1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-164.6.1.el5 (RHEL 5),
6 2.6.18-164.6.1.0.1.el5 (OEL 5),
8 2.6.27 vanilla (kernel.org).
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.27 vanilla (kernel.org)
12 * Recommended e2fsprogs version: 1.41.6.sun1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
14 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
15 removed cwd "./" (refer to Bugzilla 14399).
17 Severity : enhancement
19 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
20 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
24 Description: Sanity test_27p fails.
25 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
26 Sanity tests 27 were cleaned up and improved.
30 Description: Unify req format on client/servers
31 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
32 buffers and swabbers approach..
37 Description: drop join file stuff.
39 Severity : enhancement
41 Description: Add 2.6.27 and SLES11 (patchless client) support
43 Severity : enhancement
45 Description: Adjust locks' extents on their first enqueue, so that at the time
46 they get granted, there is no need for another pass through the
47 queues since they are already shaped into the proper forms.
51 Description: Fix in ptlrpc_expire_one_request() to print the signed time
54 Severity : enhancement
56 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
59 Severity : enhancement
61 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
63 Severity : enhancement
65 Description: correctly handle big reply message.
66 Details : send LNet event if reply is bigger then buffer and adjust this buffer
71 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
73 Severity : enhancement
75 Description: Add support for OEL5.
79 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
83 Description: client could not reconnect to OST because of an active request.
84 Details : The client now retries to reconnect to the same server, if a
85 connect request failed with EBUSY or -EAGAIN. Test case 35b
86 conf-sanity.sh has been created.
90 Description: Don't allow make backward step on assignin osc next id.
91 Details : race between allocation next id and ll_sync thread can be cause
92 of set wrong osc next id and can be kill valid ost objects.
96 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
100 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
104 Description: Error handling in osc_statfs_interpret() has been improved.
105 Details : Check in osc_statfs_interpret() for EBADR.
109 Description: Do not send statfs() requests to OSTs disabled by administrator.
110 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
112 Severity : enhancement
114 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
116 Severity : enhancement
118 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
120 Severity : enhancement
122 Description: Enhancement for lfs(1) command to use numeric uid/gid.
126 Description: Improved cleanup procedure for conf-sanity test_46a.
130 Description: Do not start bulk transfer, if client tried to reconnect.
131 Details : Do not start bulk transfer, if client tried to reconnect,
132 but failed due to active rpcs. Check for exp_abort_active_req
133 has been implemented.
137 Description: Prevent inconsistences between linux and lustre mount structures.
138 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
139 Make the sleep interruptible.
141 Severity : enhancement
143 Description: provide server to client comms path
144 Details : server to client communications path via new LDLM_SET_INFO rpc
148 Description: truncate starts GFP_FS allocation under transaction causing deadlock
149 Details : ldiskfs_truncate calls grab_cache_page which may start page
150 allocation under an open transaction. This may lead to
151 calling prune_icache with consequent lustre reentrance.
155 Description: send correctly lsm on open replay
156 Details : MDS is trust to LSM size on replay open, but client can set wrong size
161 Description: enable client interface failover
162 Details : When a child reconnects from another NID, properly update export
163 nid hash position and ldlm reverse import.
165 Severity : enhancement
167 Description: Communicate OST degraded/readonly state via statfs to MDS
168 Details : Flags in the statfs returned from OSTs indicate whether the
169 OST is in a degraded RAID state, or if the filesystem has
170 turned read-only after a filesystem error is detected.
175 Description: don't panic if EPROTO was hit when reading symlink
176 Details : correctly handling request reference in error cases.
178 Severity : enhancement
180 Description: Add LustreNetLink, a kernel-userspace communcation path.
182 Severity : enhancement
184 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
189 Description: access to llog context before init.
190 Details : move handling CATALOGS file at osc layer and forbid access to llog
194 Frequency : rare, only when client is writing to OST on local node
196 Description: Avoid deadlock for local client writes
197 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
198 memory freeing context. This allows OST threads to set the
199 PF_MEMALLOC flag on task structures in order to allocate memory
200 from reserved pools and complete IO.
201 Use GFP_HIGHUSER for OST allocations for non-local client writes,
202 so that the OST threads generate memory pressure and allow
203 inactive pages to be reclaimed.
205 Severity : enhancement
207 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
212 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
213 Details : move ldlm namespace creation in setup phase to avoid grab
214 _lprocfs_lock with cli_sem held.
218 Description: Temporarily disable grant shrink.
219 Details : Disable the feature for debugging.
222 Frequency : only when formatting test filesystems
224 Description: Unable to run several mkfs.lustre on loop devices at the same time
225 Details : mkfs.lustre returns error 256 on the concurrent loop devices
226 formatting. The solution is to proper handle the error.
228 Severity : enhancement
230 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
232 Severity : enhancement
234 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
239 Description: resolve race between obd_disconnect and class_disconnect_exports
240 Details : if obd_disconnect will be called to already disconnected export he
241 forget release one reference and osc module can't unloaded.
243 Severity : enhancement
245 Description: Allow tuning service thread via /proc
246 Details : For each service a new
247 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
248 created that can be used to set min/max thread counts, and get the
249 current number of running threads.
254 Descriptoin: don't return error if have partially created objects for file.
255 Details : lov_update_create_set uses set->set_success as index for created
256 objects, so if some requests will be failed, they will have hole
257 at end of array and qos_shrink_lsm for allocate correct lsm.
259 Severity : enhancement
261 Description: Update OFED support to 1.4.1
266 Description: Reduce small size read RPC
267 Details : Set read-ahead limit for every file and only do read-ahead when
268 available read-ahead pages are bigger than 1M to avoid small size
271 Severity : enhancement
273 Description: move AT tunable parameters for more consistent usage
274 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
276 Severity : enhancement
278 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
279 Details : allow skip disconnected ost for send statfs request and hide error
285 Description: fix racy locking of mballoc block bitmaps causing BUG
286 Details : The locking of the mballoc buddy bitmap and the in-memory
287 block bitmap was using two different spin locks in some
288 cases. This made it possible to incorrectly access the
289 mballoc bitmap while another process was modifying it,
290 causing a sanity assertion to fail. While no on-disk corruption
291 was reported, there was some risk of this happening.
294 Frequency : rare, on llog test 6
296 Descriptoin: don't allow connect to already connected import
297 Details : allowing connect to already connected import is hide connecting problem.
302 Descriptoin: allow kill process which wait statahead result
303 Details : for some reasons 'ls' can stick in waiting result from statahead,
304 in this case need way for kill this process.
306 Severity : enhancement
308 Description: Add state history info file, enhance import info file
309 Details : Track import connection state changes in a new osc/mdc proc file;
310 add overview-type data to the osc/mdc import proc file.
