention detection and lockless i/o implementation
1702 for contended files.
1706 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1709 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1711 Description: MDS may lose file striping (and hence file data) in some cases
1712 Details : If there are additional extended attributes stored on the MDS,
1713 in particular ACLs, SELinux, or user attributes (if user_xattr
1714 is specified for the client mount options) then there is a risk
1715 of attribute loss. Additionally, the Lustre file striping
1716 needs to be larger than default (e.g. striped over all OSTs),
1717 and an additional attribute must be stored initially in the
1718 inode and then increase in size enough to be moved to the
1719 external attribute block (e.g. ACL growing in size) for file
1724 Description: skiplist implementation simplification
1725 Details : skiplists are used to group compatible locks on granted list
1726 that was implemented as tracking first and last lock of each
1727 lock group the patch changes that to using doubly linked lists
1731 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1732 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1733 because LASSERT on both the data supplied by a client, and the
1734 data on disk is dangerous and incorrect.
1736 Severity : enhancement
1738 Description: Slow truncate/writes to huge files at high offsets.
1739 Details : Directly associate cached pages to lock that protect those pages,
1740 this allows us to quickly find what pages to write and remove
1741 once lock callback is received.
1745 Description: more ldlm soft lockups
1746 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1747 starve other threads from the resource lock for a long time in
1748 case of long waiting queue, so change the debug level from
1749 D_OTHER to the less frequently used D_INFO.
1751 Severity : enhancement
1753 Description: add -gid, -group, -uid, -user options to lfs find
1757 Description: Hung threads in invalidate_inode_pages2_range
1758 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1759 once one is completed. As a result, some RPCs are stuck in the
1760 queue and are never sent.
1764 Description: filter threads hungs on waiting journal commit
1765 Details : Cleanup filter group llog code, then only filter group llog will
1766 be only created in the MDS/OST syncing process.
1770 Description: Procfs and llog threads access destoryed import sometimes.
1771 Details : Sync the import destoryed process with procfs and llog threads by
1772 the import refcount and semaphore.
1774 Severity : enhancement
1776 Description: openlock cache of b1_6 port to HEAD
1781 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1782 Details : If insertion of an extent fails, then discard the inode
1783 preallocation and free data blocks else it can lead to duplicate
1788 Description: don't always update ctime in ext3_xattr_set_handle()
1789 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1790 In some cases the ctime should not be updated, for example for
1791 2.0->1.8 compatibility it is necessary to delete an xattr and it
1792 should not update the ctime.
1796 Bugzilla : 15713/16362
1797 Description: Assertion in iopen_connect_dentry in 1.6.3
1798 Details : looking up an inode via iopen with the wrong generation number can
1799 populate the dcache with a disconneced dentry while the inode
1800 number is in the process of being reallocated. This causes an
1801 assertion failure in iopen since the inode's dentry list contains
1802 both a connected and disconnected dentry.
1806 Description: assertion failure in ldlm_handle2lock()
1807 Details : fix a race between class_handle_unhash() and class_handle2object()
1808 introduced in lustre 1.6.5 by bug 13622.
1813 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1814 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1819 Description: oops in sd_iostats_seq_show()
1820 Details : unloading/reloading the scsi low level driver triggers a kernel
1821 bug when trying to access the sd iostat file.
1826 Description: Kernel panics during QLogic driver reload
1827 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1828 patch, causing memory corruption.
1833 Description: journal_dev option does not work in b1_6
1834 Details : pass mount option during pre-mount.
1836 Severity : enhancement
1838 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1839 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1840 extent information of a file. It can be used to map logical blocks
1841 in a file to physical blocks in the block device.
1845 Description: LDLM soft lockups - improvement
1846 Details : It is be possible to send the lock handle along with each read
1847 or write request because the client is already doing a lock match
1848 itself so there isn't any reason the OST should have to re-do that
1852 Frequency : only on Cray X2
1854 Description: X2 build failures
1855 Details : fix build failures on Cray X2.
1859 Description: xid & resent requests
1860 Details : Initialize RPC XID from clock at startup (randomly if clock is
1863 Severity : enhancement
1865 Description: Add lustre_start utility to start or stop multiple Lustre servers
1870 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1871 Details : In case of memory pressure, list_del() can be called twice on
1872 req->rq_history_list, causing a kernel oops.
1876 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1877 Details : fix stack overflow in the distributed lock manager by defering
1878 export eviction after a failed AST to the elt thread instead of
1879 handling it in the dlm interpret routine.
1883 Description: Convert some comments to new format.
1884 Details : Update documenting comments to match doxygen conventions.
1888 Description: Grammar fixes.
1889 Details : A couple of trivial spelling fixes.
1893 Description: OSD_COUNTERS-mandatory
1894 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1895 to simplify the code.
1899 Description: simplify cmm_device freeing logic.
1900 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1904 Description: Add lockdep support to dt_object_operations locking interface.
1905 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1906 parameter indicating lock ordering. Update mdd code to use new
1911 Description: Introduce failloc constants for lockless IO tests.
1912 Details : Add two new failloc constants to test lockless IO. Only one of
1913 them in implemented---another is checked in yet to be landed
1918 Description: Add lockdep support for inode mutex.
1919 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1920 in the situations where Lustre has to lock more than one inode
1925 Description: Add optional invariants checking support.
1926 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1927 configure switch. This macro is to be used for consistency and
1928 sanity checks that are too expensive to be left in `production'
1933 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1934 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1939 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1940 Details : Use spin_lock_nested() in (the only) situation where more than
1941 one ldlm_lock is locked simultaneously. Also, fix possible
1942 dead-lock in ldlm_lock_change_resource() by enforcing particular
1947 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1948 Details : Instead of specifying each ldlm_lock call-back through separate
1949 parameter, wrap them into struct ldlm_callback_suite.
1953 Description: Kill join_lru obd method and its callers.
1954 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1955 in memory---a requirement implemented through ->o_join_lru() obd
1956 method in HEAD. Remove this method and its users.
1960 Description: Add asynchronous ldlm ENQUEUE completion handler.
1961 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1962 case that stock ldlm_completion_ast() cannot handle as it waits
1963 until lock is granted. Introduce new ldlm_completion_ast_async()
1964 for this. Also comment ldlm_completion_ast().
1968 Description: ldlm_error <-> errno conversion.
1969 Details : Add functions to map (rather arbitrary) between LDLM error codes
1970 and standard errno values. CLIO needs this to prevent LDLM specific
1971 constants from escaping ldlm and osc.
1975 Description: Kill unused ldlm_handle2lock_ns() function.
1976 Details : Kill unused ldlm_handle2lock_ns() function.
1980 Description: Add lu_ref support to ldlm_lock
1981 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1982 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1983 and ldlm_resource. LDLM interface has to be changed, because code that
1984 releases a reference on a lock, has to "know" what reference this is.
1985 In the most frequent case
1987 lock = ldlm_handle2lock(handle);
1989 LDLM_LOCK_PUT(lock);
1991 no changes are required. When any other reference (received _not_ from
1992 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1993 instead of LDLM_LOCK_PUT().
1995 Arguably, changes are pervasive, and interface requires some discipline
1996 for proper use. On the other hand, it was very instrumental in finding
1997 a few leaked lock references.
2001 Description: Add ldlm_lock_addref_try().
2002 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2003 attempts to addref a lock that might be being canceled
2008 Description: Add ldlm_weigh_callback().
2009 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2010 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2011 hard-coded `number of pages' logic.
2015 Description: Add lockdep annotations to llog code.
2016 Details : Use appropriately tagged _nested() locking calls in the places
2017 where llog takes more than one ->lgh_lock lock.
2021 Description: Add loi_kms_set().
2022 Details : Wrap kms updates into a helper function.
2026 Description: Constify instances of struct lsm_operations.
2027 Details : Constify instances of struct lsm_operations.
2031 Description: lu_conf support.
2032 Details : On a server, a file system object is uniquely identified
2033 by a fid, which is sufficient to locate and load all object
2034 state (inode). On a client, on the other hand, more data are
2035 necessary instantiate an object. Change lu_object_find() and
2036 friends to take additional `lu_conf' argument describing object.
2037 Typically this includes layout information.
2041 Description: lu_context fixes.
2042 Details : Introduce new lu_context functions that are needed on the client
2043 side, where some system threads (ptlrpcd) are shared by multiple
2044 modules, and so cannot be stopped during module shutdown.
2048 Description: Add start and stop methods to lu_device_type_operations.
2049 Details : Introduce two new methods in lu_device_type_operations, that are
2050 invoked when first instance of a given type is created and last one
2051 is destroyed respectively. This is need by CLIO.
2055 Description: Add lu_ref support to struct lu_device.
2056 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2057 track leaked references.
2061 Description: Introduce lu_kmem_descr.
2062 Details : lu_kmem_descr and its companion interface allow to create
2063 and destroy a number of kmem caches at once.
2067 Description: Fix lu_object finalization race.
2068 Details : Fix a race between lu_object_find() finding an object and its
2069 concurrent finalization. This race is (most likely) not possible
2070 on the server, but might happen on the client.
2074 Description: Introduce lu_ref interface.
2075 Details : lu_ref is a debugging module allowing to track references to
2076 a given object. It is quite cpu expensive, and has to be
2077 explicitly enabled with --enable-lu_ref. See usage description
2082 Description: Factor lu_site procfs stats into a separate function.
2083 Details : Separate lu_site stats printing code into a separate function
2084 that can be reused on a client.
2088 Description: Constify instances of struct {lu,dt,md}_device_operations.
2089 Details : Constify instances of struct {lu,dt,md}_device_operations.
2093 Description: Introduce struct md_site and move meta-data specific parts of
2094 struct lu_site here.
2095 Details : Move md-specific fields out of struct lu_site into special struct
2096 md_site, so that lu_site can be used on a client.
2100 Description: Kill mdd_lov_destroy().
2101 Details : Remove unused mdd code.
2105 Description: Add st_block checking to multistat.c.
2106 Details : Add st_block checking to multistat.c.
2110 Description: Add lu_ref support to struct obd_device.
2111 Details : Add lu_ref tracking to obd_device.
2115 Description: Kill obd_set_fail_loc().
2116 Details : Remove unused code.
2120 Description: Add special type for ptlrpc_request interpret functions.
2121 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2122 there. Actual usage will be in CLIO.
2126 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2127 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2128 doesn't work with lockdep.
2132 Description: Add rwv.c test program.
2133 Details : New testing program exercising readv(2) and writev(2) (Qian).
2137 Description: Add sendfile.c test program.
2138 Details : New testing program exercising sendfile(2) (Jay).
2142 Description: Ratelimit a message that can be very frequent.
2143 Details : Ratelimit a memory allocation failure message that can
2148 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2149 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2153 Description: Liblustre build fixes.
2154 Details : Liblustre build fixes.
2158 Description: libcfs: add cfs_{need,cond}_resched() interface.
2159 Details : libcfs: add cfs_{need,cond}_resched() definition and
2160 implementations for Linux, NT, and liblustre.
2162 Severity : enhancement
2164 Description: More exported tunables for mballoc
2165 Details : Add support for tunable preallocation window and new tunables for
2166 large/small requests
2170 Description: Detect corruption of block bitmap and checking for preallocations
2171 Details : Checks validity of on-disk block bitmap. Also it does better
2172 checking of number of applied preallocations. When corruption is
2173 found, it turns filesystem readonly to prevent further corruptions.
2177 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2178 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2179 window to 0. If it does hit cache miss, and read-pattern is still
2180 stride-io mode, does not reset the stride window, but also does not
2181 increase the stride window length in this case.
2185 Frequency : only for big-endian servers
2186 Description: Check if system is big-endian while mounting fs with extents feature
2187 Details : Mounting a filesystem with extents feature will fail on big-endian
2188 systems since ext3-based ldiskfs is not supported on big-endian
2189 systems. This can be over-riden with "bigendian_extents" mount option.
2191 Severity : enhancement
2193 Description: The root squash functionality
2194 Details : A security feature, which is to prevent users from being able
2195 to mount lustre on their desktop, run as root, and delete
2196 all of the files in the filesystem. The goal is accomplished by
2197 remapping user id (UID) and group id (GID) of the root user to
2198 a UID and GID specified by the system administartor via Lustre
2199 configuration management server (MGS). The functionality also
2200 allows to specify sets of clients for which the remapping does
2205 Description: Excessive recovery window
2206 Details : With AT enabled, the recovery window can be excessively long (6000+
2207 seconds). To address this problem, we no longer use
2208 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2209 timeout no longer depends on the service time, it is set to
2210 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2211 time via pb_service_time.
2215 Description: Watchdog triggered on MDS failover
2216 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2217 the OSTs know that the MDS "UUID" can be reused for the same export
2218 from a different NID, so we do not need to wait for the export to be
2222 Frequency : rare, only if using MMP with Linux RAID
2224 Description: MMP doesn't work with Linux RAID
2225 Details : While using HA for Lustre servers with Linux RAID, it is possible
2226 that MMP will not detect multiple mounts. To make this work we
2227 need to unplug the device queue in RAID when the MMP block is being
2228 written. Also while reading the MMP block, we should read it from
2229 disk and not the cached one.
2231 Severity : enhancement
2233 Description: open file using fid
2234 Details : A file can be opened using just its fid, like
2235 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2238 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2241 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2242 Details : add proper locking for creation of jbd_4k slab cache
2246 Description: MMP check in ext3_remount() fails without displaying any error
2247 Details : When multiple mount protection fails during remount, proper error
2250 Severity : enhancement
2252 Description: Allow stripe size to be up to 4G-64k
2253 Details : Fix math logic to allow large stripe sizes.
2257 Description: add check for >8TB ldiskfs filesystems
2258 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2259 Don't allow >8TB ldiskfs filesystems to be mounted without
2260 force_over_8tb mount option
2265 Descriptoin: OSS complains "too many blocks" and crashes
2266 Details : In filter_commitrw_write():
2267 1. lquota_chkquota();
2268 2. obdo_from_inode();
2269 3. lquota_pending_commit();
2270 "2." has changed uid and gid in this case, then caused
2271 lquota_chkquota() and lquota_pending_commit() to use
2272 different uid and gid. That is the root of the bug.
2276 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2277 Details : Detect and print out the 'stripe count = -1' case properly.
2279 --------------------------------------------------------------------------------
2281 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2283 * Support for kernels:
2284 2.6.5-7.283 (SLES 9),
2285 2.6.9-55.EL (RHEL 4),
2286 2.6.16.46-0.14 (SLES 10),
2287 2.6.18.8 vanilla (kernel.org)
2288 * Client support for unpatched kernels:
2289 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2290 2.6.16 - 2.6.22 vanilla (kernel.org)
2291 * Due to recently discovered recovery problems, we do not recommend
2292 using patchless RHEL 4 clients with this or any earlier release.
2293 * Recommended e2fsprogs version: 1.39.cfs8
2294 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2295 * Starting with this release, the ldiskfs backing filesystem required
2296 by Lustre is now in its own package, lustre-ldiskfs. This package
2297 should be installed. It is versioned separately from Lustre and
2298 may be released separately in future.
2302 Description: block reactivating mgc import until all deactivates complete
2303 Details : Fix race when failing back MDT/MGS to itself (testing)
2306 Frequency : at statup only
2308 Description: mds_lov_synchronize race leads to various problems
2309 Details : simultaneous MDT->OST connections at startup can cause the
2310 sync to abort, leaving the OSC in a bad state.
2312 Severity : enhancement
2314 Description: add optional extra BUILD_VERSION info
2315 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2316 to override the lustre version.
2319 Frequency : 2.6.18 servers only
2321 Description: ll_kern_mount() doesn't release the module reference
2322 Details : The ldiskfs module reference count never drops down to 0
2323 because ll_kern_mount() doesn't release the module reference.
2328 Description: server LBUG when using old ost_num_threads parameter
2329 Details : Accept the old ost_num_threads parameter but warn that it
2330 is deprecated, and fix an off-by-one error that caused an LBUG.
2335 Description: Transient SCSI error results in persistent IO issue
2336 Details : iobuf->dr_error is not reinitialized to 0 between two
2340 Frequency : sometimes when underlying device returns I/O errors
2342 Description: OSTs not going read-only during write failures
2343 Details : OSTs are not remounted read-only when the journal commit threads
2344 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2345 instead of the ext3 wrappers.
2349 Description: poor connect scaling with increasing client count
2350 Details : Don't run filter_grant_sanity_check for more than 100 exports
2351 to improve scaling for large numbers of clients.
2354 Frequency : SLES10 only
2356 Description: sanity-quota.sh quotacheck failed: rc = -22
2357 Details : Quotas cannot be enabled on SLES10.
2360 Frequency : liblustre clients only
2362 Description: getdirentries does not give error when run on compute nodes
2363 Details : getdirentries does not fail when the size specified as an argument
2364 is too small to contain at least one entry
2366 Severity : enhancement
2368 Description: Add LNET router traceability for debug purposes
2369 Details : If a checksum failure occurs with a router as part of the
2370 IO path, the NID of the last router that forwarded the bulk data
2371 is printed so it can be identified.