312 Severity : enhancement
314 Description: MDS create should not wait for statfs RPC while holding DLM lock.
316 Severity : enhancement
318 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
323 Descriptoin: OST grant too much space to client even there are not enough space.
324 Details : Client will shrink its grant cache to OST if there are no write
325 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
326 this grant cache if there are already not enough avaible space
327 (left_space < total_clients * 32M).
330 Frequency : start MDS on uncleanly shutdowned MDS device
332 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
333 Details : stay in waiting mds<>ost recovery finished produce random bugs
334 due race between two ll_sync thread for one lov target. send
335 ACTIVATE event only if connect realy finished and import have
339 Frequency : rare, connect and disconnect target at same time
341 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
342 Details : don't call obd_disconnect under lov_lock. this long time
343 operation and can block ptlrpcd which answer to connect request.
348 Descriptoin: don't lose wakeup for imp_recovery_waitq
349 Details : recover_import_no_retry or invalidate_import and import_close can
350 both sleep on imp_recovery_waitq, but we was send only one wakeup
354 Frequency : always with long access acl
356 Descriptoin: mds can't pack reply with long acl.
357 Details : mds don't control size of acl but they limited by reint/getattr
360 Severity : enhancement
362 Description: Update to SLES10 kernel-2.6.16.60-0.33.
364 Severity : enhancement
366 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
369 Frequency : start MDS on uncleanly shutdowned MDS device
371 Descriptoin: aborting recovery hang on MDS
372 Details : don't throttle destroy RPCs for the MDT.
375 Frequency : on remount
377 Description: external journal device not working after the remount
378 Details : clear dev_rdonly flag for external journal devices in
384 Description: shutdown vs evict race
385 Details : client_disconnect_export vs connect request race.
386 if client will evicted at this time - we start invalidate
387 thread without referece to import and import can be freed
391 Frequency : rare, need acl's on inode.
393 Description: client can't handle ost additional correctly
394 Details : if ost was added after client connected to mds client can have
395 hit lnet_try_match_md ... to big messages to wide striped files.
396 in this case need teach client to handle config events about add
397 lov target and update client max ea size at that event.
399 Severity : enhancement
401 Description: Changelogs
402 Details : Changelogs are a lightweight mechanism to track filesystem
403 metadata and namespace changes. The changelog is recorded
404 permanently on the MDTs, and is periodically "consumed" / purged
405 when records are no longer needed.
407 Severity : enhancement
409 Description: compact fld format with extents
410 Details : Store range of seq rather than every seq in FLD. Seq
411 controller update FLD rather than clients. In Case of CMD, mdt0
412 has FLD, all other metadata server act as non persistent proxy
413 for FLD queries and cache fld entries in fld cache.
418 Description: don't skip ost target if they assigned to file
419 Details : Drop slow OSCs if we can, but not for requested start idx.
420 This means "if OSC is slow and it is not the requested
421 start OST, then it can be skipped, otherwise skip it only
422 if it is inactive/recovering/out-of-space.
426 Description: more cleanup in mds_lov
427 Details : not send LOV EA under replay, we can't know about they size at this
428 time. Don't allow client connect to mds before any ost connected,
429 for avoid problems with LOV EA size and returning EIO to client.
431 Severity : enhancement
433 Description: Interoperability at server side (Disk interoperability)
435 Severity : enhancement
437 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
439 Severity : enhancement
441 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
443 Severity : enhancement
445 Description: New client IO stack (CLIO).
447 Severity : enhancement
449 Description: Commit on sharing. Eliminate inter-client dependencies between
450 uncommitted transactions by doing transaction commits.
451 Thereby clients may recovery independently.
454 Frequency : Create a symlink file with a very long name
456 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
457 Details : If there is no extra space in the request for early cancels,
458 ldlm_req_handles_avail() returns 0 instead of a negative value.
460 Severity : enhancement
462 Description: Add /proc entry for import status
463 Details : The mdc, osc, and mgc import directories now have
464 an import directory that contains useful import data for debugging
467 Severity : enhancement
469 Description: Re-disable certain /proc logging
470 Details : Enable and disable client's offset_stats, extents_stats and
471 extents_stats_per_process stats logging on the fly.
474 Frequency : Only on FC kernels 2.6.22+
476 Description: oops in statahead
477 Details : Do not drop reference count for the dentry from VFS when lookup,
478 VFS will do that by itself.
480 Severity : enhancement
482 Description: Generic /proc file permissions
483 Details : Set /Proc file permissions in a more generic way to enable non-
484 root users operate on some /proc files.
488 Description: Hitting mdc_commit_close() ASSERTION
489 Details : Properly handle request reference release in
490 ll_release_openhandle().
494 Description: quota recovery deadlock during mds failover
495 Details : This patch includes att18982, att18236, att18237 in bz14840.
497 1. fix osts hang when mds does failover with quotaon
498 2. prevent watchdog storm when osts threads wait for the
503 Frequency : only patchless client
504 Description: add workaround for race between add/remove dentry from hash
506 Severity : enhancement
508 Description: Allow OST glimpses to return PW locks
512 Description: LBUG when llog conf file is full
513 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
517 Description: Prevent import from entering FULL state when server in recovery
521 Description: service mount cannot take device name with ":"
522 Details : Only when device name contains ":/" will mount treat it as
528 Description: replace ptlrpcd with the statahead thread to interpret the async
529 statahead RPC callback
533 Frequency : on recovery
534 Description: I/O failures after umount during fail back
535 Details : if client reconnected to restarted server we need join to recovery
536 instead of find server handler is changed and process self
537 eviction with cancel all locks.
539 Severity : enhancement
541 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
543 Severity : enhancement
545 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
547 Severity : enhancement
549 Description: Upcall on Lustre log has been dumped
550 Details : Allow for a user mode script to be called once a Lustre log has
551 been dumped. It passes the filename of the dumped log to the
552 script, the location of the script can be specified via
553 /proc/sys/lnet/debug_log_upcall.
558 Description: avoid idr_remove called for id which is not allocated.
559 Details : Move assigment s_dev for clustered nfs to end of initialization,
560 to avoid problem with error handling.
565 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
566 Details : When connection is reused this not moved from CONN_UNUSED_HASH
567 into CONN_USED_HASH and this prodice warning when put connection
568 again in unused hash.
571 Severity : enhancement
573 Description: File striping can now be set to use an arbitrary pool of OSTs.
575 Severity : enhancement
577 Description: Export bytes_read/bytes_write count on OSC/OST.
581 Description: Early reply size mismatch, MGC loses connection
582 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
583 the connect flags are properly negotiated.
588 Description: quotas are not honored with O_DIRECT
589 Details : all writes with the flag O_DIRECT will use grants which leads to
590 this problem. Now using OBD_BRW_SYNC to guard this.
594 Description: add quota statistics
595 Details : 1. sort out quota proc entries and proc code.
596 2. add quota statistics
598 Severity : enhancement
600 Description: enable quota support for HEAD.
604 Description: Properly propagate oinfo flags from lov to osc for statfs
605 Details : restore missing copy oi_flags to lov requests.
607 Severity : enhancement
609 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
610 : l_getgroups(8), lst(8), routerstat(8)
611 Details : included man pages for llobdstat(8), llstat(8),
612 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
614 Severity : enhancement
616 Description: configure's --enable-quota should check the
617 : kernel .config for CONFIG_QUOTA
618 Details : configure is terminated if --enable-quota is passed but
619 : no quota support is in kernel
623 Description: Remove portals compatibility
624 Details : Remove portals compatibility, not interoperable with releases
629 Description: Resolve device initialization race
630 Details : Prevent proc handler from accessing devices added to the
631 obd_devs array but yet be intialized.
633 Severity : enhancement
635 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
637 Severity : enhancement
639 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
644 Description: Using wrong pointer in osc_brw_prep_request
645 Details : Access to array[-1] can produce panic if kernel compiled with
646 CONFIG_PAGE_ALLOC enabled
650 Description: Client runs out of low memory
651 Details : Consider only lowmem when counting initial number of llap pages
655 Description: *optional* service tags registration
656 Details : if the "service tags" package is installed on a Lustre node
657 When the filesystem is mounted, a local-node service tag will
658 be created. See http://inventory.sun.com/ for more information
659 about the Service Tags asset management system.
663 Description: Kernel BUG tries to release flock
664 Details : Lustre does not destroy flock lock before last reference goes
665 away. So always drop flock locks when client is evicted and
666 perform unlock regardless of successfulness of speaking to MDS.
670 Description: add refcount protection for osc callbacks, avoid panic on shutdown
674 Description: sanity test 65a fails if stripecount of -1 is set
675 Details : handle -1 striping on filesystem in ll_dirstripe_verify
680 Description: ASSERTION(CheckWriteback(page,cmd)) failed
681 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
684 Severity : enhancement
686 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
690 Description: do not process already freed flock
691 Details : flock can possibly be freed by another thread before it reaches
692 to ldlm_flock_completion_ast.
696 Description: LBUG during stress test
697 Details : Need properly lock accesses the flock deadlock detection list.
701 Description: oops in page fault handler
702 Details : kernel page fault handler can return two special 'pages' in error
703 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
707 Description: timeout with invalidate import.
708 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
709 should be handled by ptlrpcd. This produce long age waiting and
710 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
712 Severity : enhancement
714 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
718 Description: enable MGS and MDT services start separately
719 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
720 a co-located MGS without starting the MGS, which is a complement
721 to 'nosvc' mount option.
725 Frequency : after recovery
726 Description: precreate to many object's after del orphan.
727 Details : del orphan st in oscc last_id == next_id and this triger growing
728 count of precreated objects. Set flag LOW to skip increase count
729 of precreated objects.
733 Frequency : rare, on clear nid stats
734 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
735 Details : when clean nid stats sometimes try destroy live entry,
736 and this produce panic in free.
740 Description: Stack overflow during MDS log replay
741 ease stack pressure by using a thread dealing llog_process.
745 Description: wait until IO finished before start new when do lock cancel.
746 Details : VM protocol want old IO finished before start new, in this case
747 need wait until PG_writeback is cleared until check dirty flag and
748 call writepages in lock cancel callback.
750 Severity : enhancement
752 Description: using special macro for print time and cleanup in includes.
756 Description: mds_mfd_close() ASSERTION(rc == 0)
757 Details : In mds_mfd_close(), we need protect inode's writecount change
758 within its orphan write semaphore to prevent possible races.
762 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
766 Frequency : rare, on shutdown ost
767 Description: don't hit live lock with umount ost.
768 Details : shrink_dcache_parent can be in long loop with destroy dentries,
769 use shrink_dcache_sb instead.
773 Description: don't panic with use echo client
774 Details : echo client pass NULL as client nid pointer and this produce null
779 Description: fix build on ppc32
780 Details : compile code with -m64 flag produce wrong object file for ppc32.
784 Description: add message levels for liblustreapi
788 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
789 Details : Move the 'good_osts' check before the 'total_bavail' check. This
790 will result in an -EAGAIN and in the exit call path we call
791 alloc_rr() which will with increasing aggressiveness attempt to
792 aquire precreated objects on the minimum number of required OSCs.
796 Description: Use old size assignment to avoid deadlock
797 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
798 the scheduling while holding a spinlock. We do not need locking
799 for size in ll_update_inode() because size is only updated from
800 the MDS for directories or files without objects, so there is no
801 other place to do the update, and concurrent access to such inodes
802 are protected by the inode lock.
806 Description: resolve "_IOWR redefined" build error on SLES10
810 Description: dump the memory debugging after all modules are unloaded to
811 suppress false negative in conf_sanity test 39
813 Severity : enhancement
815 Description: build kernel-ib packages for OFED 1.3 in our release cycle
820 Description: fix SLES kernel versioning
821 Details : the kernel version for our SLES 10 kernel did not include a "-"
822 before the "smp" at the end. while this was not a problem in
823 general, it did mean that software trying to use the kernel
824 version to try to detect a vendor specific kernel would fail.
825 this was most evident by the OFED build scripts.
829 Description: Don't update lov_desc members until making sure they are valid
830 Details : When updating lov_desc members via proc fs, need fix their
831 validities before doing the real update.
835 Description: don't put request into delay list while invalidate in flight.
836 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
837 import in flight. this produce timeout for invalidate and sometimes
838 can cause stale data.
842 Frequency : on ppc only
843 Description: not convert ost objects for directory because it's not exist.
844 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
848 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
850 Severity : enhancement
852 Description: Update to support 2.6.22.14 vanilla kernel.
856 Frequency : rare, on recovery
857 Description: read procfs can produce deadlock in some situation
858 Details : Holding lprocfs lock which send rpc can produce block for destroy
859 obd objects and this also block reconnect with -EALREADY. This
860 isn't fix all lprocfs bugs - but make it rare.
862 Severity : enhancement
864 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
867 Frequency : frequent on X2 node
869 Description: mdc_set_open_replay_data LBUG
870 Details : Set replay data for requests that are eligible for replay.
874 Description: lustre_mgs: operation 101 on unconnected MGS
875 Details : When MGC is disconnected from MGS long enough, MGS will evict the
876 MGC, and late on MGC cannot successfully connect to MGS and a lot
877 of the error messages complaining that MGS is not connected.
881 Frequency : on network error
882 Description: panic with double free request if network error
883 Details : mdc_finish_enqueue is finish request if any network error ocuring,
884 but it's true only for synchronus enqueue, for async enqueue
885 (via ptlrpcd) this incorrect and ptlrpcd want finish request
888 Severity : enhancement
890 Description: client-side metadata stat-ahead during readdir(directory readahead)
891 Details : perform client-side metadata stat-ahead when the client detects
892 readdir and sequential stat of dir entries therein
895 Frequency : on start mds
897 Description: Implement get_info(last_id) in obdfilter.
900 Frequency : occasional
902 Description: Correctly check stale fid, not start epoch if ost not support SOM
903 Details : open with flag O_CREATE need set old fid in op_fid3 because
904 op_fid2 was overwritten with new generated fid, but mds can answer
905 with one of these two fids and both is not stale. Setattr
906 incorrectly started an epoch and assume will be called
907 done_writing, but without SOM done_writing ever being called.