2376 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2377 Details : Due to a race condition, liblustre clients were occasionally
2378 evicted incorrectly.
2380 Severity : enhancement
2382 Description: lfs setstripe use optional parameters instead of postional
2385 Severity : enhancement
2387 Description: Nanosecond timestamp support for ldiskfs
2388 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2389 resolution timestamps. There is not yet support for this at
2390 the Lustre filesystem level.
2393 Frequency : during server recovery
2395 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2396 Details : request with rq_no_resend flag not awake l_wait_event if they get
2400 Frequency : nfs export on patchless client
2402 Description: connectathon hang when test nfs export over patchless client
2403 Details : Disconnected dentry cannot be found with lookup, so we do not need
2404 to unhash it or make it invalid
2407 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2408 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2409 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2414 Description: open req refcounting wrong on reconnect
2415 Details : If reconnect happened between getting open reply from server and
2416 call to mdc_set_replay_data in ll_file_open, we will schedule
2417 replay for unreferenced request that we are about to free.
2418 Subsequent close will crash in variety of ways.
2419 Check that request is still eligible for replay in
2420 mdc_set_replay_data().
2425 Description: disable writes to filesystem when reading health_check file
2426 Details : the default for reading the health_check proc file has changed
2427 to NOT do a journal transaction and write to disk, because this
2428 can cause reads of the /proc file to hang and block HA state
2429 checking on a healthy but otherwise heavily loaded system. It
2430 is possible to return to the previous behaviour during configure
2431 with --enable-health-write.
2433 Severity : enhancement
2435 Description: 64-bit inode version
2436 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2437 to the inode. This will be required for version-based recovery.
2442 Description: MDS fails to start if a duplicate client export is detected
2443 Details : in some rare cases it was possible for a client to connect to
2444 an MDS multiple times. Upon recovery the MDS would detect this
2445 and fail during startup. Handle this more gracefully.
2447 Severity : enhancement
2449 Description: Add -o localflock option to simulate old noflock
2451 Details : This will achieve local-only flock/fcntl locks
2457 Description: log_commit_thread vs filter_destroy race leads to crash
2458 Details : Take import reference before releasing llog record semaphore
2463 Description: Wrong request locking in request set processing
2464 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2465 imp_delayed_list, in this place should be used imp_lock.
2468 Frequency : when reconnection
2470 Description: Grant Leak when osc reconnect to OST
2471 Details : When osc reconnect ost, OST(filter) should check whether it
2472 should grant more space to client by comparing fed_grant and
2473 cl_avail_grant, and return the granted space to client instead
2474 of "new granted" space, because client will call osc_init_grant
2475 to update the client grant space info.
2478 Frequency : when client reconnect to OST
2480 Description: Grant Leak when osc do resend and replay bulk write
2481 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2482 bulk write request, because the grant info while be sync between
2483 OSC and OST when reconnect, and we should ignore the grant info
2484 these of resend/replay write req.
2489 Description: Grant space more than avaiable left space sometimes.
2490 Details : When then OST is about to be full, if two bulk writing from
2491 different clients came to OST. Accord the avaliable space of the
2492 OST, the first req should be permitted, and the second one
2493 should be denied by ENOSPC. But if the seconde arrived before
2494 the first one is commited. The OST might wrongly permit second
2495 writing, which will cause grant space > avaiable space.
2498 Frequency : when client is evicted
2500 Description: Grant might be wrongly erased when osc is evicted by OST
2501 Details : when the import is evicted by server, it will fork another
2502 thread ptlrpc_invalidate_import_thread to invalidate the
2503 import, where the grant will be set to 0. While the original
2504 thread will update the grant it got when connecting. So if
2505 the former happened latter, the grant will be wrongly errased
2506 because of this race.
2511 Description: Checking Stale with correct fid
2512 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2513 to check whether it is stale, because sometimes, we want the
2514 enqueue happened anyway, and op_data.fid2 will not be initialized.
2516 Severity : enhancement
2518 Description: update patchless client
2519 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2522 Frequency : only with 2.4 kernel
2524 Description: random memory corruption
2525 Details : size of struct ll_inode_info is to big for union inode.u and this
2526 can be cause of random memory corruption.
2531 Description: Memory leak in recovery
2532 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2533 It should also check obd_fail before fsfilt_start, otherwise if
2534 fsfilt_start return -EROFS,(failover mds during mds recovery).
2535 then the req will return with repmsg->transno = 0 and rc = EROFS.
2536 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2537 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2538 be freed no matter whether fsfilt_commit success or not.
2541 Frequency : only with huge count clients
2543 Description: Prevents from taking the superblock lock in llap_from_page for
2545 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2546 for llap_from_page call in ll_removepage prevents from taking the
2547 superblock lock for a soon died page.
2552 Description: Not check open intent error before release open handle
2553 Details : in some rare cases, the open intent error is not checked before
2554 release open handle, which may cause
2555 ASSERTION(open_req->rq_transno != 0), because it tries to release
2556 the failed open handle.
2561 Description: Set cat log bitmap only after create log success.
2562 Details : in some rare cases, the cat log bitmap is set too early. and it
2563 should be set only after create log success.
2567 Description: Accessing a block bevice can re-enable I/O when Lustre is
2568 tearing down a device.
2569 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2574 Description: service threads may hog cpus when there are a lot of requests
2575 Details : Insert cond_resched to give other threads a chance to use some CPU
2580 Description: the cat log was not initialized in recovery
2581 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2582 unlink log on mds will not be initialized until mds post
2583 recovery. And also in mds post recovery, the unlink log will
2584 initialization will be done asynchronausly, so there will be race
2585 between add unlink log and unlink log initialization.
2589 Description: brw_stats were being printed incorrectly
2590 Details : brw_stats were being printed as log2 but all of them were not
2591 recorded as log2. Also remove some code duplication arising from
2592 filter_tally_{read,write}.
2596 Frequency : rare, only in recovery.
2597 Description: ASSERTION(req->rq_type != LI_POISON) failed
2598 Details : imp_lock should be held while iterating over imp_sending_list for
2599 prevent destroy request after get timeout in ptlrpc_queue_wait.
2603 Description: replay-single.sh test 52 fails
2604 Details : A lock's skiplist need to be cleanup when it being unlinked
2605 from its resource list.
2609 Description: Short directio read returns full requested size rather than
2611 Details : Direct I/O operations should return actual amount of bytes
2612 transferred rather than requested size.
2614 Severity : enhancement
2616 Description: metadata RPC reduction (e.g. for rm performance)
2617 Details : decrease the amount of synchronous RPC between clients and servers
2618 by canceling conflicing lock before the operation on the client
2619 and packing thier handles into the main operation RPC to server.
2621 Severity : enhancement
2623 Description: Async OSC create to avoid the blocking unnecessarily.
2624 Details : If a OST has no remain object, system will block on the creating
2625 when need to create a new object on this OST. Now, ways use
2626 pre-created objects when available, instead of blocking on an
2627 empty osc while others are not empty. If we must block, we block
2628 for the shortest possible period of time.
2632 Description: improve handling recoverable errors
2633 Details : if request processig with error which can be recoverable on server
2634 request should be resend, otherwise page released from cache and
2637 Severity : enhancement
2639 Description: refine locking for avoid write wrong info into lov_objid file
2640 Details : fix possible races with add new target and write/update data in
2643 --------------------------------------------------------------------------------
2645 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2650 Frequency : on some architectures
2652 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2653 Details : Uninitialized flags sometimes cause configuration commands to
2657 Frequency : patchless clients only
2659 Description: missing __iget() symbol export
2660 Details : The __iget() symbol export is missing. To avoid the need for
2661 this on patchless clients the deathrow inode reaper is turned
2662 off, and we depend on the VM to clean up old inodes. This
2663 dependency was during via the fix for bug 12181.
2665 --------------------------------------------------------------------------------
2667 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2669 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2670 INTEROPERATE with older versions automatically. In many cases a
2671 special upgrade step is needed. Please read the
2672 user documentation before upgrading any part of a live system.
2673 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2674 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2675 * WARNING: Lustre configuration and startup changes are required with
2676 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2680 Severity : enhancement
2682 Description: MountConf
2683 Details : Lustre configuration is now managed via mkfs and mount
2684 commands instead of lmc and lconf. New obd types (MGS, MGC)
2685 are added for dynamic configuration management. See
2686 https://mail.clusterfs.com/wikis/lustre/MountConf for
2689 Severity : enhancement
2691 Description: dynamic OST addition
2692 Details : OSTs can now be added to a live filesystem
2694 Severity : enhancement
2696 Description: startup order invariance
2697 Details : MDTs and OSTs can be started in any order. Clients only
2698 require the MDT to complete startup.
2700 Severity : enhancement
2702 Description: parallel, asynchronous orphan cleanup
2703 Details : orphan cleanup is now performed in separate threads for each
2704 OST, allowing parallel non-blocking operation.
2706 Severity : enhancement
2708 Description: optimized stripe assignment
2709 Details : stripe assignments are now made based on ost space available,
2710 ost previous usage, and OSS previous usage, in order to try
2711 to optimize storage space and networking resources.
2713 Severity : enhancement
2715 Description: Permanently set tunables
2716 Details : All writable /proc/fs/lustre tunables can now be permanently
2717 set on a per-server basis, at mkfs time or on a live system.
2719 Severity : enhancement
2721 Description: Lustre message v2
2722 Details : Add lustre message format v2.
2724 Severity : enhancement
2726 Description: client OST exclusion list
2727 Details : Clients can be started with a list of OSTs that should be
2728 declared "inactive" for known non-responsive OSTs.
2732 Description: SPEC SFS validation failure on NFS v2 over lustre.
2733 Details : Changes the blocksize for regular files to be 2x RPC size,
2734 and not depend on stripe size.
2736 Severity : enhancement
2738 Description: Multiple MD RPCs in flight.
2739 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2740 To avoid overly-overloading MDS, introduce a limit on number of
2741 MDS RPCs in flight for a single client and add /proc controls
2742 to adjust this limit.
2744 Severity : enhancement
2746 Description: client read/write statistics
2747 Details : Add client read/write call usage stats for performance
2748 analysis of user processes.
2749 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2750 file access. extents_stats shows chunk size distribution.
2751 extents_stats_per_process show chunk size distribution per
2754 Severity : enhancement
2756 Description: mds statistics
2757 Details : Add detailed mds operations statistics in
2758 /proc/fs/lustre/mds/*/stats.
2762 Description: Failure of copying files with lustre special EAs.
2763 Details : Client side always return success for setxattr call for lustre
2764 special xattr (currently only "trusted.lov").
2769 Description: Refcount LNET uuids
2770 Details : The global LNET uuid list grew linearly with every startup;
2771 refcount repeated list entries instead of always adding to
2774 Severity : enhancement
2776 Description: Dynamic service threads
2777 Details : Within a small range, start extra service threads
2778 automatically when the request queue builds up.
2781 Frequency : mixed-endian client/server environments
2783 Description: mixed-endian crashes
2784 Details : The new msg_v2 system had some failures in mixed-endian
2787 Severity : enhancement
2789 Description: Easy OST removal
2790 Details : OSTs can be permanently deactivated with e.g. 'lctl
2791 conf_param lustre-OST0001.osc.active=0'
2793 Severity : enhancement
2795 Description: MGS proc entries
2796 Details : Added basic proc entries for the MGS showing what filesystems
2799 Severity : enhancement
2801 Description: provide MGS failover
2802 Details : Added config lock reacquisition after MGS server failover.
2804 Severity : enhancement
2806 Description: add Linux 2.4 support
2807 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2811 Description: a large application tries to do I/O to the same resource and dies
2812 in the middle of it.
2813 Details : Check the req->rq_arrival time after the call to
2814 ost_brw_lock_get(), but before we do anything about
2815 processing it & sending the BULK transfer request. This
2816 should help move old stale pending locks off the queue as
2817 quickly as obd_timeout.
2820 Frequency : when an incorrect nid is specified during startup
2822 Description: ptlrpc connect to non-existant node causes kernel crash
2823 Details : LNET can't be re-entered from an event callback, which
2824 happened when we expire a message after the export has been
2825 cleaned up. Instead, hand the zombie cleanup off to another
2828 Severity : enhancement
2830 Description: plain/inodebits lock performance improvement
2831 Details : Grouping plain/inodebits in granted list by their request modes
2832 and bits policy, thus improving the performance of search through
2836 Frequency : only if OST filesystem is corrupted
2838 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2839 Details : for a short time RPCs with bulk IO are in the replay list,
2840 but replay of bulk IOs is unimplemented. If the OST filesystem
2841 is corrupted due to disk cache incoherency and then replay is
2842 started it is possible to trip an assertion. Avoid putting
2843 committed RPCs into the replay list at all to avoid this issue.
2846 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2848 Description: System hang on startup
2849 Details : This bug allowed the liblustre (e.g. catamount) client to
2850 return to the app before handling all startup RPCs. This
2851 could leave the node unresponsive to lustre network traffic
2852 and manifested as a server ptllnd timeout.
2854 Severity : enhancement
2856 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2857 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2858 boolean to control whether to print peer debug info when a
2859 client's RPC times out.
2862 Frequency : only for kernels with patches from Lustre below 1.4.3
2864 Description: Remove old rdonly API
2865 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2868 Frequency : only for devices with external journals
2870 Description: Set external device read-only also
2871 Details : During a commanded failover stop, we set the disk device
2872 read-only while the server shuts down. We now also set any
2873 external journal device read-only at the same time.
2876 Frequency : when upgrading from 1.4 while trying to change parameters
2878 Description: The wrong (new) MDC name was used when setting parameters for
2879 upgraded MDT's. Also allows changing of OSC (and MDC)
2880 parameters if --writeconf is specified at tunefs upgrade time.
2883 Frequency : when setting specific ost indicies
2885 Description: QOS code breaks on skipped indicies
2886 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2887 created with --index need not be sequential.
2889 Severity : enhancement
2891 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2892 Details : The uninit_groups feature works in conjunction with the kernel
2893 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2894 up the pass1 processing of e2fsck. This is a read-only feature
2895 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2896 mount filesystems that have had this feature enabled.
2898 Severity : enhancement
2900 Description: Improve multi-block allocation algorithm to avoid fragmentation
2901 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2902 allocation locality and avoid filesystem fragmentation.
2904 ------------------------------------------------------------------------------
2906 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2908 * Support for kernels:
2909 2.4.21-47.0.1.EL (RHEL 3)
2910 2.6.5-7.283 (SLES 9)
2911 2.6.9-42.0.10.EL (RHEL 4)
2912 2.6.12.6 vanilla (kernel.org)
2913 2.6.16.27-0.9 (SLES 10)
2914 * Recommended e2fsprogs version: 1.39.cfs5
2916 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2920 Frequency : occasional, depends on client load and configuration
2921 Bugzilla : 12181, 12203
2922 Description: data loss for recently-modified files
2924 Details : In some cases it is possible that recently written or created
2925 files may not be written to disk in a timely manner (this should
2926 normally be within 30s unless client IO load is very high).
2927 The problem appears as zero-length files or files that are a
2928 multiple of 1MB in size after a client crash or client eviction
2929 that are missing data at the end of the file.
2931 This problem is more likely to be hit on clients where files are
2932 repeatedly created and unlinked in the same directory, clients
2933 have a large amount of RAM, have many CPUs, the filesystem has
2934 many OSTs, the clients are rebooted frequently, and/or the files
2935 are not accessed by other nodes after being written.
2937 The presence of the problem can be detected by looking at
2938 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2939 smaller than the second (nr_unused) then dirty files will not
2940 be flushed automatically to disk. "sync; sleep 10" should be
2941 run several times on the node before unmounting it to update
2942 Lustre (this is also safe to run on nodes without this problem).
2944 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2945 that can cause this same problem, so customers using that kernel
2946 also need to update the kernel in addition to Lustre. In order
2947 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2949 It is normal that files written just before a client crash (less
2950 than 30s) may not yet have been flushed to disk, even for local
2954 Frequency : frequent on thin XT3 nodes
2956 Description: UUID collision on thin XT3 Linux nodes
2957 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2958 randomly, since we relied on an insufficiently-seeded PRNG.
2963 Description: OSS hangs after "All ost request buffers busy"
2964 Details : A deadlock between quota and journal operations caused OSS
2965 hangs after printing "All ost request buffers busy."
2968 Frequency : always on liblustre builds
2970 Description: Cleanup compiler warnings on liblustre
2973 Frequency : always on liblustre builds on XT3
2975 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2979 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2980 > 32000 subdirectories
2981 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2982 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2983 subdirectory count crosses 32000. This will aid e2fsck to
2984 correctly handle more than 32000 subdirectories.
2987 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2989 Description: System hang on startup
2990 Details : This bug allowed the liblustre (e.g. catamount) client to
2991 return to the app before handling all startup RPCs. This
2992 could leave the node unresponsive to lustre network traffic
2993 and manifested as a server ptllnd timeout.
2995 Severity : enhancement
2997 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2998 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2999 boolean to control whether to print peer debug info when a
3000 client's RPC times out.