910 Frequency : rare, depends on device drivers and load
912 Description: MDS or OSS nodes crash due to stack overflow
913 Details : Code changes in 1.8.0 increased the stack usage of some functions.
914 In some cases, in conjunction with device drivers that use a lot
915 of stack the MDS (or possibly OSS) service threads could overflow
916 the stack. One change which was identified to consume additional
917 stack has been reworked to avoid the extra stack usage.
920 Frequency : occasional
922 Description: Do not fail import if osc_interpret_create gets -EAGAIN
923 Details : If osc_interpret_create got -EAGAIN it immediately exits and
924 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
925 oscc_has_objects and see OSC has no objests and call
926 oscc_internal_create to resend create request.
928 Severity : enhancement
930 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
932 Severity : enhancement
934 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
937 Frequency : very rare
939 Description: Fix replay if there is an un-replied request and open
940 Details : In some cases, older replay request will revert the
941 mcd->mcd_last_xid on MDS which is used to record the client's
944 Severity : enhancement
946 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
948 Severity : enhancement
950 Description: Add rhel5 support to HEAD.
952 Serverity : enhancement
954 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
959 Description: Don't allow skipping OSTs if index has been specified.
960 Details : Don't allow skipping OSTs if index has been specified, make
961 locking in internal create lots better.
965 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
966 Details : access to bitfield in structure is always rounded to long
967 and this produce problem with not atomic change any bit.
971 Description: Lustre make rpms failed.
972 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
973 builds Lustre rpms from tarball.
976 Frequency : rare on shutdown ost
978 Description: If llog cancel was not send before clean_exports phase, this can
979 produce deadlock in llog code.
980 Details : If llog thread has last reference to obd and call class_import_put
981 this produce deadlock because llog_cleanup_commit_master wait when
982 last llog_commit_thread exited, but this never success because was
983 called from llog_commit_thread.
987 Description: allow userland application know is lost one of stripes.
988 Details : fill lvb_blocks with error code on ost and return it to
989 application if error flag found.
993 Description: NULL lov_tgts causing MDS oops
994 Details : more safe checks for NULL lov_tgts for avoid oops.
996 Severity : enhancement
998 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1002 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1003 Details : If an OST is down the MDS will hang indefinitely in
1004 obd_statfs_rqset() waiting for the statfs data. While for
1005 MDS QOS usage of statfs, it should not stuck in waiting.
1007 Severity : enhancement
1009 Description: remote_acl support
1010 Details : Support ACL-based permission check for remote user.
1011 Support setfacl/getfacl for remote user with the utils
1012 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1013 the system "{s,g}etfacl" utils.
1015 Severity : enhancement
1017 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1019 Severity : enhancement
1021 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1025 Description: make mgs_setparam() handle fsname containing dash
1026 Details : fsname containing a dash does not work with lctl conf_param
1028 Severity : enhancement
1030 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1032 Severity : enhancement
1034 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1038 Frequency : rare, at shutdown
1039 Description: access already free / zero obd_namespace.
1040 Details : if client_disconnect_export was called without force flag set,
1041 and exist connect request in flight, this can produce access to
1042 NULL pointer (or already free pointer) when connect_interpret
1043 store ocd flags in obd_namespace.
1047 Frequency : only at startup
1048 Description: not alloc memory with spinlock held.
1049 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1050 if any spinlock held.
1052 Severity : enhancement
1054 Description: make lustre randomly fail allocating memory
1055 Details : Make lustre randomly failed allocating memory for testing purpose.
1057 Severity : enhancement
1059 Description: lost problems with lov objid file
1060 Details : Fixes some scability and access to not inited memory problems
1061 in work with lov objdid file.
1066 Description: lfs find does not continue on file error
1067 Details : Continue other files processing when a file/dir is absent.
1071 Description: Inconsistent usage of lustre_pack_reply()
1072 Details : Standardize the usage of lustre_pack_reply() such that it
1073 always generate a CERROR on failure.
1076 Frequency : occasional
1078 Description: MDS hang or stay in waiting lock
1079 Details : If client receive lock with CBPENDING flag ldlm need send lock
1080 cancel as separate rpc, to avoid situation when cancel request
1081 can't processed due all i/o threads stay in wait lock.
1085 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1089 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1093 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1097 Description: Update to RHEL4 latest kernel.
1099 Severity : enhancement
1101 Description: Build SLES10 patchless client fails
1102 Details : The configure was broken by run ./configure with
1103 --with-linux-obj=.... argument for patchless client. When the
1104 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1105 can't search header adequately. Use absolute path such as
1106 -I($LINUX)/include instead.
1110 Description: interrupt oig_wait produce painc on resend.
1111 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1112 private set, and this produce situation when rq_ptlrpcd_data not
1113 copyed to new allocated request and triggered LBUG on assert
1114 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1115 ptlrpcd set if request is changed and can be safety changed to use
1120 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1121 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1122 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1123 these macros and replace them with their difinitions instead.
1127 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1128 Details : Update which_patch & target file for SLES10 latest kernel.
1131 Description: Remove client patches from SLES 10 kernel.
1132 Details : This causes SLES 10 clients to behave as patchless clients
1133 even on a Lustre-patched (server) kernel.
1135 Severity : enhancement
1137 Description: self-adjustable client's lru lists
1138 Details : use adaptive algorithm for managing client cached locks lru
1139 lists according to current server load, other client's work
1140 pattern, memory activities, etc. Both, server and client
1141 side namespaces provide number of proc tunables for controlling
1144 Severity : enhancement
1146 Description: light-weight GSS support
1147 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1148 and reduce performance overhead.
1150 Severity : enhancement
1152 Description: Linux keyring support
1153 Details : Support using service of Linux keyring for Lustre GSS internal
1154 context refresh/cache mechanism.
1158 Description: Fix errors in lfs documentation
1159 Details : Fixes man pages
1163 Description: don't use GFP_* in generic Lustre code.
1164 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1168 Description: obdclass is limited by single OBD_ALLOC(idarray)
1169 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1173 Description: warnings with build patchless client with vanila 2.6.19 and up
1174 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1175 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1179 Frequency : Only for SLES
1180 Description: sanity_quota fail test_1
1181 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1182 file, which makes TSTID[2] inunique.
1186 Frequency : Only for SLES9
1187 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1188 Details : Update target/ChangeLog/which_patch .