3005 Description: make O_SYNC working on 2.6 kernels
3006 Details : 2.6 kernels use different method for mark pages for write,
3007 so need add a code to lustre for O_SYNC work.
3012 Description: Failure to close file and release space on NFS
3013 Details : Put inode details into lock acquired in ll_intent_file_open.
3014 Use mdc_intent_lock in ll_intent_open to properly
3015 detect all kind of errors unhandled by mdc_enqueue.
3020 Description: proc file read during shutdown sometimes raced obd removal,
3022 Details : Add lock to prevent obd access after proc file removal.
3025 Frequency : Only for files larger than 4GB on 32-bit clients.
3027 Description: improperly doing page alignment of locks
3028 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3029 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3034 Description: RPCs being resent when they shouldn't be
3035 Details : Some RPCs that should not be resent are being resent. This
3036 can cause inconsistencies in the RPC state machine. Do not
3037 resend such requests.
3040 Frequency : rare, only with NFS export
3042 Description: Crash on NFS re-export node
3043 Details : under very unusual load conditions an assertion is hit in
3044 ll_intent_file_open()
3047 Frequency : only if OST filesystem is corrupted
3049 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3050 Details : for a short time RPCs with bulk IO are in the replay list,
3051 but replay of bulk IOs is unimplemented. If the OST filesystem
3052 is corrupted due to disk cache incoherency and then replay is
3053 started it is possible to trip an assertion. Avoid putting
3054 committed RPCs into the replay list at all to avoid this issue.
3059 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3060 Details : Large single O_DIRECT read and write calls can fail to allocate
3061 a sufficiently large buffer to process the request. In case of
3062 allocation failure the allocation is retried with a smaller
3063 buffer and broken into smaller requests.
3065 Severity : enhancement
3067 Description: Add -o localflock option to simulate old noflock behaviour.
3068 Details : This will achieve local-only flock/fcntl locks coherentness.
3073 Description: versioning check is incomplete
3074 Details : Checking the version difference of client vs. server, report
3075 error if the gap is too big.
3080 Description: add support PG_writeback bit
3081 Details : add support for PG_writeback bit for Lustre, for more carefull
3082 work with page cache in 2.6 kernel. This also fix some deadlocks
3083 and remove hack for work O_SYNC with 2.6 kernel.
3085 Severity : enhancement
3087 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3088 Details : The uninit_groups feature works in conjunction with the kernel
3089 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3090 up the pass1 processing of e2fsck. This is a read-only feature
3091 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3092 mount filesystems that have had this feature enabled.
3094 Severity : enhancement
3096 Description: Improve multi-block allocation algorithm to avoid fragmentation
3097 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3098 allocation locality and avoid filesystem fragmentation.
3100 ------------------------------------------------------------------------------
3102 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3104 * Support for kernels:
3105 2.6.9-42.0.3.EL (RHEL 4)
3106 2.6.5-7.276 (SLES 9)
3107 2.4.21-47.0.1.EL (RHEL 3)
3108 2.6.12.6 vanilla (kernel.org)
3109 2.6.16.21-0.8 (SLES10)
3110 * Recommended e2fsprogs version: 1.39.cfs2-0
3112 * The backwards-compatible /proc/sys/portals symlink has been removed
3113 in this release. Before upgrading, please ensure that you change
3114 any configuration scripts or /etc/sysctl.conf files that access
3115 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3116 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3117 in advance of the upgrade on any system running Lustre 1.4.6 or
3118 newer, since /proc/sys/lnet was added in that version.
3119 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3123 Frequency : only when quota is used
3125 Description: avoid scanning export list for quota master
3126 Details : Change the algorithms to avoid scanning export list in order
3127 to improve the efficiency.
3130 Frequency : MDS failover only, very rarely
3132 Description: "went back in time" messages on mds failover
3133 Details : The greatest transno may be lost when the current operation
3134 finishes with an error (transno==0) and the client's last_rcvd
3135 record is over-written. Save the greatest transno in the
3136 mds_last_transno for this case.
3139 Frequency : always for specific kernels and striping counts
3141 Description: client may get "Matching packet too big" without ACL support
3142 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3143 when trying to access files in certain configurations. The
3144 clients should in fact be denied when mounting because they do
3145 not understand ACLs.
3148 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3150 Description: many clients connecting with IO in progress causes connect timeouts
3151 Details : Avoid synchronous journal commits to avoid delays caused by many
3152 clients connecting/disconnecting when bulk IO is in progress.
3153 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3154 OST_IO_PORTAL to avoid delays behind potentially many pending
3158 Frequency : occasionally with multiple writers to a single file
3160 Description: shared writes to file may result in wrong size reported by stat()
3161 Details : Allow growing of kms when extent lock is cancelled
3164 Frequency : always with random mmap IO to multi-striped file
3166 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3167 Details : Only if the hole is at the end of OST object so that kms is too
3168 small. Fix is to increase kms accordingly in ll_nopage.
3171 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3173 Description: writes to a missing object would leak memory on the OST
3174 Details : If there is an inconsistency between the MDS and OST filesystems,
3175 such that the MDS references an object that doesn't exist, writes
3176 to that object will leak memory due to incorrect cleanup in the
3177 error handling path, eventually running out of memory on the OST.
3182 Description: Creating too long symlink causes lustre errors
3183 Details : Check symlink and name lengths before sending requests to MDS.
3186 Frequency : only if flock is enabled (not on by default)
3188 Description: posix locks not released on fd closure on 2.6.9+
3189 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3190 kernels, this caused such locks not to be released on fd close and
3191 then assertions on fs unmount about still used locks.
3194 Frequency : MDS failover only, very rarely
3196 Description: clients may get ASSERTION(granted_lock != NULL)
3197 Details : When request was taking a long time, and a client was resending
3198 a getattr by name lock request. The were multiple lock requests
3199 with the same client lock handle and
3200 mds_getattr_name->fixup_handle_for_resent_request found one of the
3201 lock handles but later failed with ASSERTION(granted_lock != NULL).
3206 Description: handle->h_buffer_credits > 0, assertion failure
3207 Details : h_buffer_credits is zero after truncate, causing assertion
3208 failure. This patch extends the transaction or creates a new
3212 Frequency : NFS re-export or patchless client
3213 Bugzilla : 11179, 10796
3214 Description: Crash on NFS re-export node (__d_move)
3215 Details : We do not want to hash the dentry if we don't have a lock.
3216 But if this dentry is later used in d_move, we'd hit uninitialised
3217 list head d_hash, so we just do this to init d_hash field but
3218 leave dentry unhashed.
3221 Frequency : NFS re-export or patchless client
3223 Description: NFS exports has problem with symbolic link
3224 Details : lustre client didn't properly install dentry when re-exported
3225 to NFS or running patchless client.
3228 Frequency : NFS re-export or patchless client
3230 Description: Various nfs/patchless fixes.
3231 Details : fixes reuse disconected alias for lookup process - this fixes
3232 warning "find_exported_dentry: npd != pd",
3233 fix permission error with open files at nfs.
3234 fix apply umask when do revalidate.
3237 Frequency : occasional
3239 Description: Crash on NFS re-export node
3240 Details : calling clear_page() on the wrong pointer triggered oops in
3241 generic_mapping_read().
3244 Frequency : rarely, using O_DIRECT IO
3246 Description: unaligned directio crashes client with LASSERT
3247 Details : check for unaligned buffers before trying any requests.
3250 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3252 Description: stale data returned from RAID cache
3253 Details : If only a small amount of IO is done to the RAID device before
3254 reading it again it is possible to get stale data from the RAID
3255 cache instead of reading it from disk.
3258 Frequency : always for sles10 kernel
3260 Description: sles10 support
3261 Details : ll_follow_link: compile fixes and using of nd_set_link
3262 under newer kernels.
3265 Frequency : depends on arch, kernel and compiler version, always on sles10
3268 Description: recursive or deep enough symlinks cause stack overflow
3269 Details : getting rid of large stack-allocated variable in
3273 Frequency : depends on hardware
3275 Description: lustre write performance loss in the SLES10 kernel
3276 Details : the performance loss is caused by using of write barriers in the
3277 ext3 code. The SLES10 kernel turns barrier support on by
3278 default. The fix is to undo that change for ldiskfs.
3280 ------------------------------------------------------------------------------
3282 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3284 * Support for kernels:
3285 2.6.9-42.0.3EL (RHEL 4)
3286 2.6.5-7.276 (SLES 9)
3287 2.4.21-47.0.1.EL (RHEL 3)
3288 2.6.12.6 vanilla (kernel.org)
3292 Frequency : quota enabled and large files being deleted
3294 Description: releasing more than 4GB of quota at once hangs OST
3295 Details : If a user deletes more than 4GB of files on a single OST it
3296 will cause the OST to spin in an infinite loop. Release
3297 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3302 Description: statfs data retrieved from /proc may be stale or zero
3303 Details : When reading per-device statfs data from /proc, in the
3304 {kbytes,files}_{total,free,avail} files, it may appear
3305 as zero or be out of date.
3308 Frequency : systems with MD RAID1 external journal devices
3310 Description: lconf's call to blkid is confused by RAID1 journal devices
3311 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3312 of returning all block devices that match the journal UUID.
3315 Frequency : always, for aggregate stripe size over 4GB
3317 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3318 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3319 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3322 Frequency : always if "lfs find" used on a local file/directory
3324 Description: "lfs find" segfaults if used on a local file/directory
3325 Details : The case where a directory component was not specified wasn't
3326 handled correctly. Handle this properly.
3329 Frequency : always on ppc64
3331 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3332 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3333 perform endianness conversion.
3336 Frequency : rarely (truncate to non-zero file size after write under load)
3337 Bugzilla : 10730, 10687
3338 Description: Files padded with zeros to next 4K multiple
3339 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3340 it is possible that files that are truncated to a non-zero size
3341 immediately after being written are filled with zero bytes beyond
3342 the truncated size. No file data is lost.
3344 Severity : enhancement
3346 Description: Allow recovery/failover for liblustre clients.
3347 Details : liblustre clients were unaware of failover configurations until
3350 Severity : enhancement
3352 Description: user file locks should fail when not mounting with flock option
3353 Details : Set up an error-returning stub in ll_file_operations.lock field
3354 to prevent incorrect behaviour when client is mounted without
3355 flock option. Also, set up properly f_op->flock field for
3359 Frequency : always on ia64
3361 Description: "lfs df" loops on printing out MDS statfs information
3362 Details : The obd_ioctl_data was not initialized and in some systems
3363 this caused a failure during the ioctl that did not return
3364 an error. Initialize the struct and return an error on failure.
3367 Frequency : SLES 9 only
3369 Description: Error of copying files with lustre special EAs as root
3370 Details : Client side always return success for setxattr call for lustre
3371 special xattr (currently only "trusted.lov").
3374 Frequency : rarely on clusters with both ia64+i386 clients
3376 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3377 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3378 clients are doing concurrent writes to the same file it is
3379 possible that the ia64 clients may LASSERT because the OST
3380 extent locks are not PAGE_SIZE aligned. Ensure that grown
3381 locks are always aligned on the request boundary.
3384 Frequency : specific use, occasional
3386 Description: Overwriting in use executable truncates on-disk binary image
3387 Details : If one node attempts to overwrite an executable in use by
3388 another node, we now correctly return ETXTBSY instead of
3389 truncating the file.
3391 Severity : enhancement
3393 Description: Async OSC create to avoid the blocking unnecessarily.
3394 Details : If a OST has no remain object, system will block on the creating
3395 when need to create a new object on this OST. Now, ways use
3396 pre-created objects when available, instead of blocking on an
3397 empty osc while others are not empty. If we must block, we block
3398 for the shortest possible period of time.
3403 Description: chmod on Lustre root is propagated to other clients
3404 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3405 invalid by the follow_mount time.
3408 Frequency : liblustre clients only
3410 Description: Race in 'instant cancel' lock handling could lead to such locks
3411 never to be granted in case of SMP MDS
3412 Details : Do not destroy not yet granted but cbpending locks in
3416 Frequency : replay/resend of open
3418 Description: non null lock assetion failure in mds_intent_policy
3419 Details : Trying to replay/resend lockless open requests resulted in
3420 mds_open() returning 0 with no lock. Now it sets a flag if
3421 a lock is going to be returned.
3423 Severity : enhancement
3425 Description: Checksum enhancements
3426 Details : New checksum enhancements allow for resending RPCs that failed
3429 Severity : enhancement
3431 Description: Tunables on number of dirty pages in cacche
3432 Details : Allow to set limit on number of dirty pages cached.
3437 Description: client crash on unmount - lock still has references
3438 Details : In some error handling cases it was possible to leak a lock
3439 reference on a client while accessing a file. This was not
3440 harmful to the client during operation, but would cause the
3441 client to crash when the filesystem is unmounted.
3444 Frequency : specific case, rare
3446 Description: ETXTBSY on mds though file not in use by client
3447 Details : ETXTBSY is no longer incorrectly returned when attempting to
3448 chmod or chown a directory that the user previously tried to
3449 execute or a currently-executing binary.
3452 Frequency : extremely rare except on liblustre-based clients
3454 Description: Lustre space not freed when files are deleted
3455 Details : Clean up open-unlinked files after client eviction. Previously
3456 the unlink was skipped and the files remained as orphans.
3461 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3462 Details : In some cases it was possible to send a blocking callback to
3463 a client doing a glimpse, even though that client didn't get
3464 a lock granted. When the glimpse lock is cancelled on the OST
3465 the freed lock is left on the waiting list and corrupted the list.
3468 Frequency : all core dumps
3470 Description: Broke core dumps to lustre
3471 Details : Negative dentry may be unhashed if parent does not have UPDATE
3472 lock, but some callers, e.g. do_coredump, expect dentry to be
3473 hashed after successful create, hash it in ll_create_it.
3475 ------------------------------------------------------------------------------
3477 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3479 * Support for kernels:
3480 2.6.9-42.0.2.EL (RHEL 4)
3481 2.6.5-7.276 (SLES 9)
3482 2.4.21-40.EL (RHEL 3)
3483 2.6.12.6 vanilla (kernel.org)
3487 Frequency : always on RHEL 3
3489 Description: Number of open files grows over time
3490 Details : The number of open files grows over time, whether or not
3491 Lustre is started. This was due to a filp leak introduced
3492 by one of our kernel patches.
3494 ------------------------------------------------------------------------------
3496 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3498 * Support for kernels:
3499 2.6.9-42.EL (RHEL 4)
3500 2.6.5-7.276 (SLES 9)
3501 2.4.21-40.EL (RHEL 3)
3502 2.6.12.6 vanilla (kernel.org)
3507 Bugzilla : 5719, 9635, 9792, 9684
3508 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3509 Details : If a server is under heavy load and cannot reply to new
3510 connection requests before the client resends the (re)connect,
3511 the connection handling code can behave badly if two service
3512 threads are concurrently handing separate (re)connections from
3513 the same client. Add better locking to the connection handling
3514 code, and ensure that only a single connection will be processed
3515 for a given client UUID, even if the lock is dropped.
3517 Severity : enhancement
3519 Description: add TCP zero-copy support to kernel
3520 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3521 sends if the hardware supports scatter-gather and checksumming.
3522 This allows socklnd to do client-write and server-read more
3523 efficiently and reduce CPU utilization from skbuf copying.
3526 Frequency : only if NFS exporting from client
3528 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3529 Details : If setting attributes on a file created under NFS that had
3530 never been opened it would be possible to oops the client
3531 if the file had no objects.
3534 Frequency : always for liblustre
3536 Description: liblustre client does MDS+OSTs setattr RPC for each write
3537 Details : When doing a write from a liblustre client, the client
3538 incorrectly issued an RPC to the MDS and each OST the file was
3539 striped over in order to update the timestamps. When writing
3540 with small chunks and many clients this could overwhelm the MDS
3541 with RPCs. In all cases it would slow down the write because
3542 these RPCs are unnecessary.
3544 Severity : enhancement
3546 Description: allow number of MDS service threads to be changed at module load
3547 Details : It is now possible to change the number of MDS service threads
3548 running. Adding "options mds mds_num_threads={N}" to the MDS's
3549 /etc/modprobe.conf will set the number of threads for the next
3550 time Lustre is restarted (assuming the "mds" module is also
3551 reloaded at that time). The default number of threads will
3552 stay the same, 32 for most systems.
3557 Description: OST crash if filesystem is unformatted or corrupt
3558 Details : If an OST is started on a device that has never been formatted
3559 or if the filesystem is corrupt and cannot even mount then the
3560 error handling cleanup routines would dereference a NULL pointer.
3565 Description: NULL pointer deref in llap_from_page.
3566 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3567 incorrect) mapping if the page was truncated/reclaimed while it was
3568 searched for. Check for this condition and skip such pages when
3569 doing readahead. Introduce extra check to llap_from_page() to
3570 verify page->mapping->host is non-NULL (so page is not anonymous).