1192 Frequency : Only for test_10 in sanity-quota.sh
1193 Description: a bug in quota test code
1194 Details : checking a test flag in a wrong place
1198 Frequency : Only for test_13 in sanity-quota.sh
1199 Description: a bug in quota test script
1200 Details : shouldn't check overfull allocated quota
1204 Description: jbd statistics
1205 Details : Port older jbd statistics patch for sles10
1207 Severity : enhancement
1209 Description: Add printing inode info into message about error in writepage.
1214 Description: reply_lock_interpret crash due to race with it and lock cancel.
1215 Details : Do not replay locks that are being cancelled. Do not reference
1216 locks by their address during replay, just by their handle.
1220 Frequency : When flocks are used.
1221 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1222 Details : Flock locks might destroy just granted lock if it could be merged
1223 with another existing flock, this is done in completion handler,
1224 so teach ldlm_cli_enquque_fini that this is a valid case for
1230 Description: Oops in read and write path when failing to allocate lock.
1231 Details : Check if lock allocation failed and return error back.
1235 Description: lstripe command fails for valid OST index
1236 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1237 instead of lov->desc.ld_active_tgt_count.
1241 Description: sanity.sh failed test 103
1242 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1247 Description: df doesn't work properly if diskfs blocksize != 4K
1248 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1252 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1253 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1255 Severity : enhancement
1257 Description: Adaptive timeouts
1258 Details : RPC timeouts adapt to changing server load and network
1259 conditions to reduce resend attempts and improve recovery time.
1263 Description: llapi_file_create() does not allow some changes
1264 Details : add llapi_file_open() that allows specifying the mode and
1265 open flags, and also returns an open file handle.
1269 Description: merge and cleanup kernel patches.
1270 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1274 Description: Add journal checksum support.(Kernel part)
1275 Details : The journal checksum feature adds two new flags i.e
1276 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1277 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1278 indicates that the commit block contains the checksum for
1279 the blocks described by the descriptor blocks. Now commit
1280 record can be sent to disk without waiting for descriptor
1281 blocks to be written to disk. This behavior is controlled
1282 using JBD2_FEATURE_ASYNC_COMMIT flag.
1286 Description: OSS needs mutliple precreate threads
1287 Details : Add ability to start more than one create thread per OSS.
1291 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1292 Details : execute lfs setstripe on client
1296 Description: mds_obd_create error creating tmp object
1297 Details : When the user sets quota on root, llog will be affected and can't
1298 create files and write files.
1301 Frequency : Always on ia64 patchless client, and possibly others.
1303 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1304 Details : This allows the patchless client to be loaded on architectures
1305 without this export.
1309 Description: RedHat Update kernel for RHEL5
1310 Details : Add the kernel config file for RHEL5.
1314 Description: RedHat Update kernel for RHEL5
1315 Details : Modify the kernel config file more closer RHEL5.
1319 Description: Build failure against Centos5 (RHEL5)
1320 Details : Define PAGE_SIZE when it isn't present.
1324 Description: test_7 on sanity-quota.sh can't be used on separate machines
1325 Details : add facet to handle it
1329 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1330 Details : with NFS, the anon dentry's parent was set to itself in
1331 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1332 corresponding dentry other than use rec->ur_name.
1334 Severity : enhancement
1336 Description: lfs setstripe enhancement
1337 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1341 Description: enable data checksumming by default
1342 Details : enable checksum by default, allow --disable-checksum
1343 configure option and "-o nochecksum" mount option. Checksums
1344 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1348 Description: lustre support for RHEL5
1349 Details : Add support for RHEL5.
1353 Description: Client eviction due to failover config
1354 Details : after a connection loss, the lustre client should attempt to
1355 reconnect to the last active server first before trying the
1356 other potential connections.
1360 Description: when mds and osts use different quota unit(32bit and 64bit),
1361 quota will be released repeatly.
1362 Details : void sending multiple quota reqs to mds, which will keep the
1363 status between the reqs.
1367 Description: osts not allocated evenly to files
1368 Details : change the condition to increase offset_idx
1371 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1373 Description: Data corruption for OSTs that are formatted larger than 2TB
1375 Details : When generating the bio request for lustre file writes the
1376 sector number would overflow a temporary variable before being
1377 used for the IO. The data reads correctly from Lustre (which
1378 will overflow in a similar manner) but other file data or
1379 filesystem metadata may be corrupted in some cases.
1383 Description: Tune the kernel for good SCSI performance.
1384 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1385 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1390 Description: rewrite ext2-derived code in obdclass/uuid.c
1391 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1392 from scratch preserving functionality.
1396 Description: "lfs find -obd UUID" prints directories
1397 Details : "lfs find -obd UUID" will return all directory names instead
1398 of just file names. It is incorrect because the directories
1399 do not reside on the OSTs.
1403 Description: use i_size_read and i_size_write in 2.6 port
1404 Details : replace inode->i_size access with i_size_read/write()
1407 Frequency : when removing large files
1409 Description: scheduling issue during removal of large Lustre files
1410 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1411 It causes scheduling issues when removing large files (17TB in the
1415 Frequency : only with liblustre clients on XT3
1417 Description: evictions taking too long
1418 Details : allow llrd to evict clients directly on OSTs
1421 Frequency : only on ppc
1423 Description: /proc/fs/lustre/devices broken on ppc
1424 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1425 We should make sure the type of 'index' is loff_t and then cast
1426 explicitly as needed below. Do not assign an explicitly cast
1430 Frequency : only for rhel5
1432 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1433 Details : Modify the target file & which_kernel.
1435 Severity : enhancement
1437 Description: omit set fsid for export NFS
1438 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1439 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1443 Frequency : Always, for kernels after 2.6.16
1444 Description: Fix warning idr_remove called for id=.. which is not allocated.
1445 Details : Last kernels save old s_dev before kill super and not allow
1446 to restore from callback - restore it before call kill_anon_super.
1450 Description: buffer overruns could theoretically occur
1451 Details : llapi_semantic_traverse() modifies the "path" argument by
1452 appending values to the end of the origin string, and a
1453 overrun may occur. Adding buffer overrun check in liblustreapi.
1457 Description: Fix error on 'ls .' at the top of the Lustre mount.
1458 Details : Don't revalidate dentry if it is a root dentry.
1462 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1463 Details : Modify vm-tunables-rhel4.patch.
1467 Description: Kernel config for 2.6.18-vanilla.
1468 Details : Modify targets/2.6-vanilla.target.in.
1469 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1470 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1471 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1472 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1476 Description: improve handling recoverable errors
1477 Details : if request processig with error which can be recoverable on server
1478 request should be resend, otherwise page released from cache and
1483 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1484 Details : Modify target file & which_patch.
1485 A flaw was found in the IA32 system call emulation provided
1486 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1487 value could be stored in the %RAX register, which could trigger an
1488 out-of-bounds system call table access. An untrusted local user
1489 could exploit this flaw to run code in the kernel
1490 (ie a root privilege escalation). (CVE-2007-4573).
1494 Description: O_DIRECT bypasses client statistics.
1495 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1496 read_ahead_stats were updated. copied the stats section from
1497 osc_send_oap_rpc() into async_internal().