3573 Frequency : Sometimes when using sys_sendfile
3575 Description: "page not covered by a lock" warnings from ll_readpage
3576 Details : sendfile called ll_readpage without right page locks present.
3577 Now we introduced ll_file_sendfile that does necessary locking
3578 around call to generic_file_sendfile() much like we do in
3582 Frequency : with certain MDS communication failures at client mount time
3584 Description: NULL pointer deref after failed client mount
3585 Details : a client connection request may delayed by the network layer
3586 and not be sent until after the PTLRPC layer has timed out the
3587 request. If the client fails the mount immediately it will try
3588 to clean up before the network times out the request. Add a
3589 reference from the request import to the obd device and delay
3590 the cleanup until the network drops the request.
3593 Frequency : occasionally during client (re)connect
3595 Description: assertion failure during client (re)connect
3596 Details : processing a client connection request may be delayed by the
3597 client or server longer than the client connect timeout. This
3598 causes the client to resend the connection request. If the
3599 original connection request is replied in this interval, the
3600 client may trip an assertion failure in ptlrpc_connect_interpret()
3601 which thought it would be the only running connect process.
3604 Frequency : only with obd_echo servers and clients that are rebooted
3606 Description: kernel BUG accessing uninitialized data structure
3607 Details : When running an obd_echo server it did not start the ping_evictor
3608 thread, and when a client was evicted an uninitialized data
3609 structure was accessed. Start the ping_evictor in the RPC
3610 service startup instead of the OBD startup.
3612 Severity : enhancement
3613 Bugzilla : 10193 (patchless)
3614 Description: Remove dependency on various unexported kernel interfaces.
3615 Details : No longer need reparent_to_init, exit_mm, exit_files,
3616 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3619 Frequency : rare (only users of deprecated and unsupported LDAP config)
3621 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3622 Details : LDAP apparently contains 'lustreName' attributes instead of
3623 'name'. A simple remapping of the name is sufficient.
3626 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3628 Description: waiting_locks_callback trips kernel BUG if client is evicted
3629 Details : Running with the dump_on_timeout debug flag turned on makes
3630 it possible that the waiting_locks_callback() can try to dump
3631 the Lustre kernel debug logs from an interrupt handler. Defer
3632 this log dumping to the expired_lock_main() thread.
3634 Severity : enhancement
3636 Description: Support NFS exporting on 2.6 kernels.
3637 Details : Implement non-rawops metadata methods for NFS server to use without
3638 changing NFS server code.
3641 Frequency : very rare (synthetic metadata workload only)
3643 Description: two racing renames might cause an MDS thread to deadlock
3644 Details : Running the "racer" program may cause one MDS thread to rename
3645 a file from being the source of a rename to being the target of
3646 a rename at exactly the same time that another thread is doing
3647 so, and the second thread has already enqueued these locks after
3648 doing a lookup of the target and is trying to relock them in
3649 order. Ensure that we don't try to re-lock the same resource.
3652 Frequency : only very large systems with liblustre clients
3654 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3655 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3656 all OSTs in parallel. This allows the eviction of stale liblustre
3657 clients to proceed much faster than if they were done in series,
3658 and also offers similar improvements for other set_info RPCs.
3663 Description: excessive CPU usage during initial read phase on client
3664 Details : During the initial read phase on a client, it would agressively
3665 retry readahead on the file, consuming too much CPU and impacting
3666 performance (since 1.4.5.8). Improve the readahead algorithm
3667 to avoid this, and also improve some other common cases (read
3668 of small files in particular, where "small" is files smaller than
3669 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3674 Description: MDS crash when receiving packet with unknown intent.
3675 Details : Do not LBUG in unknown intent case, just return -EFAULT
3677 Severity : enhancement
3678 Bugzilla : 9293, 9385
3679 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3680 Details : Do not serialize getattr (non-intent version) and statfs.
3683 Frequency : occasional, when OST network is overloaded/intermittent
3685 Description: client evicted by OST after bulk IO timeout
3686 Details : If a client sends a bulk IO request (read or write) the OST
3687 may evict the client if it is unresposive to its data GET/PUT
3688 request. This is incorrect if the network is overloaded (takes
3689 too long to transfer the RPC data) or dropped the OST GET/PUT
3690 request. There is no need to evict the client at all, since
3691 the pinger and/or lock callbacks will handle this, and the
3692 client can restart the bulk request.
3695 Frequency : Always when mmapping file with no objects
3697 Description: client crashes when mmapping file with no objects
3698 Details : Check that we actually have objects in a file before doing any
3699 operations on objects in ll_vm_open, ll_vm_close and
3705 Description: Request leak when working with deleted CWD
3706 Details : Introduce advanced request refcount tracking for requests
3707 referenced from lustre intent.
3709 Severity : Enhancement
3711 Description: Cache open file handles on client.
3712 Details : MDS now will return special lock along with openhandle, if
3713 requested and client is allowed to hold openhandle, even if unused,
3714 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3715 closing files for every read/write openration.
3717 Severity : Enhancement
3719 Description: Cache open negative dentries on client when possible.
3720 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3721 negative dentries on lock revocation.
3726 Description: Remounting a client read-only wasn't possible with a zconf mount
3727 Details : It wasn't possible to remount a client read-only with llmount.
3729 Severity : enhancement
3730 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3731 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3732 MPICH is included to add Lustre-specific ADIO interfaces.
3733 This is based closely on the UFS ADIO layer and only differs
3734 in file creation, in order to allow the OST striping to be set.
3735 This is user-contributed code and not supported by CFS.
3740 Description: extended inode attributes (immutable, append-only) work improperly
3741 when 2.4 and 2.6 kernels are used on client/server or vice versa
3742 Details : Introduce kernel-independent values for these flags.
3744 Severity : enhancement
3747 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3748 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3749 granularity. Now, Fractional MB granularity is be supported,
3750 this is very useful for low memory system.
3752 Severity : enhancement
3754 Description: Getattr by fid
3755 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3756 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3757 have locks and avoids getattr rpc requests for it on every stat.
3760 Frequency : Always, for filesystems larger than 2TB
3762 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3763 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3764 filesystems larger than 2TB to be mounted with mballoc enabled.
3767 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3769 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3770 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3771 filesystems larger than 2TB is mounted then the kernel will
3772 silently corrupt the start of the filesystem. CONFIG_LBD is
3773 enabled for all CFS-supported kernels, but the possibility of
3774 this happening with a modified kernel config exists.
3776 Severity : enhancement
3778 Description: add client O_DIRECT support for 2.6 kernels
3779 Details : It is now possible to do O_DIRECT reads and writes to files
3780 in the Lustre client mountpoint on 2.6 kernel clients.
3782 Severity : enhancement
3784 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3785 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3786 parallel, not waiting for response from every OST before sending
3787 a rpc to the next OST.
3792 Description: setattr vs write race when updating file timestamps
3793 Details : Client processes that update a file timestamp into the past
3794 right after writing to the file (e.g. tar) it is possible that
3795 the updated file modification time can be reset to the current
3796 time due to a race between processing the setattr and write RPC.
3798 Severity : enhancement
3800 Description: Bring 'lfs find' closer in line with regular Linux find.
3801 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3802 -print0 options and obtains all the needed info through the lustre
3805 Severity : enhancement
3807 Description: support up to 1024 configured devices on one node
3808 Details : change obd_dev array from statically allocated to dynamically
3809 allocated structs as they are first used to reduce memory usage
3814 Description: Flush dirty partially truncated pages during truncate
3815 Details : Immediatelly flush partially truncated pages in filter_setattr,
3816 this way we completely avoid having any pages in page cache on OST
3817 and can retire ugly workarounds during writes to flush such pages.
3822 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3823 Details : protect inode from truncation within vfs_unlink() context
3824 just take a reference before calling vfs_unlink() and release it
3825 when parent's i_sem is free.
3830 Description: last_id value checked outside lock on OST caused LASSERT failure
3831 Details : If there were multiple MDS->OST object precreate requests in
3832 flight, it was possible that the OST's last object id was checked
3833 outside a lock and incorrectly tripped an assertion. Move checks
3834 inside locks, and discard old precreate requests.
3837 Frequency : always, if extents are used on OSTs
3839 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3840 Details : OSTs using the extents format would not zero the high 16 bits of
3841 the index physical block number. This is not a problem for any
3842 OST filesystems smaller than 16TB, and no kernels support ext3
3843 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3844 new/modified files) and can be fixed for existing filesystems
3845 with e2fsprogs-1.39-cfs1.
3850 Description: import connection selection may be incorrect if timer wraps
3851 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3852 import connections to be ignored if the 32-bit jiffies counter
3853 wraps. Use a 64-bit jiffies counter.
3856 Frequency : very large clusters immediately after boot
3858 Description: LNET request buffers exhausted under heavy short-term load
3859 Details : If a large number of client requests are generated on a service
3860 that has previously never seen so many requests it is possible
3861 that the request buffer growth cannot keep up with the spike in
3862 demand. Instead of dropping incoming requests, they are held in
3863 the LND until the RPC service can accept more requests.
3866 Frequency : Sometimes during replay
3868 Description: Assertion failure in ll_local_open after replay.
3869 Details : If replay happened on an open request reply before we were able
3870 to set replay handler, reply will become not swabbed tripping the
3871 assertion in ll_local_open. Now we set the handler right after
3872 recognising of open request
3875 Frequency : very rare
3877 Description: kernel reports "badness in vsnprintf"
3878 Details : Reading from the "recovery_status" /proc file in small chunks
3879 may cause a negative length in lprocfs_obd_rd_recovery_status()
3880 call to vsnprintf() (which is otherwise harmless). Exit early
3881 if there is no more space in the output buffer.
3883 Severity : enhancement
3885 Description: clear OBD RPC statistics by writing to them
3886 Details : It is now possible to clear the OBD RPC statistics by writing
3887 to the "stats" file.
3892 Description: Client mtime is not the same on different clients after utimes
3893 Details : In some cases, the client was using the utimes() syscall on
3894 a file cached on another node. The clients now validate the
3895 ctime from the MDS + OSTs to determine which one is right.
3900 Description: Inability to activate failout mode
3901 Details : lconf script incorrectly assumed that in pythong string's numeric
3902 value is used in comparisons.
3905 Frequency : always with multiple stripes per file
3907 Description: Inefficient object allocation for mutli-stripe files
3908 Details : When selecting which OSTs to stripe files over, for files with
3909 a stripe count that divides evenly into the number of OSTs,
3910 the MDS is always picking the same starting OST for each file.
3911 Return the OST selection heuristic to the original design.
3916 Description: mount failures may take full timeout to return an error
3917 Details : Under some heavy load conditions it is possible that a
3918 failed mount can wait for the full obd_timeout interval,
3919 possibly several minutes, before reporting an error.
3920 Instead return an error as soon as the status is known.
3922 Frequency : quota enabled and large files being deleted
3924 Description: releasing more than 4GB of quota at once hangs OST
3925 Details : If a user deletes more than 4GB of files on a single OST it
3926 will cause the OST to spin in an infinite loop. Release
3927 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3932 Description: statfs data retrieved from /proc may be stale or zero
3933 Details : When reading per-device statfs data from /proc, in the
3934 {kbytes,files}_{total,free,avail} files, it may appear
3935 as zero or be out of date.
3938 Frequency : systems with MD RAID1 external journal devices
3940 Description: lconf's call to blkid is confused by RAID1 journal devices
3941 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3942 of returning all block devices that match the journal UUID.
3945 Frequency : always, for aggregate stripe size over 4GB
3947 Description: assertion fails when trying to use 4GB stripe size
3948 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3949 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3952 Frequency : always on ppc64
3954 Description: the first write on an ext3 filesystem with mballoc got stuck
3955 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3956 perform endianness conversion.
3958 ------------------------------------------------------------------------------
3960 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3962 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3963 INTEROPERATE with older versions automatically. Please read the
3964 user documentation before upgrading any part of a live system.
3965 * WARNING: Lustre networking configuration changes are required with
3966 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3969 * Support for kernels:
3970 2.6.9-22.0.2.EL (RHEL 4)
3971 2.6.5-7.244 (SLES 9)
3972 2.6.12.6 vanilla (kernel.org)
3975 Severity : enhancement
3976 Bugzilla : 7981/8208
3977 Description: Introduced Lustre Networking (LNET)
3978 Details : LNET is new networking infrastructure for Lustre, it includes
3979 a reorganized network configuration mode (see the user
3980 documentation for full details) as well as support for routing
3981 between different network fabrics. Lustre Networking Devices
3982 (LNDs) for the supported network fabrics have also been
3983 created for this new infrastructure.
3985 Severity : enhancement
3986 Description: Introduced Access control lists
3987 Details : clients can set ACLs on files and directories in order to have
3988 more fine-grained permissions than the standard Unix UGO+RWX.
3989 The MDS must be started with the "-o acl" mount option.
3991 Severity : enhancement
3992 Description: Introduced filesystem quotas
3993 Details : Administrators may now establish per-user quotas on the
3996 Severity : enhancement
3998 Description: Configuration change for the XT3
3999 The PTLLND is now used to run Lustre over Portals on the XT3
4000 The configure option(s) --with-cray-portals are no longer used.
4001 Rather --with-portals=<path-to-portals-includes> is used to
4002 enable building on the XT3. In addition to enable XT3 specific
4003 features the option --enable-cray-xt3 must be used.
4008 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4009 Details : A race between allocating a new llcd and re-getting the llcd_lock
4010 allowed another thread to grab newly-allocated llcd.
4012 Severity : enhancement
4014 Description: 2.6 OST async journal commit and locking fix to improve performance
4015 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4016 2.6 kernels are now async as they already were in 2.4 kernels so
4017 that they can commit concurrently with the network bulk transfer.
4018 For block-allocated files the filter allocation semaphore is held
4019 to avoid filesystem fragmentation during allocation. BKL lock
4020 removed for 2.6 xattr operations where it is no longer needed.
4025 Description: lconf incorrectly determined whether two IP networks could talk
4026 Details : In some more complicated routing and multiple-network
4027 configurations, lconf will avoid trying to make a network
4028 connection to a disjoint part of the IP space. It was doing the
4029 math incorrectly for one set of cases.
4034 Description: Fix for potential infinite loop processing records in an llog.
4035 Details : If an llog record is corrupted/zeroed, it is possible to loop
4036 forever in llog_process(). Validate the llog record length
4037 and skip the remainder of the block on error.
4040 Frequency : occasional (liblustre only)
4042 Description: liblustre could not open files whose last component is a symlink
4043 Details : sysio_path_walk() would incorrectly pass the open intent to
4044 intermediate path components.
4047 Frequency : rare (liblustre only with non-standard tuning)
4048 Bugzilla : 7201 (7350)
4049 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4050 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4051 after it has been released, only do this for referenced locks
4053 Severity : enhancement
4055 Description: specifying an (invalid) directory default stripe_size of -1
4056 would reset the directory default striping
4057 Details : stripe_size -1 was used internally to signal directory stripe
4058 removal, now use "all default" to signal dir stripe removal
4059 as a directory striping of "all default" is not useful
4062 Frequency : common for large clusters running liblustre clients
4064 Description: doing an ls when liblustre clients are running is slow
4065 Details : sending a glimpse AST to a liblustre client waits for every AST
4066 to time out, as liblustre clients will not respond. Since they
4067 cannot cache data we refresh the OST lock LVB from disk instead.
4069 Severity : enhancement
4071 Description: doing an ls at the same time as file IO can be slow
4072 Details : enqueue and other "small" requests can be blocked behind many
4073 large IO requests. Create a new OST IO portal for non-IO
4074 requests so they can be processed faster.
4077 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4079 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4080 created file with mode 000
4081 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4082 allow an ftruncate() as a non-root user to a file with mode 000.
4083 Lustre now respects this flag to disable mode checks when
4084 truncating a file owned by the user
4087 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4090 Description: Revoking locks from clients that went dead or catatonic might take
4092 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4093 cancellation of such locks instant on servers without waiting for
4094 any reply from clients. Clients drops these locks when cancel
4095 notification from server is received without replying.
4098 Frequency : liblustre-only, when liblustre client dies or becomes busy
4100 Description: Doing ls on Linux clients can take a long time with active
4102 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4103 granting such locks to it in the first place if possible. Locks
4104 are taken by proxy on the OST during the read or write and
4105 dropped immediately afterward. Add connect flags handling, do
4106 not grant locks to liblustre clients for glimpse ASTs.
4108 Severity : enhancement
4110 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4111 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4112 sequential reads and behaves badly with applications doing random
4113 reads. Improve it by reading ahead at least read region, and
4114 avoiding excessive large RPC for small reads.
4116 Severity : enhancement
4118 Description: Creating more than 1000 files for a single job may cause a load
4119 imbalance on the OSTs if there are also a large number of OSTs.
4120 Details : qos_prep_create() uses an OST index reseed value that is an
4121 even multiple of the number of available OSTs so that if the
4122 reseed happens in the middle of the object allocation it will
4123 still utilize the OSTs as uniformly as possible.
4128 Description: OST or MDS may oops in ping_evictor_main()
4129 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4130 but doesn't restart at beginning of obd_exports_timed list
4133 Severity : enhancement
4135 Description: improve by-nid export eviction on the MDS and OST
4136 Details : allow multiple exports with the same NID to be evicted at one
4137 time without re-searching the exports list.
4140 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4142 Description: MDS may oops in groups_free()
4143 Details : in rare race conditions a newly allocated group_info struct is
4144 freed again, and this can be NULL. The 2.4 compatibility code
4145 for groups_free() checked for a NULL pointer, but 2.6 did not.
4148 Frequency : common for liblustre clients doing little filesystem IO
4149 Bugzilla : 9352, 7313
4150 Description: server may evict liblustre clients accessing contended locks
4151 Details : if a client is granted a lock or receives a completion AST
4152 with a blocking AST already set it would not reply to the AST
4153 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4156 Frequency : lfs setstripe, only systems with more than 160 OSTs
4158 Description: unable to set striping with a starting offset beyond OST 160
4159 Details : llapi_create_file() incorrectly limited the starting stripe
4160 index to the maximum single-file stripe count.
4163 Frequency : LDAP users only
4165 Description: lconf did not handle in-kernel recovery with LDAP properly
4166 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4168 Severity : enhancement
4170 Description: bind OST threads to NUMA nodes to improve performance
4171 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4172 node and do their allocations there to localize memory access
4174 Severity : enhancement
4176 Description: llmount can determine client NID directly from Myrinet (GM)
4177 Details : the client NID code from gmnalnid was moved directly into
4178 llmount, removing the need to use this or specifying the
4179 client NID explicitly when mounting GM clients with zeroconf
4182 Frequency : if client is started with down MDS
4184 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4185 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4186 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4187 ptlrpc_import_delayed_req() because we didn't check if the
4188 request was interrupted, and we also didn't break out of the
4189 event loop if there was no timeout
4194 Description: data loss during non-page-aligned writes to a single file from
4195 both multiple nodes and multiple threads on one node at same time
4196 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4197 inconsistency led to false short-reads, that were cached and later
4198 used by ->prepare_write() to fill in partially written page,
4199 leading to data loss.
4202 Frequency : always, if lconf --abort_recovery used
4204 Description: lconf --abort_recovery fails with 'Operation not supported'
4205 Details : lconf was attempting to abort recovery on the MDT device and not
4208 Severity : enhancement
4210 Description: remove cleanup logs
4211 Details : replace lconf-generated cleanup logs with lustre internal
4212 cleanup routines. Eliminates the need for client-cleanup and
4215 Severity : enhancement
4217 Description: add support for EAs (user and system) on lustre filesystems
4218 Details : it is now possible to store extended attributes in the Lustre
4219 client filesystem, and with the user_xattr mount option it
4220 is possible to allow users to store EAs on their files also
4222 Severity : enhancement
4224 Description: Add possibility (config option) to show minimal available OST free
4226 Details : When compiled with --enable-mindf configure option, statfs(2)
4227 (and so, df) will return least minimal free space available from
4228 all OSTs as amount of free space on FS, instead of summary of
4229 free spaces of all OSTs.
4231 Severity : enhancement
4233 Description: do not expand extent locks acquired on OST-side
4234 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4235 by server: they are not cached anyway.
4238 Frequency : when mmap is used/binaries executed from Lustre
4240 Description: Unmmap pages before throwing them away from read cache.
4241 Details : llap_shrink cache now attempts to unmap pages before discarding
4242 them (if unmapping failed - do not discard). SLES9 kernel has
4243 extra checks that trigger if this unmapping is not done first.
4248 Description: lconf didn't resolve symlinks before checking to see whether a
4249 given mountpoint was already in use
4252 Frequency : when migrating failover services
4253 Bugzilla : 6395, 9514
4254 Description: When migrating a subset of services from a node (e.g. failback
4255 from a failover service node) the remaining services would
4256 time out and evict clients.
4257 Details : lconf --force (implied by --failover) sets the global obd_timeout
4258 to 5 seconds in order to quickly disconnect, but this caused
4259 other RPCs to time out too quickly. Do not change the global
4260 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4262 Severity : enhancement
4263 Frequency : if MDS is started with down OST
4264 Bugzilla : 9439,5706
4265 Description: Allow startup/shutdown of an MDS without depending on the
4266 availability of the OSTs.
4267 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4268 Add appropriate locking and lov-osc refcounts for safe
4269 cleaning. Add osc abort_inflight calls in case the
4270 synchronize never started.
4273 Frequency : occasional (Cray XT3 only)
4275 Description: root not authorized to access files in CRAY_PORTALS environment
4276 Details : The client process capabilities were not honoured on the MDS in
4277 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4278 been accessed by an authorized user then root was able to access
4279 the file on the local client also. The root user capabilities
4280 are now allowed on the MDS, as this environment has secure UID.
4283 Frequency : occasional
4285 Description: ldiskfs "too long searching" message happens too often
4286 Details : A debugging message (otherwise harmless) prints too often on
4287 the OST console. This has been reduced to only happen when
4288 there are fragmentation problems on the filesystem.
4293 Description: Division by zero in statfs when all OSCs are inactive
4294 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4295 lov_statfs divides by value returned by lov_get_stripecnt().
4299 Bugzilla : 9489, 3273
4300 Description: First write from each client to each OST was only 4kB in size,
4301 to initialize client writeback cache, which caused sub-optimal
4302 RPCs and poor layout on disk for the first writen file.
4303 Details : Clients now request an initial cache grant at (re)connect time
4304 and so that they can start streaming writes to the cache right
4305 away and always do full-sized RPCs if there is enough data.
4306 If the OST is rebooted the client also re-establishes its grant
4307 so that client cached writes will be honoured under the grant.
4312 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4313 Details : Now I/O RPCs go to different portal number and (presumably) fast
4314 lock requests (and glimses) and other RPCs get their own service
4315 threads pool that should be able to service those RPCs
4318 Severity : enhancement
4320 Description: Ability to exchange lustre version between client and servers and
4321 issue warnings at client side if client is too old. Also for
4322 liblustre clients there is ability to refuse connection of too old
4324 Details : New 'version' field is added to connect data structure that is
4325 filled with version info. That info is later checked by server and
4329 Frequency : rare, liblustre only.
4330 Bugzilla : 9296, 9581
4331 Description: Two simultaneous writes from liblustre at offset within same page
4332 might proceed at the same time overwriting eachother with stale
4334 Details : I/O lock withing llu_file_prwv was released too early, before data
4335 actually was hitting the wire. Extended lock-holding time until
4336 server acknowledges receiving data.
4339 Frequency : extremely rare. Never observed in practice.
4341 Description: avoid generating lustre_handle cookie of 0.
4342 Details : class_handle_hash() generates handle cookies by incrementing
4343 global counter, and can hit 0 occasionaly (this is unlikely, but
4344 not impossible, because initial value of cookie counter is
4345 selected randonly). Value of 0 is used as a sentinel meaning
4346 "unassigned handle" --- avoid it. Also coalesce two critical
4347 sections in this function into one.
4349 Severity : enhancement
4351 Description: allow liblustre clients to delegate truncate locking to OST
4352 Details : To avoid overhead of locking, liblustre client instructs OST to
4353 take extent lock in ost_punch() on client's behalf. New connection
4354 flag is added to handle backward compatibility.
4356 Severity : enhancement
4357 Bugzilla : 4928, 7341, 9758
4358 Description: allow number of OST service threads to be specified
4359 Details : a module parameter allows the number of OST service threads
4360 to be specified via "options ost ost_num_threads={N}" in the
4361 OSS's /etc/modules.conf or /etc/modprobe.conf.
4365 Bugzilla : 6146, 9635, 9895
4366 Description: servers crash with bad pointer in target_handle_connect()
4367 Details : In rare cases when a client is reconnecting it was possible that
4368 the connection request was the last reference for that export.
4369 We would temporarily drop the export reference and get a new
4370 one, but this may have been the last reference and the export
4371 was just destroyed. Get new reference before dropping old one.
4373 Severity : enhancement
4374 Frequency : if client is started with failover MDS
4376 Description: Allow multiple MDS hostnames in the mount command
4377 Details : Try to read the configuration from all specified MDS
4378 hostnames during a client mount in case the "primary"
4381 Severity : enhancement
4383 Description: Stop sending data to evicted clients as soon as possible.
4384 Details : Check if the client we are about to send or are sending data to
4385 was evicted already. (Check is done every second of waiting,
4386 for which l_wait_event interface was extended to allow checking
4387 of exit condition at specified intervals).
4390 Frequency : rare, normally only when NFS exporting is done from client
4392 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4394 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4397 Frequency : always, if srand() is called before liblustre initialization
4399 Description: Liblustre uses system PRNG disturbing its usage by user application
4400 Details : Introduce internal to lustre fast and high-quality PRNG for
4401 lustre usage and make liblustre and some other places in generic
4402 lustre code to use it.
4404 Severity : enhancement
4405 Bugzilla : 9477, 9557, 9870
4406 Description: Verify that the MDS configuration logs are updated when xml is
4407 Details : Check if the .xml configuration logs are newer than the config
4408 logs stored on the MDS and report an error if this is the case.
4409 Request --write-conf, or allow starting with --old_conf.
4411 Severity : enhancement
4413 Description: Handle symlinks in the path when checking if Lustre is mounted.
4414 Details : Resolve intermediate symlinks when checking if a client has
4415 mounted a filesystem to avoid duplicate client mounts.
4420 Description: lconf can hit an error exception but still return success.
4421 Details : The lconf command catches the Command error exception at the top
4422 level script context and will exit with the associated exit
4423 status, but doesn't ensure that this exit status is non-zero.
4428 Description: failure of ptlrpc thread startup can cause oops
4429 Details : Starting a ptlrpc service thread can fail if there are a large
4430 number of threads or the server memory is very fragmented.
4431 Handle this without oopsing.
4434 Frequency : always, only if liblustre and non-default acceptor port was used
4436 Description: liblustre cannot connect to servers with non-default acceptor port
4437 Details : tcpnal_set_default_params() was not called and was therefore
4438 ignoring the environment varaible TCPNAL_PORT, as well as other
4439 TCPNAL_ environment variables
4444 Description: two objects could be created on the same OST for a single file
4445 Details : If an OST is down, in some cases it was possible to create two
4446 objects on a single OST for a single file. No problems other
4447 than potential performance impact and spurious error messages.
4451 Bugzilla : 5681, 9562
4452 Description: Client may oops in ll_unhash_aliases
4453 Details : Client dcache may become inconsistent in race condition.
4454 In some cases "getcwd" can fail if the current directory is
4460 Description: Inode refcounting problems in NFS export code
4461 Details : link_raw functions used to call d_instantiate without obtaining
4462 extra inode reference first.
4466 Bugzilla : 9942, 9903
4467 Description: Referencing freed requests leading to crash, memleaks with NFS.
4468 Details : We used to require that call to ll_revalidate_it was always
4469 followed by ll_lookup_it. Also with revalidate_special() it is
4470 possible to call ll_revalidate_it() twice for the same dentry
4471 even if first occurence returned success. This fix changes semantic
4472 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4473 reference on a request referred from the intent.
4474 ll_intent_release() then releases such a request.
4477 Frequency : rare, normally benchmark loads only
4479 Description: unlinked inodes were kept in memory on the client
4480 Details : If a client is repeatedly creating and unlinking files it
4481 can accumulate a lot of stale inodes in the inode slab cache.
4482 If there is no other client load running this can cause the
4483 client node to run out of memory. Instead flush old inodes
4484 from client cache that have the same inode number as a new inode.
4487 Frequency : SLES9 2.6.5 kernel and long filenames only
4488 Bugzilla : 9969, 10379
4489 Description: utime reports stale NFS file handle
4490 Details : SLES9 uses out-of-dentry names in some cases, which confused
4491 the lustre dentry revalidation. Change it to always use the
4495 Frequency : rare, unless heavy write-truncate concurrency is continuous
4496 Bugzilla : 4180, 6984, 7171, 9963, 9331
4497 Description: OST becomes very slow and/or deadlocked during object unlink
4498 Details : filter_destroy() was holding onto the parent directory lock
4499 while truncating+unlinking objects. For very large objects this
4500 may block other threads for a long time and slow overall OST
4501 responsiveness. It may also be possible to get a lock ordering
4502 deadlock in this case, or run out of journal credits because of
4503 the combined truncate+unlink. Solution is to do object truncate
4504 first in one transaction without parent lock, and then do the
4505 final unlink in a new transaction with the parent lock. This
4506 reduces the lock hold time dramatically.
4509 Frequency : rare, 2.4 kernels only
4511 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4512 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4513 before unmounting their filesystem in order to avoid deadlock.
4514 The kernel_locked() function in 2.4 kernels only checks whether
4515 the kernel lock is held, not whether it is this process that is
4516 holding it as 2.6 kernels do.
4521 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4522 Details : The client ptlrpc code may be trying to reconnect to a down
4523 server before a previous connection attempt has timed out.
4524 Increase the reconnect interval to be longer than the connection
4525 timeout interval to avoid sending duplicate connections to
4529 Frequency : echo_client brw_test command
4531 Description: fix echo_client to work with OST preallocated code
4532 Details : OST preallocation code (5137) didn't take echo_client IO path
4533 into account: echo_client calls filter methods outside of any
4534 OST thread and, hence, there is no per-thread preallocated
4535 pages and buffers to use. Solution: hijack pga pages for IO. As
4536 a byproduct, this avoids unnecessary data copying.
4540 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4541 Description: Client can oops in mdc_commit_close() after open replay
4542 Details : It was possible for the MDS to return an open request with no
4543 transaction number in mds_finish_transno() if the client was
4544 evicted, but without actually returning an error. Clients
4545 would later try to replay that open and may trip an assertion
4546 Simplify the client close codepath, and always return an error
4547 from the MDS in case the open is not successful.
4550 Frequency : rare, 2.6 OSTs only
4552 Description: OST may deadlock under high load on fragmented files
4553 Details : If there was a heavy load and highly-fragmented OST filesystems
4554 it was possible to have all the OST threads deadlock waiting on
4555 allocation of biovecs, because the biovecs were not released
4556 until the entire RPC IO was completed. Instead, release biovecs
4557 as soon as they are complete to ensure forward IO progress.
4559 Severity : enhancement
4561 Description: Support for specifying external journal device at mount
4562 Details : If an OST or MDS device is formatted with an external journal
4563 device, this device major/minor is stored in the ext3 superblock
4564 and may not be valid for failover. Allow detecting and
4565 specifying the external journal at mount time.
4570 Description: Mounting an MDS with pending unlinked files may cause oops
4571 Details : target_finish_recovery() calls mds_postrecov() which returned
4572 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4573 considers this an error and immediately begins cleaning up the
4574 lov, just after starting the mds_lov process
4576 Severity : enhancement
4578 Description: Implement 'lfs df' to report actual free space on per-OST basis
4579 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4580 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4581 usage of objects; '-h' to report in human readable format.
4583 ------------------------------------------------------------------------------
4585 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4592 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4593 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4594 at startup during orphan unlinking. Instead, skip buddy update
4595 before setup, as it will be regenerated after recovery is complete.
4600 Description: If an OST is inactive, its locks might reference stale inodes.
4601 Details : lov_change_cbdata() must iterate over all namespaces, even if
4602 they are inactive to clear inode references from the lock.
4604 Severity : enhancement
4605 Frequency : occasional, if non-standard max_dirty_mb used
4607 Description: Client will block write RPCs if not enough grant
4608 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4609 then the client will block writes while waiting for another RPC
4610 to complete instead of consuming its dirty limit. With change
4611 we get improved performance when max_dirty_mb is small.
4613 Severity : enhancement
4614 Bugzilla : 3389, 6253
4615 Description: Add support for supplementary groups on the MDS.
4616 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4617 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4618 lookups for user supplementary groups into a cache.
4622 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4623 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4624 for liblustre O_EXCL is enforced by MDS.
4628 Description: readdir never returns NULL in liblustre.
4629 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4630 dentry in struct dirent.
4634 Frequency : liblustre only, depends on application IO pattern
4635 Description: liblustre clients evicted if not contacting servers
4636 Details : Don't put liblustre clients into the ping_evictor list, so
4637 they will not be evicted by the pinger ever.
4639 Severity : enhancement
4641 Description: Add ability to evict clients by NID from MDS.
4642 Details : By echoing "nid:$NID" string into
4643 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4644 $NID would be instantly evicted from this MDS and from all active
4645 OSTs connected to it.
4649 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4650 Details : lookup_it_finish() used to query file size from OSTs that was not
4655 Description: service threads change working directory to that of init
4656 Details : Starting lustre service threads may pin the working directory
4657 of the parent thread, making that filesystem busy. Threads
4658 now change to the working directory of init to avoid this.
4662 Frequency : during shutdown only
4663 Description: shutdown with a failed MDS or OST can cause unmount to hang
4664 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4669 Frequency : 2.6 only
4670 Description: chmod/chown may include an extra supplementary group
4671 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4672 supplementary group and if none is specified this is used.