1501 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1505 Description: change order of libsysio includes
1506 Details : '#include sysio.h' should always come before '#include xtio.h'
1510 Description: handle "serious error: objid * already exists" more gracefully
1511 Details : If LAST_ID value on disk is smaller than the objects existing in
1512 the O/0/d* directories, it indicates disk corruption and causes an
1513 LBUG(). If the object is 0-length, then we should use the existing
1514 object. This will help to avoid a full fsck in most cases.
1516 Severity : enhancement
1518 Description: adapt the lustre_config script to support the upgrade case
1519 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1520 server targets to 1.6 in parallel.
1524 Description: To avoid grant space > avaible space when the disk is almost
1525 full. Without this patch you might see the error "grant XXXX >
1526 available" or some LBUG about grant, when the disk is almost
1528 Details : In filter_check_grant, for non_grant cache write, we should
1529 check the left space by if (*left > ungranted + bytes), instead
1530 of (*left > ungranted), because only we are sure the left space
1531 is enough for another "bytes", then the ungrant space should be
1532 increase. In client, we should update cl_avail_grant only there
1533 is OBD_MD_FLGRANT in the reply.
1537 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1538 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1539 described in CVE-2007-4573. This problem could allow untrusted
1540 local users to gain root access.
1543 Frequency : when using O_DIRECT and quotas
1545 Description: Incorrect file ownership on O_DIRECT output files
1546 Details : block usage reported by 'lfs quota' does not take into account
1547 files that have been written with O_DIRECT.
1552 Description: touch file failed when fs is not full
1553 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1554 otherwise we can get ENOSP while fs is not full.
1558 Description: parallel lock callbacks
1559 Details : Instead of sending blocking and completion callbacks as separated
1560 requests, adding them to a set and sending in parallel.
1563 Frequency : only for Cray XT3
1564 Bugzilla : 12829/13455
1565 Description: Changing primary group doesn't change the group lustre assigns to
1567 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1568 overridden with the primary group provided by the group upcall,
1569 whereas the supplied fsgid can be trusted if it is in the list of
1570 supplementary groups returned by the group upcall.
1572 Severity : enhancement
1574 Description: Allow masking D_WARNING, D_ERROR messages from console
1575 Details : Console messages can now be disabled via lnet.printk.
1579 Description: User code with malformed file open parameter crashes client node
1580 Details : Before packing join_file req, all the related reference should be
1581 checked carefully in case some malformed flags cause fake
1582 join_file req on client.
1586 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1587 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1588 mode is equal to the granted mode. However, it should also check
1589 that LDLM_FL_CP_REQD is not set.
1593 Description: Heavy nfs access might result in deadlocks
1594 Details : After ELC code landed, it is now improper to enqueue any mds
1595 locks under och_sem, because enqueue might want to decide to
1596 cancel open locks for same inode we are holding och_sem for.
1600 Description: Client eviction while running blogbench
1601 Details : A lot of unlink operations with concurrent I/O can lead to a
1602 deadlock causing evictions. To address the problem, the number of
1603 oustanding OST_DESTROY requests is now throttled to
1604 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1605 callbacks are processed in priority.
1609 Description: enable ACLs on MDS by default
1610 Details : ACLs must be enabled on MDS by default.
1613 Frequency : PPC/PPC64 only
1615 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1616 Details : fix duplicated definitions between asm-ppc64/types.h and
1617 lustre_types.h on PPC.
1620 Frequency : PPC/PPC64 only
1622 Description: asm-ppc/segment.h does not exist
1623 Details : fix compile issue on PPC.
1627 Description: better handle error messages in extents code
1630 Frequency : RHEL4 only
1632 Description: mkfs is very slow on IA64/RHEL4
1633 Details : A performance regression has been discovered in the MPT Fusion
1634 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1635 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1636 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1638 Severity : enhancement
1640 Description: SNMP support enhancement
1641 Details : Adding total number of sampled request for an MDS node in snmp
1644 Severity : enhancement
1646 Description: Optimize ldlm waiting list processing for PR extent locks
1647 Details : When processing waiting list for read extent lock and meeting read
1648 lock that is same or wider to it that is not contended, skip
1649 processing rest of the list and immediatelly return current
1650 status of conflictness, since we are guaranteed there are no
1651 conflicting locks in the rest of the list.
1655 Description: Time out and refuse to reconnect
1656 Details : When the failover node is the primary node, it is possible
1657 to have two identical connections in imp_conn_list. We must
1658 compare not conn's pointers but NIDs, otherwise we can defeat
1659 connection throttling.
1663 Description: port llog fixes from b1_6 into HEAD
1664 Details : Port llog reference couting and some llog cleanups from b1_6
1665 (bug 10800) into HEAD, for protect from panic and access to already
1666 free llog structures.
1670 Description: Detect stride IO mode in read-ahead
1671 Details : When a client does stride read, read-ahead should detect that and
1672 read-ahead pages according to the detected stride pattern.
1676 Description: data checksumming impacts single node performance
1677 Details : add support for several checksum algorithm. Currently, only CRC32
1678 and Adler-32 are supported. The checksum type can be changed on
1679 the fly via /proc/fs/lustre/osc/*/checksum_type.
1683 Description: use adler32 for page checksums
1684 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1689 Description: build for x2 fails
1690 Details : fix compile issue on Cray systems.
1694 Description: Properly match for duplicate locks
1695 Details : Due to different lock order from skiplists code, we need to
1696 traverse entire list for now
1699 Frequency : only on PPC/SLES10
1701 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1702 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1703 be defined. Add a hack in mkfs_lustre.c to work around this compile
1708 Description: LASSERT on MDS when client holding flock lock dies
1709 Details : ldlm pool logic depends on number of granted locks equal to
1710 number of released locks which is not true for flock locks, so
1711 just exclude such locks from consideration.
1715 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1716 Details : Use fsfilt_sync() for both the whole filesystem sync and
1717 individual file sync to eliminate dangerous inode locking
1718 with I_LOCK that can lead to a deadlock.
1722 Description: performance in 1.6.3
1723 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1724 sure that 1MB requests are not fragmented by the block layer.
1726 Severity : enhancement
1728 Description: organize the server-side client stats on per-nid basis
1729 Details : Change the structure of stats under obdfilter and mds to
1737 The "uuid"s file would list the uuids of _active_ exports.
1738 And the clear entry is to clear all stats and stale nids.
1740 Severity : enhancement
1742 Description: eliminate client locks in face of contention
1743 Details : file contention detection and lockless i/o implementation
1744 for contended files.