4675 Bugzilla : 5479 (6816)
4677 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4678 Details : If lookup is in progress on a file that is unlinked we might try
4679 to revalidate the inode and fail in revalidate after lookup is
4680 complete and ll_file_open() enqueues the open again but
4681 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4684 Frequency : always, if lconf --abort_recovery used
4686 Description: lconf --abort_recovery fails with 'Operation not supported'
4687 Details : lconf was attempting to abort recovery on the MDT device and not
4690 ------------------------------------------------------------------------------
4692 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4697 Frequency : rare (only unsupported configurations with a node running as an
4699 Bugzilla : 6514, 5137
4700 Description: Mounting a Lustre file system on a node running as an OST could
4702 Details : OSTs now preallocates memory needed to write out data at
4703 startup, instead of when needed, to avoid having to
4704 allocate memory in possibly low memory situations.
4705 Specifically, if the file system is mounted on on OST,
4706 memory pressure could force it to try to write out data,
4707 which it needed to allocate memory to do. Due to the low
4708 memory, it would be unable to do so and the node would
4709 become unresponsive.
4711 Severity : enhancement
4713 Description: Addition of lconf --service command line option
4714 Details : lconf now accepts a '--service <arg>' option, which is
4715 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4717 Severity : enhancement
4719 Description: Failover mode is now the default for OSTs.
4720 Details : By default, OSTs will now run in failover mode. To return to
4721 the old behaviour, add '--failout' to the lmc line for OSTs.
4723 Severity : enhancement
4725 Description: Health checks are now provided for MDS and OSTs
4726 Details : Additional detailed health check information on MSD and OSTs
4727 is now provided through the procfs health_check value.
4730 Frequency : occasional, depends on IO load
4732 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4733 after numerous create/delete cycles
4734 Details : The ext3 inode allocation policy would not allocate new inodes
4735 very well on the OSTs because there are no new directories
4736 being created. Instead we look for groups with free space if
4737 the parent directories are nearly full.
4742 Description: Network or server problems during mount may cause partially
4743 mounted clients instead of returning an error.
4744 Details : The config llog parsing code may overwrite the error return
4745 code during mount error handling, returning success instead
4751 Description: MDS can fail to allocate large reply buffers
4752 Details : After long uptimes the MDS can fail to allocate large reply
4753 buffers (e.g. zconf client mount config records) due to memory
4754 fragmentation or consumption by the buffer cache. Preallocate
4755 some large reply buffers so that these replies can be sent even
4756 under memory pressure.
4760 Frequency : rare (liblustre)
4761 Description: fsx running with liblustre complained that using truncate() to
4762 extend the file doesn't work. This patch corrects that issue.
4763 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4764 ATTR_SIZE and lsm use in llu_setattr_raw.
4768 Frequency : rare, only 2.6 kernels
4769 Description: Unusual file access patterns on the MDS may result in inode
4770 data being lost in very rare circumstances.
4771 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4772 code in the 2.6 kernel caused the inode and/or EA data not to
4773 be read from disk, causing single-file corruption.
4777 Frequency : rare, only 2.6 filesystems using extents
4778 Description: Heavy concurrent write and delete load may cause data corruption.
4779 Details : It was possible under high-load situations to have an extent
4780 metadata block in the block device cache from a just-unlinked
4781 file overwrite a newly-allocated data block. We now unmap any
4782 metadata buffers that alias just-allocated data blocks.
4786 Frequency : filesystems with default stripe_count larger than 77
4787 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4788 Details : lconf specifies an inode size of 4096 bytes when the default
4789 stripe_count is larger than 77. This conflicts with the default
4790 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4791 this case to avoid pinning too much memory for large EAs.
4793 ------------------------------------------------------------------------------
4795 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4800 Frequency : rare (extremely heavy IO load with hundreds of clients)
4802 Description: Client is evicted, gets IO error writing to file
4803 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4804 caused clients to be evicted for AST timeouts. The fixes in
4805 bug 5192 mean we no longer need to have such short AST timeouts
4806 so ldlm_timeout has been increased.
4809 Frequency : occasional during --force or --failover shutdown under load
4810 Bugzilla : 5949, 4834
4811 Description: Server oops/LBUG if stopped with --force or --failover under load
4812 Details : a collection of import/export refcount and cleanup ordering
4813 issues fixed for safer force cleanup
4816 Frequency : only filesystems larger than 120 OSTs
4817 Bugzilla : 5990, 6223
4818 Description: lfs getstripe would oops on a very large filesystem
4819 Details : lov_getconfig used kfree on vmalloc'd memory
4822 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4823 Bugzilla : 6242, 6243
4824 Description: reading from files that had been truncated to a non-zero size
4825 but never opened returned no data
4826 Details : ll_file_read() reads zeros from no-object files to EOF
4831 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4832 Details : There's little harm in aborting MDS/OSS recovery and letting it
4833 try again, so I removed the LASSERT and return an error instead.
4835 Severity : enhancement
4837 Description: New debugging infrastructure for tracking down data corruption
4838 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4839 (b) cover more of the client-side code path, and (c) try to narrow
4840 down where problems occurred
4844 Bugzilla : 3819, 4364, 4397, 6313
4845 Description: Racing close and eviction MDS could cause assertion in mds_close
4846 Details : It was possible to get multiple mfd references during close and
4847 client eviction, leading to one thread referencing a freed mfd.
4849 Severity: : enhancement
4850 Bugzilla : 3262, 6359
4851 Description: Attempts to reconnect to servers are now more aggressive.
4852 Details : This builds on the enhanced upcall-less recovery that was added
4853 in 1.4.2. When trying to reconnect to servers, clients will
4854 now try each server in the failover group every 10 seconds. By
4855 default, clients would previously try one server every 25 seconds.
4860 Description: After recovery, certain operations trigger a failed
4861 assertion on a client.
4862 Details : Failing over an mds, using lconf -d --failover, while a
4863 client was doing a readdir() call would cause the client to
4864 LBUG after recovery completed and the readdir() was resent.
4866 Severity : enhancement
4868 Description: Default groups are now added by lconf
4869 Details : You can now run lconf --group <servicename> without having to
4870 manually add groups with lmc.
4873 Frequency : occasional
4875 Description: Nodes with an elan id of 0 trigger a failed assertion
4878 Frequency : always when accessing e.g. tty/console device nodes
4880 Description: tty and some other devices nodes cannot be used on lustre
4881 Details : file's private_data field is used by device data and lustre
4882 values in there got lost. New field was added to struct file to
4883 store fs-specific private data.
4886 Frequency : when exporting Lustre via NFS
4888 Description: NFSD failed occasionally when looking up a path component
4889 Details : NFSD is looking up ".." which was broken in ext3 directories
4890 that had grown large enough to become hashed.
4893 Frequency : Clusters with multiple interfaces not on the same subnet
4895 Description: Nodes will repeatedly try to reconnect to an interface which it
4896 cannot reach and report an error to the log.
4897 Details : Extra peer list entries will be created by lconf with some peers
4898 unreachable. lconf now validates the peer before adding it.
4901 Frequency : Only if a default stripe is set on the filesystem root.
4903 Description: Setting a default stripe on the filesystem root prevented the
4904 filesystem from being remounted.
4905 Details : The client was sending extra request flags in the root getattr
4906 request and did not allocate a reply buffer for the dir EA.
4909 Frequency : occasional, higher if lots of files are accessed by one client
4910 Bugzilla : 6159, 6097
4911 Description: Client trips assertion regarding lsm mismatch/magic
4912 Details : While revalidating inodes the VFS looks up inodes with ifind()
4913 and in rare cases can find an inode that is being freed.
4914 The ll_test_inode() code will free the lsm during ifind()
4915 when it finds an existing inode and then the VFS later attaches
4916 this free lsm to a new inode.
4920 Bugzilla : 6422, 7030
4921 Description: MDS deadlock between mkdir and client eviction
4922 Details : Creating a new file via mkdir or mknod (starting a transaction
4923 and getting the ns lock) can deadlock with client eviction
4924 (gets ns lock and trying to finish a synchronous transaction).
4927 Frequency : occasional
4928 Description: While starting a server, the fsfilt_ext3 module could not be
4930 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4931 kernels. Previously, lconf would still use the ext3 name
4932 when trying to load modules. Now, it will correctly use
4933 ext3 on 2.4 and ldiskfs on 2.6.
4935 Severity : enhancement
4936 Description: The default stripe count has been changed to 1
4937 Details : The interpretation of the default stripe count (0, to lfs
4938 or lmc) has been changed to mean striping across a single
4939 OST, rather than all available. For general usage we have
4940 found a stripe count of 1 or 2 works best.
4942 Severity : enhancement
4943 Description: Add support for compiling against Cray portals.
4944 Details : Conditional compiling for some areas that are different
4948 Frequency : occasional
4949 Bugzilla : 6409, 6834
4950 Description: Creating files with an explicit stripe count may lead to
4951 a failed assertion on the MDS
4952 Details : If some OSTs are full or unavailable, creating files may
4953 trigger a failed assertion on the MDS. Now, Lustre will
4954 try to use other servers or return an error to the
4958 Frequency : occasional
4960 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4961 caused lock timeout during readahead (since 1.4.2).
4962 Details : Processes doing readahead might match a lock that hasn't been
4963 granted yet if there are overlapping and conflicting lock
4964 requests. The readahead process waits on ungranted lock
4965 (original lock is CBPENDING), while OST waits for that process
4966 to cancel CBPENDING read lock and eventually evicts client.
4968 Severity : enhancement
4970 Description: Initial enabling of flock support for clients
4971 Details : Implements fcntl advisory locking and file status functions.
4972 This feature is provided as an optional mount flag (default
4973 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4974 locking are implemented yet, and those that are are not guaranteed
4975 to be completely correct in production environments.
4976 mount -t lustre -o [flock|noflock] ...
4979 Frequency : occasional
4981 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4982 assertion in the ext3 JBD (journaling) layer.
4983 Details : The b_committed_data struct is protected by the big kernel lock
4984 in 2.4 kernels, serializing journal_commit_transaction() and
4985 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4986 access to this struct. In 2.6 kernels there is finer grained
4987 locking to improve SMP performance of the JBD layer.
4991 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4996 Description: merge and cleanup kernel patches.
4999 -----------------------------------------------------------------------------
5001 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5003 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5004 versions of Lustre. Please update all servers and clients to
5005 version 1.4.2 or later at the same time. You must also run
5006 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5007 to update the configuration logs.
5009 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5010 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5011 - swab llog records of type '0' so we get proper header size/idx (5861)
5012 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5013 - fix rename of one directory over another leaking an inode (5953)
5014 - avoid SetPageDirty on 2.6 (5981)
5015 - don't re-add just-being-destroyed locks to the waiting list (5653)
5016 - when creating new directories, inherit the parent's custom
5017 striping settings if present parent (3048)
5018 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5019 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5020 - don't hold client locks on temporary worklist from l_lru (5666)
5021 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5022 - automatically evict dead clients (5921)
5023 - Update file size properly in create+truncate+fstat case (6196)
5024 - Do not unhash mountpoint dentries, do not allow removal of
5026 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5027 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5028 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5029 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5030 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5031 - Correctly handle reads of files with no objects (6243)
5032 - lctl recover will also mark a device active if deactivate used (5933)
5034 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5035 - allow --write-conf on an MDS with different nettype than client (5619)
5036 - don't write config llogs to MDS for mounts not from that MDS (5617)
5037 - lconf should create multiple TCP connections from a client (5201)
5038 - init scripts are now turned off by default; run chkconfig --on
5039 lustre and chkconfig --on lustrefs to use them
5040 - upcalls are no longer needed for clients to recover to failover
5042 - add --abort-recovery option to lconf to abort recovery on device
5044 - add support for an arbitrary number of OSTs (3026)
5045 - Quota support protocol changes.
5046 - forward compatibility changes to wire structs (6007)
5047 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5048 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5049 - improved "lustre status" script
5050 - initialize blocksize for non-regular files (6062)
5051 - added --disable-server and --disable-client configure options (5782)
5052 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5053 - Vanilla 2.4.29 support
5054 - increase maximum number of obd devices to 520 (6242)
5055 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5056 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5057 SLES 9 2.6.5 kernels
5058 - limit stripes per file to 160 (the maximum EA size) (6093)
5060 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5063 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5064 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5065 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5066 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5067 - small ext3 extents cleanups and fixes (5733)
5068 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5069 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5070 - avoid needless client->OST connect, fix handle mismatch (5317)
5071 - fix DLM error path that led to out-of-sync client, long delays (5779)
5072 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5073 - fix several locking issues related to i_size (5492,5624,5654,5672)
5074 - don't move pending lock onto export if it is already evicted (5683)
5075 - fix kernel oops when creating .foo in unlinked directory (5548)
5076 - fix deadlock in obdfilter statistics vs. object create (5811)
5077 - use time_{before,after} to avoid timer jiffies wrap (5882)
5078 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5079 - Do not leak request if server was not able to process it (5154)
5080 - If mds_open unable to find parent dir, make that negative lookup(5154)
5081 - don't create new directories with extent-mapping (5909, 5936)
5083 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5084 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5085 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5086 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5087 - support for RHEL4 kernel on i686 (5773)
5088 - provide error messages when incompatible logs are encountered (5898)
5090 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5091 * version 1.4.0.10 (1.4.1 release candidate 1)
5093 - don't keep a lock reference when lock is not granted (4238)
5094 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5095 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5096 - drop import inflight refcount on signal_completed_replay error (5255)
5097 - unlock page after async write error during send (3677)
5098 - handle missing objects in filter_preprw_read properly (5265)
5099 - no transno return for symlink open, don't save no-trasno open (3440)
5100 - don't try to complete elan receive that already failed (4012)
5101 - free RPC server reply state on error (5406)
5102 - clean up thread from ptlrpc_start_thread() on error (5160)
5103 - readahead could read extra page into cache that wasn't ejected (5388)
5104 - prevent races in class_attach/setup/cleanup/detach (5260)
5105 - don't dereference de->d_inode after l_dput of de (5458)
5106 - use "int" for stripe value returned from lock_to_stripe (5544)
5107 - mballoc allocation and error-checking fixes in 2.6 (5504)
5108 - block device patches to fix I/O request sizes in 2.6 (5482)
5109 - look up hostnames for IB nals (5602)
5110 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5111 - don't start multiple acceptors for the same port (5277)
5112 - fix incorrect LASSERT in mds_getattr_name (5635)
5113 - export a proc file for general "ping" checking (5628)
5114 - fix "lfs check" to not block when the MDS is down (5628)
5116 - service request history (4965)
5117 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5118 - create an "evict_client" /proc entry on OSTs, like the MDS has
5119 - fix mount usage message, return errors per mount(8) (5168)
5120 - change grep [] to grep "[]" in tests so they work in more UMLs
5121 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5122 - remove ancient LOV_MAGIC_V0 EA support (5047)
5123 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5124 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5125 - allow readahead allocations to fail when low on memory (5383)
5126 - mmap locking landed again, after considerable improvement (2828)
5127 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5129 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5132 - send OST transaction number in read/write reply to free req (4966)
5133 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5134 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5135 - the watchdog thread now runs as interruptible (5246)
5136 - flock/lockf fixes (but it's still disabled, pending 5135)
5137 - don't use EXT3 constants in llite code (5094)
5138 - memory shortage at startup could cause assertion (5176)
5140 - reorganization of lov code
5141 - single portals codebase
5143 - add extents/mballoc support (5025)
5144 - direct I/O reads in the obdfilter (4048)
5145 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5147 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5150 - send OST transaction number in read/write reply to free req (4966)
5151 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5152 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5153 - the watchdog thread now runs as interruptible (5246)
5154 - handle missing objects in filter_preprw_read properly (5265)
5155 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5156 - drop import inflight refcount on signal_completed_replay error (5255)
5157 - unlock page after async write error during send (3677)
5158 - return original error code on reconstructed replies (3761)
5159 - no transno return for symlink open, don't save no-trasno open (3440)
5161 - add pid to ldlm debugging output (4922)
5162 - bump the watchdog timeouts -- we can't handle 30sec yet
5163 - extra debugging for orphan dentry/inode bug (5259)
5165 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5168 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5169 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5170 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5171 - fix NULL dereference in /proc/sys/portals/routes (4827)
5172 - allow failed mdc_close() operations to be interrupted (4561)
5173 - stop precreate on OST before MDS would time out on it (4778)
5174 - don't send partial-page writes before EOF from client (4410)
5175 - discard client grant for sub-page writes on large-page clients (4520)
5176 - don't free dentries not owned by NFS code, check generation (4806)
5177 - fix lsm leak if mds_create_objects() fails (4801)
5178 - limit debug_daemon file size, always print CERROR messages (4789)
5179 - use transno after validating reply (3892)
5180 - process timed out requests if import state changes (3754)
5181 - update mtime on OST during writes, return in glimpse (4829)
5182 - add mkfsoptions to LDAP (4679)
5183 - use ->max_readahead method instead of zapping global ra (5039)
5184 - don't interrupt __l_wait_event() during strace
5186 - add software watchdogs to catch hung threads quickly (4941)
5187 - make lustrefs init script start after nfs is mounted
5188 - fix CWARN/ERROR duplication (4930)
5189 - return async write errors to application if possible (2248)
5190 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5191 - print NAL number in %x format (4645)
5192 - update barely-supported suse-2.4.21-171 series (4842)
5193 - support for sles 9 %post scripts
5194 - support for building 2.6 kernel-source packages
5195 - support for sles km_* packages
5197 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5200 - ignore -ENOENT errors in osc_destroy (3639)
5201 - notify osc create thread that OSC is being cleaned up (4600)
5202 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5203 - reconstruct ost_handle() like mds_handle() (4657)
5204 - create a new thread to do import eviction to avoid deadlock (3969)
5205 - let lconf resolve symlinked-to devices (4629)
5206 - don't unlink "objects" from directory with default EA (4554)
5207 - hold socknal file ref over connect in case target is down (4394)
5208 - allow more than 32000 subdirectories in a single directory (3244)
5209 - fix blocks count for O_DIRECT writes (3751)
5210 - OST returns ENOSPC from object create when no space left (4539)
5211 - don't send truncate RPC if file size isn't changing (4410)
5212 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5213 - bind to privileged port in socknal and tcpnal (3689)
5215 - rate limit CERROR/CWARN console message to avoid overload (4519)
5216 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5217 - basic mmap support (3918)
5218 - kernel patch series update from b1_4 (4711)
5220 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5223 - avoid crash during MDS cleanup with OST shut down (2775)
5224 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5225 - don't use bad inodes on the MDS (3744)
5226 - dynamic object preallocation to improve recovery speed (4236)
5227 - don't hold spinlock over lock dumping or change debug flags (4401)
5228 - don't zero obd_dev when it is force cleaned (3651)
5229 - print grants to console if they go negative (4431)
5230 - "lctl deactivate" will stop automatic recovery attempts (3406)
5231 - look for existing locks in ldlm_handle_enqueue() (3764)
5232 - don't resolve lock handle twice in recovery avoiding race (4401)
5233 - revalidate should check working dir is a directory (4134)
5235 - don't always mark "slow" obdfilter messages as errors (4418)
5237 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5240 - don't close LustreDB during write_conf until it is done (3860)
5241 - fix typo in lconf for_each_profile (3821)
5242 - allow dumping logs from multiple threads at one time (3820)
5243 - don't allow multiple threads in OSC recovery (3812)
5244 - fix debug_size parameters (3864)
5245 - fix mds_postrecov to initialize import for llog ctxt (3121)
5246 - replace config semaphore with spinlock (3306)
5247 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5248 - don't allow enqueue to complete on a destroyed export (3822)
5249 - down write_lock before checking llog header bitmap (3825)
5250 - recover from lock replay timeout (3764)
5251 - up llog sem before sending rpc (3652)
5252 - reduce ns lock hold times when setting kms (3267)
5253 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5254 - fix NULL deref and obd_dev leak on setup error (3312)
5255 - replace some LBUG about llog ops with error handling (3841)
5256 - don't match INVALID dentries from d_lookup and spin (3784)
5257 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5258 - fix invalid assertion in ptlrpc_set_wait (3880)
5260 - add libwrap support for the TCP acceptor (3996)
5261 - add /proc/sys/portals/routes for non-root route listing (3994)
5262 - allow setting MDS UUID in .xml (2580)
5263 - print the stack of a process that LBUGs (4228)
5265 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5268 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5269 - make sure to unset replay flag from failed open requests (3440)
5270 - if default stripe count is 0, use OST count for inode size (3636)
5271 - update parent mtime/ctime on client for create/unlink (2611)
5272 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5273 - free recovery state on server during a forced cleanup (3571)
5274 - unregister_reply for resent reqs (3063)
5275 - loop back devices mounting and status check on 2.6 (3563)
5276 - fix resource-creation race that can provoke i_size == 0 (3513)
5277 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5278 - more debugging for page-accounting assertion (3746)
5279 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5280 - avoid deadlock after precreation failure (3758)
5281 - fix race and lock order deadlock in orphan handling (3450, 3750)
5282 - add validity checks when grabbing inodes from l_ast_data (3599)
5284 - add /proc/.../recovery_status to obdfilter (3428)
5285 - lightweight CDEBUG infrastructure, debug daemon (3668)
5286 - change default OSC RPC parameters to be better on small clusters
5287 - turn off OST read cache for files smaller than 32MB
5288 - install man pages and include them in rpms (3100)
5289 - add new init script for (un)mounting lustre filesystems (2593)
5290 - run chkconfig in %post for init scripts (3701)
5291 - drop scimac NAL (unmaintained)
5293 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5296 - clean kiobufs before and after use (3485)
5297 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5298 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5299 - change init script to fail more clearly if not run as root (1528)
5300 - allow clients to reconnect during replay (1742)
5301 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5302 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5303 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5304 - fix cond_resched (3554)
5305 - only evict unfinished clients after recovery (3515)
5306 - allow bulk resend, prevent data loss (3570)
5307 - dynamic ptlrpc request buffer allocation (2102)
5308 - don't allow unlinking open directory if it isn't empty (2904)
5309 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5310 - remove extraneous obd dereference causing LASSERT failure (3334)
5311 - don't use get_cycles() when creating temp. files on the mds (3156)
5312 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5313 - handle EEXIST for set-stripe, set proper directory name (3336)
5315 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5316 - fix ksocknal_fmb_callback() error messages (2918)
5318 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5321 - don't copy lvb into (possibly NULL) reply on error (2983)
5322 - don't deref dentry after dput, don't free lvb on error (2922)
5323 - use the kms to determine writeback rpc length (2947)
5324 - increment oti_logcookies when osc is inactive (2948)
5325 - update client's i_blocks count via lvb messages (2543)
5326 - handle intent open/close of special files properly (1557)
5327 - mount MDS with errors=remount-ro, like obdfilter (2009)
5328 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5329 - don't use cancelling-locks' kms values (2947)
5330 - use highest lock extent for kms, not last one (2925)
5331 - don't dereference ERR_PTR() dentry in error handling path (3107)
5332 - fix thread race in portals_debug_dumplog() (3122)
5333 - create lprocfs device entries at setup instead of at attach (1519)
5334 - common AST error handler, don't evict client on completion race (3145)
5335 - zero nameidata in detach_mnt in 2.6 (3118)
5336 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5337 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5338 - initialize RPC timeout timer earlier for 2.6 (3219)
5339 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5340 - print nal/nid for unknown nid (3258)
5341 - additional checks for oscc recovery before doing precreate (3284)
5342 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5343 - don't crash in mdc_close for bad permissions on open (3285)
5344 - zero i_rdev for non-device files (3147)
5345 - clear page->private before handing to FS, better assertion (3119)
5346 - tune the read pipeline (3236)
5347 - fix incorrect decref of invalidated dentry (2350)
5348 - provide read-ahead stats and refine rpc in flight stats (3328)
5349 - don't hold journal transaction open across create RPC (3313)
5350 - update atime on MDS at close time (3265)
5351 - close LDAP connection when recovering to avoid server load (3315)
5352 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5353 - don't leak open file on MDS after open resend (3325)
5354 - serialize filter_precreate and filter_destroy_precreated (3329)
5355 - loop device shouldn't call sync_dev() for nul device (3092)
5356 - clear page cache after eviction (2766)
5357 - resynchronize MDS->OST in background (2824)
5358 - refuse to mount the same filesystem twice on same mountpoint (3394)
5359 - allow llmount to create routes for mounting behind routers (3320)
5360 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5361 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5362 - fix rare problem with rename on htree directories (3417)
5364 - allow default OST striping configuration per directory (1414)
5365 - fix compilation for qswnal for 2.6 kernels (3125)
5366 - increase maximum number of MDS request buffers for large systems
5367 - change liblustreapi to be useful for external progs like lfsck (3098)
5368 - increase local configuration timeout for slow disks (3353)
5369 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5371 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5374 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5375 - don't overwrite extent policy data in reply if lock was blocked (2901)
5376 - drop filter export grants atomically with removal from device (2663)
5377 - del obd_self_export from work_list in class_disconnect_exports (2908)
5378 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5379 - swab reply message in mdc_close, other PPC fixes (2464)
5380 - fix destroying of named logs (2325)
5381 - overwrite old logs when running lconf --write_conf (2264)
5382 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5383 - fix race in target_handle_connect (2898)
5384 - mds_reint_create() should take same inode create lock (2926)
5385 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5386 - don't close files for self_export to avoid uninitialized obd (2936)
5387 - allow MDS with the same name as client node (2939)
5388 - hold dentry reference for closed log files for unlink (2325)
5389 - reserve space for all logs during transactions (2059)
5390 - don't evict page beyond end of stripe extent (2925)
5391 - don't oops on a deleted current working directory (2399)
5392 - handle hard links to targets without a parent properly (2517)
5393 - don't dereference NULL lock when racing during eviction (2867)
5394 - don't grow lock extents when lots of conflicting locks (2919)
5396 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5399 - account for cache space usage on clients to avoid data loss (974)
5400 - lfsck support in lustre kernel code (2349)
5401 - reduce journal credits needed for BRW writes (2370)
5402 - orphan handling to avoid losing space on client/server crashes
5403 - ptlrpcd can be blocked, stopping ALL progress (2477)
5404 - use lock value blocks to assist in proper KMS, faster stat (1021)
5405 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5406 - recovery for initial connections (2355)
5407 - fixes for mds_cleanup_orphans (1934)
5408 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5409 - block all file creations until orphan recovery completes (1901)
5410 - client remove rq_connection from request struct (2423)
5411 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5412 - recovery timer race (2670)
5413 - mdc_close recovey bug (2532)
5414 - ptlrpc cleanup bug (2710)
5415 - mds timeout on local locks (2588)
5416 - namespace lock held during RPCs (2431)
5417 - handle interrupted sync write properly (2503)
5418 - don't try to handle a message that hasn't been replied to (2699)
5419 - client assert failure during cleanup after abort recovery (2701)
5420 - leak mdc device after failed mount (2712)
5421 - ptlrpc_check_set allows timedout requests to complete (2714)
5422 - wait for inflight reqs when ptlrpcd finishes (2710)
5423 - make sure unregistered services are removed from the srv_list
5424 - reset bulk XID's when resending them (caught by 1138 test)
5425 - unregister_bulk after timeout
5426 - fix lconf error (2694)
5427 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5428 - readahead locks pages, leaves pending causing memory pressure (2673)
5429 - increase OST request buffers to 4096 on large machines (2729)
5430 - fix up permission of existing directories in simple_mkdir (2661)
5431 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5432 - don't assign transno to errored transactions (2742)
5433 - don't delete objects on OST if given a bogus objid from MDS (2751)
5434 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5435 - if rq_no_resend, then timeout request after recovery (2432)
5436 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5437 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5438 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5439 - MDS recovery shouldn't delete objects that it has given out (2730)
5440 - if enqueue arrives after completion, don't clobber LVB (2819)
5441 - don't unlock pages twice when trigger_group_io returns error (2814)
5442 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5443 - don't write pages to disk if there was an error (1450)
5444 - don't ping imports that have recovery disabled (2676)
5445 - take buffered bytes into account when balancing socknal conn (2817)
5446 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5447 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5448 - remove little-endian swabbing of llog records (1987)
5449 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5450 - retry reposting request buffers if they fail (1191)
5451 - grow extent at grant time to avoid granting a revoked lock (2809)
5452 - lock revoke doesn't evict page if covered by a second lock (2765)
5453 - disable VM readahead to avoid reading outside lock extents (2805)
5455 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5456 - updated kernel patches for hp-2.4.20 kernel (2681)
5458 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5461 - fix truncated write corruption (2366)
5462 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5464 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5465 - ensure len(uuid) < 37 in lmc (1171)
5466 - fix ia64 OOPS in llog_test (2255)
5467 - zero end of page at obdfilter for partial page writes (2648)
5468 - don't leave stale dentries around after renames (bug 2428)
5469 - fix timeouts when evicting a client with a single lock held (2642)
5470 - set deadline for the initial HELLO message to drain (2634)
5471 - print out dotted-quad IP addresses in the socknal (2302)
5473 - additional debugging for MDS client eviction problem (2443)
5474 - fix mkfsoptions support for osts (2603, 2604)
5476 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5479 - add series for the vanilla 2.6.0 kernel
5480 - add series for the vanilla 2.4.24 kernel
5481 - add series for a cray x86/64 UL kernel drop
5482 - fix xattr patches for the vanilla 2.4.19 series
5484 - generate true UUIDs in lmc (1171)
5485 - have portals stack dumping break in UML (2466)
5486 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5487 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5488 - wake callback waiting threads on client eviction (2460)
5489 - Add --ptldebug and --subsystem to lmc (1719)
5490 - update assertion to allow safe interrupt allocation
5491 - set rq_no_resend for cancel requests (2432)
5492 - recalculate ptlrpcd timeout after resend (2494)
5493 - call vfs_rmdir when removing pending directories (2368)
5494 - fix renaming a file to itself (2429)
5495 - lmc creates a default one-stripe lov (2454)
5496 - expand procfs space to handle large clusters (2326)
5497 - increase UML stack to avoid overflow
5498 - update lconf's list of debug and subsystem masks
5499 - fix lfs find --obd (2510)
5500 - /proc tunable for disabling filter read caching (2591)
5501 - stop rpm packages from altering slapd.conf (2301)
5502 - disable nagle in the socknal under 0conf (2578)
5503 - choose mds inode size based on stripe count (2572)
5504 - fix kernel-source rpm problems (2516)
5506 - add --disable-doc to avoid pdf generation (2421)
5507 - update documentation, tests, type-os, comments
5508 - avoid format warnings on ia64
5509 - remove the TOE NAL
5510 - tiny code cleanups by removing unused fields
5512 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5515 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5516 - remove the most blatant lies from BUILDING file (2371)
5517 - change default debug level to reasonable production setting
5518 - reduce client side cache size to reduce cache flush time
5519 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5520 - make TCP zerocopy and pinger support enabled by default (2476)
5521 - sync writes completed after process exits caused crashes (2319)
5522 - maintain correct mount count on the MDS (2356)
5523 - backout 1557, because 2316 wasn't really fixed
5524 - better file I/O statistics gathering in /proc
5525 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5526 - another kernel patch to fix zero-copy TCP function export
5527 - don't take duplicate lock when processing re-sent getattr (2420)
5528 - lctl uses obd_self_export instead of creating new conn (2353)
5529 - MDS/OST recovery case which requires object creation asserted (2425)
5530 - move lfs from /usr/sbin to /usr/bin in packages
5531 - fix race between mds_client_add and mds_client_free (2417)
5532 - use kmalloc instead of slabs in portals (2430)
5533 - don't create duplicate records when a failover MDS is present (2442)
5534 - remove unnecessary mount age check (2332)
5535 - don't remove directory inodes from locks prematurely (2451)
5536 - don't break if MDS service name is the same as hostname (2103)
5537 - fix races in client write RPC generation when cache full (2482)
5539 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5542 - remove now-unused request->rq_obd (278)
5543 - if an allocation fails, print out how much memory we've used (1933)
5544 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5545 - add the "configurable stack size" patch to most series files (1256)
5546 - ability to write large log records, for 100+ OST configs (2306)
5547 - fix NULL deref when filter_prep fails (2314)
5548 - fix operator precedence error in filter_sync
5549 - dynamic allocation of socknal TX descriptors (2315)
5550 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5551 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5552 - ensure that CWARN messages in Portals always get to the syslog
5553 - __init/__exit are not for prototype decls (ldlm_init/exit)
5554 - x86-64 compile warning fixes
5555 - fix gateway LMC keyword conflict (2318)
5556 - fix MDS lock inversions in getattr/reint paths (1844)
5557 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5558 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5559 - DLM race condition prevented some lock evictions (2328)
5560 - ENOMEM detection and retry on socknal sends (2230)
5561 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5562 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5563 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5564 - better error messages when a client is rejected during recovery (1505)
5565 - avoid cancelling locks which were never granted, after failure (2330)
5566 - fix i_sem/journal inversion in mds_client_add (2333)
5567 - fix truncate/getattr lock cycle deadlock (2334)
5568 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5569 - fix two rare exit paths which could leak an l_lock() ref (2321)
5570 - fencepost error in MDS/OST orphan recovery (2226)
5571 - make log record alignment 8 bytes (1988)
5572 - lstripe now fails when requested offset > ost_count (2237)
5573 - ensure that all kernel series have a complete list.h (1607)
5574 - fix crashes in special-file operations (2316)
5575 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5577 - allow configurable automake binary, for testing new versions