1748 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1751 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1753 Description: MDS may lose file striping (and hence file data) in some cases
1754 Details : If there are additional extended attributes stored on the MDS,
1755 in particular ACLs, SELinux, or user attributes (if user_xattr
1756 is specified for the client mount options) then there is a risk
1757 of attribute loss. Additionally, the Lustre file striping
1758 needs to be larger than default (e.g. striped over all OSTs),
1759 and an additional attribute must be stored initially in the
1760 inode and then increase in size enough to be moved to the
1761 external attribute block (e.g. ACL growing in size) for file
1766 Description: skiplist implementation simplification
1767 Details : skiplists are used to group compatible locks on granted list
1768 that was implemented as tracking first and last lock of each
1769 lock group the patch changes that to using doubly linked lists
1773 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1774 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1775 because LASSERT on both the data supplied by a client, and the
1776 data on disk is dangerous and incorrect.
1778 Severity : enhancement
1780 Description: Slow truncate/writes to huge files at high offsets.
1781 Details : Directly associate cached pages to lock that protect those pages,
1782 this allows us to quickly find what pages to write and remove
1783 once lock callback is received.
1787 Description: more ldlm soft lockups
1788 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1789 starve other threads from the resource lock for a long time in
1790 case of long waiting queue, so change the debug level from
1791 D_OTHER to the less frequently used D_INFO.
1793 Severity : enhancement
1795 Description: add -gid, -group, -uid, -user options to lfs find
1799 Description: Hung threads in invalidate_inode_pages2_range
1800 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1801 once one is completed. As a result, some RPCs are stuck in the
1802 queue and are never sent.
1806 Description: filter threads hungs on waiting journal commit
1807 Details : Cleanup filter group llog code, then only filter group llog will
1808 be only created in the MDS/OST syncing process.
1812 Description: Procfs and llog threads access destoryed import sometimes.
1813 Details : Sync the import destoryed process with procfs and llog threads by
1814 the import refcount and semaphore.
1816 Severity : enhancement
1818 Description: openlock cache of b1_6 port to HEAD
1823 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1824 Details : If insertion of an extent fails, then discard the inode
1825 preallocation and free data blocks else it can lead to duplicate
1830 Description: don't always update ctime in ext3_xattr_set_handle()
1831 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1832 In some cases the ctime should not be updated, for example for
1833 2.0->1.8 compatibility it is necessary to delete an xattr and it
1834 should not update the ctime.
1838 Bugzilla : 15713/16362
1839 Description: Assertion in iopen_connect_dentry in 1.6.3
1840 Details : looking up an inode via iopen with the wrong generation number can
1841 populate the dcache with a disconneced dentry while the inode
1842 number is in the process of being reallocated. This causes an
1843 assertion failure in iopen since the inode's dentry list contains
1844 both a connected and disconnected dentry.
1848 Description: assertion failure in ldlm_handle2lock()
1849 Details : fix a race between class_handle_unhash() and class_handle2object()
1850 introduced in lustre 1.6.5 by bug 13622.
1855 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1856 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1861 Description: oops in sd_iostats_seq_show()
1862 Details : unloading/reloading the scsi low level driver triggers a kernel
1863 bug when trying to access the sd iostat file.
1868 Description: Kernel panics during QLogic driver reload
1869 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1870 patch, causing memory corruption.
1875 Description: journal_dev option does not work in b1_6
1876 Details : pass mount option during pre-mount.
1878 Severity : enhancement
1880 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1881 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1882 extent information of a file. It can be used to map logical blocks
1883 in a file to physical blocks in the block device.
1887 Description: LDLM soft lockups - improvement
1888 Details : It is be possible to send the lock handle along with each read
1889 or write request because the client is already doing a lock match
1890 itself so there isn't any reason the OST should have to re-do that
1894 Frequency : only on Cray X2
1896 Description: X2 build failures
1897 Details : fix build failures on Cray X2.
1901 Description: xid & resent requests
1902 Details : Initialize RPC XID from clock at startup (randomly if clock is
1905 Severity : enhancement
1907 Description: Add lustre_start utility to start or stop multiple Lustre servers
1912 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1913 Details : In case of memory pressure, list_del() can be called twice on
1914 req->rq_history_list, causing a kernel oops.
1918 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1919 Details : fix stack overflow in the distributed lock manager by defering
1920 export eviction after a failed AST to the elt thread instead of
1921 handling it in the dlm interpret routine.
1925 Description: Convert some comments to new format.
1926 Details : Update documenting comments to match doxygen conventions.
1930 Description: Grammar fixes.
1931 Details : A couple of trivial spelling fixes.
1935 Description: OSD_COUNTERS-mandatory
1936 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1937 to simplify the code.
1941 Description: simplify cmm_device freeing logic.
1942 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1946 Description: Add lockdep support to dt_object_operations locking interface.
1947 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1948 parameter indicating lock ordering. Update mdd code to use new
1953 Description: Introduce failloc constants for lockless IO tests.
1954 Details : Add two new failloc constants to test lockless IO. Only one of
1955 them in implemented---another is checked in yet to be landed
1960 Description: Add lockdep support for inode mutex.
1961 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1962 in the situations where Lustre has to lock more than one inode
1967 Description: Add optional invariants checking support.
1968 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1969 configure switch. This macro is to be used for consistency and
1970 sanity checks that are too expensive to be left in `production'
1975 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1976 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1981 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1982 Details : Use spin_lock_nested() in (the only) situation where more than
1983 one ldlm_lock is locked simultaneously. Also, fix possible
1984 dead-lock in ldlm_lock_change_resource() by enforcing particular
1989 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1990 Details : Instead of specifying each ldlm_lock call-back through separate
1991 parameter, wrap them into struct ldlm_callback_suite.
1995 Description: Kill join_lru obd method and its callers.
1996 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1997 in memory---a requirement implemented through ->o_join_lru() obd
1998 method in HEAD. Remove this method and its users.
2002 Description: Add asynchronous ldlm ENQUEUE completion handler.
2003 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2004 case that stock ldlm_completion_ast() cannot handle as it waits
2005 until lock is granted. Introduce new ldlm_completion_ast_async()
2006 for this. Also comment ldlm_completion_ast().
2010 Description: ldlm_error <-> errno conversion.
2011 Details : Add functions to map (rather arbitrary) between LDLM error codes
2012 and standard errno values. CLIO needs this to prevent LDLM specific
2013 constants from escaping ldlm and osc.
2017 Description: Kill unused ldlm_handle2lock_ns() function.
2018 Details : Kill unused ldlm_handle2lock_ns() function.
2022 Description: Add lu_ref support to ldlm_lock
2023 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2024 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2025 and ldlm_resource. LDLM interface has to be changed, because code that
2026 releases a reference on a lock, has to "know" what reference this is.
2027 In the most frequent case
2029 lock = ldlm_handle2lock(handle);
2031 LDLM_LOCK_PUT(lock);
2033 no changes are required. When any other reference (received _not_ from
2034 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2035 instead of LDLM_LOCK_PUT().
2037 Arguably, changes are pervasive, and interface requires some discipline
2038 for proper use. On the other hand, it was very instrumental in finding
2039 a few leaked lock references.