5578 - small update to the lfs documentation
5580 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5582 * fix negative export reference count in fsfilt_sync (2312)
5584 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5585 * release candidate 0.9.1
5587 - orphans are moved into the PENDING directory for possible recovery
5588 - replayed opens now open by fid for orphan/rename safety (1042)
5589 - last close of an orphan inode generates a transno (683)
5590 - chdir() and mount() now pin the directory entry (1020)
5591 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5592 - discard very old requests without processing them (1502)
5593 - remove some common, well-understood CERRORs (1505)
5594 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5595 - clear "grant" flags in OST replies until OST grant code lands (1644)
5596 - fix read performance by not clobbering i_blksize on client (1598)
5597 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5598 - make LRU size a /proc tunable, clears locks when reduced (707)
5599 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5600 - server locks take references on exports now (1558)
5601 - build fixes for 2.4.20-rh trees (1663)
5602 - return an error from lov_create if all OSCs are inactive (1751)
5603 - fix import levels when a reconnect happens without a timeout (1597)
5604 - exit early from mds_open if we get a lookup error (1749)
5605 - partial page read at EOF wouldn't wait for disk before sending (1642)
5606 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5607 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5608 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5609 - write/truncate lock inversion (1639)
5610 - Don't auto-load obdclass, portals modules during cleanup (1495)
5611 - fix timestamps from jumping to "now" (1763)
5612 - extra journal assertions (1648)
5613 - add an extra multiunlink test (1771)
5614 - fix read_record/write_record API (1776)
5615 - fix leak of offset_extent, possible incorrect i_size later (1772)
5616 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5617 - Debugging for the kqswnal_get_idle_tx problems (1820)
5618 - Allow recovery to be attempted multiple times (1536)
5619 - Write out MDS last_rcvd file after it is first created (1600)
5620 - Fix tx_descriptor leak in failed transmit situations (1827)
5621 - ext3 journaling fixes for assertion failure after IO error (1871)
5622 - class_export_put() on freed export after completion AST error (1896)
5623 - Fix revalidate looping in VFS (1322)
5624 - Don't access a freed export during MDS_REINT timeout (1521)
5625 - Add open-unlink recovery support on the MDS (1673,1764)
5626 - Return an error if no MDS data was read from last_rcvd (1946)
5627 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5628 - Don't setup a disk device that doesn't match exported UUID (317)
5629 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5630 - avoid committing NULL handle in force close
5631 - local.sh is now a one-stripe LOV configuration
5632 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5633 - fix ext3 htree duplicate directory entry corruption (1516)
5634 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5635 - update to new LOV EA format (2097)
5636 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5638 2003-06-15 Phil Schwan <phil@clusterfs.com>
5641 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5642 - per-import/export recovery handling (958, 931, 959)
5643 - multiple last-rcvd slots, for serving multiple FSes (949)
5644 - connections are again shared between multiple imp/exports (963, 964)
5645 - "umount -f" would hang if any requests needed to be sent (393, 978)
5646 - avoid pinning large req buffer by copying for queued messages (989)
5647 - add "uuid" to "lctl device" command to help upcalls (991)
5648 - "open" RPCs with transnos would confuse recovery counters (1037)
5649 - do proper endian conversion of all wire messages (288, 340, 891)
5650 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5651 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5652 - fix ptlrpc_connection leak in target_handle_connect (1174)
5653 - fix import refcounting bug in OST and MDS cleanup (1134)
5654 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5655 - fix very unlikely obd_types race condition (501)
5656 - remove osc_open hack for echo_client (1187)
5657 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5658 - a failure in read_inode2 leads to deadlock (1139)
5659 - cancel ack-locks as soon as transaction is committed (1072)
5660 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5661 - make sure to commitrw after any preprw to avoid deadlock (1162)
5662 - failing to execute a file in a lustre FS would lock inode (1203)
5663 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5664 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5665 - obd_uuid2tgt would walk off the end of the list (1255)
5666 - on IA64 the osc would give portals incorrect bulk size (1258)
5667 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5668 - fix lock inversion caused by new llite matching code (1282)
5669 - limit the number of dirty pages on a client to 10MB (1286)
5670 - timed out locks were not being corrected cancelled (1289)
5671 - fix O_DIRECT above 4GB on IA-32 (1292)
5672 * major user-visible changes
5673 - fail out/fail over policy now controlled by the upcall (993)
5675 - add OBD_PING to check server availability and failure (954)
5676 - lustre messages are now sent in sending host order (288, 340, 891)
5677 - add eadatalen to MDS getattr reply (340)
5678 - OST read replies may contain second buffer, with per-page status (593)
5680 2003-03-11 Phil Schwan <phil@clusterfs.com>
5683 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5684 - failed open()s could cause deadlock; fixed (867, 869)
5685 - stop cancelling OST locks when files are closed (481)
5686 - overlapping XID spaces caused network corruption (851, 853)
5687 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5688 - setattr_raw added, to do single-RPC, server-side setattrs
5689 - lmc/lconf syntax change for OST UUIDs
5690 - fix crashy race condition between ptlrpc_free_req and osc_close
5691 - don't use request in mdc_enqueue if we hit a timeout (889)
5692 - don't set the inode i_size for regular files from the MDS (896)
5693 - handle out of order completion AST (842)
5694 - don't LBUG if a lock request times out after receiving AST (913)
5695 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5696 - if a bad lock AST arrives, send an error instead of dropping entirely
5697 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5698 - fix leak in bulk IO when only partially completed (899, 900, 926)
5699 - fix O_DIRECT for ia64 (55)
5700 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5701 - C-z after timeout could hang a process forever; fixed (977)
5703 - client-side I/O cache (678, 924, 929, 941, 970)
5705 - READPAGE and SETATTRs which don't take server-side locks get
5708 2003-02-11 Phil Schwan <phil@clusterfs.com>
5711 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5712 - Fix fsfilt_extN_readpage() to read a full page of directory
5713 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5714 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5715 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5716 - Most intent operations were removed, in favour of a new RPC mode
5717 that does a single RPC to the server and bypasses most of the VFS
5718 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5719 - Aggressively cancel local locks on DLM servers
5720 - mds_reint_unlink sends EA to the client if it's the last nlink.
5721 client uses that EA to unlink OST objects.
5722 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5723 - recursive symlinks were fixed (439)
5724 - fixed NULL deref in DEBUG_REQ
5725 - filter_update_lastobjid no longer calls sync, which annoyed extN
5726 - fixed multi-client small-writes to a single file problem (445)
5727 - fixed mtime updates during file writes (607)
5728 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5729 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5730 - fixed Linux OST size reporting problem (444, 656)
5731 - OST now updates object mtime with writes or setattr (607, 619)
5732 - client verifies file size before zeroing page past EOF (445)
5733 - OST now writes last allocated objid to disk with allocation (108)
5734 - LOV on echo now works (409)
5736 - mds_reint_unlink sends a new buffer, with the EA included. this
5737 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5738 set if a regular file was being unlinked, and it was the last link
5739 - use PtlGet from the target for bulk writes (315)
5740 - OST now updates object mtime with writes or setattr (607, 619)
5741 - LDLM now has a grant-time callback to revalidate locked items, if
5743 - Many MDS operations were reorganized to combat race conditions
5745 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5746 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5748 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5751 - Fully reactivate OST imports after reconnection (512, others)
5752 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5753 - More graceful error handling for truncating on dead OST (515)
5754 - Don't error out unless we're actually accessing dead stripes (474)
5755 - Fix garbage sizes when stripes are missing (410)
5756 - LRU counters were broken, causing constant lock purge (433, 432)
5757 - garbage on read from stripes with failed OSTs (441)
5758 - mark OSCs as active before reconnecting during recovery (438)
5759 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5760 - lfind did not preserve OST order in output (443)
5761 - symlinks cause hung clients, incorrect data (439)
5762 - stop dereferencing request after dropping refcount (457)
5763 - don't LASSERT(spin_is_locked) on non-SMP (455)
5764 - fixes for many rename() bugs
5765 - fstat didn't correctly synchronize attributes (399)
5766 - server must handle lock cancellation during blocking AST prep (487)
5767 - bulk descriptors were free()d too soon (511)
5768 - fix paths in lconf, which would load incorrect modules (451, 507)
5769 - fix confusing lconf 'host not found' error message (386)
5770 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5771 - fix race condition in mdc_blocking_ast() for inode access (526)
5772 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5773 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5774 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5775 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5776 - fix lstripe in several ways (kernel side) (527)
5777 - fix request leak in ldlm_cli_enqueue (262)
5778 - incorrect OSC was marked inactive after OST failure
5779 - call mds_fs_cleanup before unmounting filesystem (524)
5780 - fix races between taking ns_lock and ldlm_lock_change_resource
5781 - fix races updating LOV export open file list
5782 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5783 - fix recovery NULL deref in ldlm_cli_cancel_unused
5784 - fix some DLM races by using new hash table for lock handles (419)
5785 - permit the client to specify desired inodes, at replay
5786 - duplicate requests when we queue them for replay reintegration
5787 - fix last_rcvd offset calculation
5788 - sync after each recovered transaction, so we always make progress
5789 - never, not always, ERESTART requests without transnos
5790 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5792 - skip replay if the MDS says that the client is already connected
5793 - don't check for a recovery-enabled export to match lctl's UUID
5794 - don't INC_USE_COUNT for phantom exports
5795 - don't crash when cleaning up phantom exports (567)
5796 - don't double-finish or set replay data for errored mdc_open requests
5797 - abort requests when they time out, so we don't get old replies
5798 - send/receive replies for AST messages again
5799 - if the client says that it doesn't have the lock, cancel it on the
5801 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5802 mark it as destroyed, it will all work out when decref is called
5803 - fix module use counts (22, 581)
5805 - ASTs now expect a reply (server cancels lock on error reply)
5807 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5810 - fix many simultaneous client startup (392)
5811 - fix dentry->d_it clobbering
5812 - credentials weren't being shipped for readdir/getattr operations
5813 - remove invalid assertions triggered during some concurrent MD
5815 - proper Lustre versions added (336, 389)
5816 - fix memory leak for create error case (398)
5817 - fix LOV locking bug that would get cli/srv out of sync
5818 - fix echo client over LOV (409)
5819 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5820 - fix double-O_EXCL intent crash (424)
5821 - avoid sending multiple lock CANCELs (352)
5823 - MDS can do multi-client recovery (modulo bugs in new code)
5825 - many updates, edits, cleanups
5827 2002-11-18 Phil Schwan <phil@clusterfs.com>
5830 - fix null d_it dereference (346)
5831 - fix full OST/dbench hang (333)
5832 - fix permission problem with file removal (286)
5833 - fix removal of OSCs from LOV when they fail
5834 - fix NULL deref during bulk timeout (214)
5835 - fix problems related to multiple filesystems on one MDS (241)
5836 - fixed serious subtle metadata locking bugs
5837 - free locks on clients when inodes are removed due to memory
5839 - fix inode pointer in lock data (285)
5840 - partial support for multiple MDS on a single host (241)
5841 - data locks weren't cancelled at clear_inode time (290, 311)
5842 - intent locks could lead to unbounded lock growth (205)
5843 - added a maximum lock count, an LRU list, and a flusher
5844 - fix multiple rename (365)
5845 - properly abstracted the echo client
5846 - OSC locked 1 byte too many; fixed
5847 - rewrote brw callback code:
5848 - fixed recovery bugs related to LOVs (306)
5849 - fixed too-many-pages-in-one-write crash (191)
5850 - fixed (again) crash in sync_io_timeout (214)
5851 - probably fixed callback-related race (385)
5853 - Add capability to MDS protocol
5854 - LDLM cancellations and callbacks on different portals
5856 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5859 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5860 - defer open object destruction to close time (601981, 138)
5861 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5862 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5863 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5864 - don't call obd_setattr in ll_file_release for destroyed objects
5865 * protocol change to lustre_msg: move |version| and add |flags|
5866 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5867 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5868 * added replay of create, unlink, link and rename operations during
5869 MDS failover; recovery should be much more robust now
5870 * remove failed OSCs from LOVs (only lov_create uses this so far)
5871 * the lustre-HOWTO was brought (more) up to date (582544)
5873 2002-10-23 Phil Schwan <phil@clusterfs.com>
5876 - in-use dentries weren't being reused properly (617851)
5877 - prevent multiple LDLM setup (599178)
5878 - fix LOV size calculations for truncate (617853)
5879 - fix client handling of MDS intent errors (POSIX)
5880 - fix permission bug in lovstripe.c test (624321)
5881 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5882 - truncate past end of file could corrupt data
5883 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5884 - a race in recovery could return ETIMEDOUT to apps (623947)
5885 - building outside the source directory was fixed
5886 * the lustre-HOWTO was brought (more) up to date (582544)
5887 * major progress was made on recovery functionality
5889 2002-10-10 Phil Schwan <phil@clusterfs.com>
5892 - recovery deadlock fix
5893 - rm -rf causes LBUG fix (617817)
5894 - file open by multiple tasks fix (618962)
5895 - directory permissions bugs (602707 and 620007)
5896 - journal_stop fixed with locking (611313)
5897 - O_APPEND failures resolved (618273, perhaps 614459)
5898 - lconf PATH fix (619770)
5899 - IA64 build fix (621450)
5900 - RPC buffer sizes scale with amount of memory
5902 2002-10-01 Phil Schwan <phil@clusterfs.com>
5905 - locks would be cancelled without throwing away data pages,
5906 resulting in inconsistent data (605627)
5907 - inode attributes were not always being refreshed (605627, 612449)
5908 - lconf now continues to cleanup after lctl reports an error
5909 - MDS now enforces user permissions (602707)
5910 - lprocfs cleanup fixed, but not yet enabled (614157)
5911 - fixed infinite server hang, should a client not respond to an AST
5912 - avoid going into recovery if user calls readlink() with a buffer
5913 that's too small (613941)
5914 - AST RPCs no longer require replies (614867) -- this may be changed
5915 - don't crash server if client sends an IOV that's too big (611336)
5916 - fixed lock conversion deadlock (611892)
5917 - fixed the following of symlinks (614622)
5918 * recovery: the server can remove locks from a client that dies, other
5919 clients can make progress
5920 * more extN patch fixes
5921 * compile-time configurable ptlrpc buffer allocations
5923 - collaborative read cache document
5924 - Lustre Lite Performance CDR document-in-progress
5926 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5929 - fix typo in patch-2.4.18
5931 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5934 - clear ptlrpc request each time in handle_incoming_request()
5935 - unlink of files now destroys the object on the OST
5937 2002-09-19 Peter Braam <braam@clusterfs.com>
5939 * add hard link support
5940 * change obdfile creation method
5941 * kernel patch changed
5943 2002-09-19 Peter Braam <braam@clusterfs.com>
5946 - stack overflow bug in extN fixed
5948 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5950 * documentation updates
5951 - add man pages for config tools
5952 - update tests/README to describe testing with new config tools
5953 - finish metadata API descriptions
5954 * bug fixes and cleanups
5955 - statfs workaround for 16TB limit
5956 - LOV stripe allocation improved, can stripe on subset of OSTs
5957 - LOV file size/IO offset was wrong for files > 4GB in size
5958 - object EA data was being dropped, caused files to be unreadable
5959 - memory overflow with non-LOV OST caused memory corruption
5960 - fixed regression tests to work with new config tools, obdfilter
5961 - fixed bug when directory size became larger than 1 block
5962 - fixed bug (for single client case) when PWD was deleted
5963 - invalidate local directory pages when doing intent-based ops
5964 - avoid LDLM oops when lock callback contained bad data
5966 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5968 * documentation updates
5969 * bug fixes and cleanups
5970 - configuration tools
5973 - 64-bit compile warnings
5974 - 64-bit internal statfs data
5976 * test_brw on persistent OST devices
5978 * lprocfs (disabled)
5980 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5982 * documentation updates
5983 * bug fixes and cleanups
5984 * configuration tools
5986 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5989 * many small fixes to 0_5_4
5990 * io/network handling
5991 * thinkos in MDS operations
5993 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5996 * crucial basic fixes to 0.5.3
5997 * IOR, Iozone work over Elan
6000 2002-08-07 Phil Schwan <phil@clusterfs.com>
6001 * version 0_5_3, our first alpha
6002 * we use the new Portals iovs
6003 * documentation updates
6004 * bug fixes and cleanups
6005 * small changes in the DLM wire protocol
6007 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6008 * version 0_5_1 with some initial stability,
6009 * locking on MD and file I/O.
6010 * documentation updates
6011 * several bug fixes since 0.5.0
6012 * small changes in wire protocol
6014 2002-07-18 Phil Schwan <phil@clusterfs.com>
6016 * delivered as Lustre Light Alpha
6017 * fixed a crash after handling invalid MDS requests
6018 * fixed directory pages for architectures with non-4k pages sizes
6020 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6021 * release version v0_4_4
6022 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6024 * Fixes a number of interruption problems with OST operations.
6025 * Update documentation for portals header changes
6026 * Move all wire protocol structs/defines to lustre_idl.h
6027 * Fixes symlink length bug.
6028 * Add tcpdump to repository.
6030 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6031 * release version v0_4_3
6032 * Fixes statfs for inodes on extN.
6033 * Fixes bug in runtests which would delete /etc/hosts.
6034 * Use 64-bit object IDs wherever possible (not into VFS though)
6035 Remove ost_get_info, which is unused by lustre, and out of date.
6037 2002-07-03 Peter Braam <braam@clusterfs.com>
6038 * release version v0_4_2 Fixes a lookup error (type not passed)
6039 * move forward to head of Portals
6040 * move forward to latest Lustre kernel
6042 2002-06-25 Peter Braam <braam@clusterfs.com>
6043 * release version v0_4_1. Hopefully stable on single node use.