2043 Description: Add ldlm_lock_addref_try().
2044 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2045 attempts to addref a lock that might be being canceled
2050 Description: Add ldlm_weigh_callback().
2051 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2052 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2053 hard-coded `number of pages' logic.
2057 Description: Add lockdep annotations to llog code.
2058 Details : Use appropriately tagged _nested() locking calls in the places
2059 where llog takes more than one ->lgh_lock lock.
2063 Description: Add loi_kms_set().
2064 Details : Wrap kms updates into a helper function.
2068 Description: Constify instances of struct lsm_operations.
2069 Details : Constify instances of struct lsm_operations.
2073 Description: lu_conf support.
2074 Details : On a server, a file system object is uniquely identified
2075 by a fid, which is sufficient to locate and load all object
2076 state (inode). On a client, on the other hand, more data are
2077 necessary instantiate an object. Change lu_object_find() and
2078 friends to take additional `lu_conf' argument describing object.
2079 Typically this includes layout information.
2083 Description: lu_context fixes.
2084 Details : Introduce new lu_context functions that are needed on the client
2085 side, where some system threads (ptlrpcd) are shared by multiple
2086 modules, and so cannot be stopped during module shutdown.
2090 Description: Add start and stop methods to lu_device_type_operations.
2091 Details : Introduce two new methods in lu_device_type_operations, that are
2092 invoked when first instance of a given type is created and last one
2093 is destroyed respectively. This is need by CLIO.
2097 Description: Add lu_ref support to struct lu_device.
2098 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2099 track leaked references.
2103 Description: Introduce lu_kmem_descr.
2104 Details : lu_kmem_descr and its companion interface allow to create
2105 and destroy a number of kmem caches at once.
2109 Description: Fix lu_object finalization race.
2110 Details : Fix a race between lu_object_find() finding an object and its
2111 concurrent finalization. This race is (most likely) not possible
2112 on the server, but might happen on the client.
2116 Description: Introduce lu_ref interface.
2117 Details : lu_ref is a debugging module allowing to track references to
2118 a given object. It is quite cpu expensive, and has to be
2119 explicitly enabled with --enable-lu_ref. See usage description
2124 Description: Factor lu_site procfs stats into a separate function.
2125 Details : Separate lu_site stats printing code into a separate function
2126 that can be reused on a client.
2130 Description: Constify instances of struct {lu,dt,md}_device_operations.
2131 Details : Constify instances of struct {lu,dt,md}_device_operations.
2135 Description: Introduce struct md_site and move meta-data specific parts of
2136 struct lu_site here.
2137 Details : Move md-specific fields out of struct lu_site into special struct
2138 md_site, so that lu_site can be used on a client.
2142 Description: Kill mdd_lov_destroy().
2143 Details : Remove unused mdd code.
2147 Description: Add st_block checking to multistat.c.
2148 Details : Add st_block checking to multistat.c.
2152 Description: Add lu_ref support to struct obd_device.
2153 Details : Add lu_ref tracking to obd_device.
2157 Description: Kill obd_set_fail_loc().
2158 Details : Remove unused code.
2162 Description: Add special type for ptlrpc_request interpret functions.
2163 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2164 there. Actual usage will be in CLIO.
2168 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2169 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2170 doesn't work with lockdep.
2174 Description: Add rwv.c test program.
2175 Details : New testing program exercising readv(2) and writev(2) (Qian).
2179 Description: Add sendfile.c test program.
2180 Details : New testing program exercising sendfile(2) (Jay).
2184 Description: Ratelimit a message that can be very frequent.
2185 Details : Ratelimit a memory allocation failure message that can
2190 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2191 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2195 Description: Liblustre build fixes.
2196 Details : Liblustre build fixes.
2200 Description: libcfs: add cfs_{need,cond}_resched() interface.
2201 Details : libcfs: add cfs_{need,cond}_resched() definition and
2202 implementations for Linux, NT, and liblustre.
2204 Severity : enhancement
2206 Description: More exported tunables for mballoc
2207 Details : Add support for tunable preallocation window and new tunables for
2208 large/small requests
2212 Description: Detect corruption of block bitmap and checking for preallocations
2213 Details : Checks validity of on-disk block bitmap. Also it does better
2214 checking of number of applied preallocations. When corruption is
2215 found, it turns filesystem readonly to prevent further corruptions.
2219 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2220 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2221 window to 0. If it does hit cache miss, and read-pattern is still
2222 stride-io mode, does not reset the stride window, but also does not
2223 increase the stride window length in this case.
2227 Frequency : only for big-endian servers
2228 Description: Check if system is big-endian while mounting fs with extents feature
2229 Details : Mounting a filesystem with extents feature will fail on big-endian
2230 systems since ext3-based ldiskfs is not supported on big-endian
2231 systems. This can be over-riden with "bigendian_extents" mount option.
2233 Severity : enhancement
2235 Description: The root squash functionality
2236 Details : A security feature, which is to prevent users from being able
2237 to mount lustre on their desktop, run as root, and delete
2238 all of the files in the filesystem. The goal is accomplished by
2239 remapping user id (UID) and group id (GID) of the root user to
2240 a UID and GID specified by the system administartor via Lustre
2241 configuration management server (MGS). The functionality also
2242 allows to specify sets of clients for which the remapping does
2247 Description: Excessive recovery window
2248 Details : With AT enabled, the recovery window can be excessively long (6000+
2249 seconds). To address this problem, we no longer use
2250 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2251 timeout no longer depends on the service time, it is set to
2252 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2253 time via pb_service_time.
2257 Description: Watchdog triggered on MDS failover
2258 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2259 the OSTs know that the MDS "UUID" can be reused for the same export
2260 from a different NID, so we do not need to wait for the export to be
2264 Frequency : rare, only if using MMP with Linux RAID
2266 Description: MMP doesn't work with Linux RAID
2267 Details : While using HA for Lustre servers with Linux RAID, it is possible
2268 that MMP will not detect multiple mounts. To make this work we
2269 need to unplug the device queue in RAID when the MMP block is being
2270 written. Also while reading the MMP block, we should read it from
2271 disk and not the cached one.
2273 Severity : enhancement
2275 Description: open file using fid
2276 Details : A file can be opened using just its fid, like
2277 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2280 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2283 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2284 Details : add proper locking for creation of jbd_4k slab cache
2288 Description: MMP check in ext3_remount() fails without displaying any error
2289 Details : When multiple mount protection fails during remount, proper error
2292 Severity : enhancement
2294 Description: Allow stripe size to be up to 4G-64k
2295 Details : Fix math logic to allow large stripe sizes.
2299 Description: add check for >8TB ldiskfs filesystems
2300 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2301 Don't allow >8TB ldiskfs filesystems to be mounted without
2302 force_over_8tb mount option
2307 Descriptoin: OSS complains "too many blocks" and crashes
2308 Details : In filter_commitrw_write():
2309 1. lquota_chkquota();
2310 2. obdo_from_inode();
2311 3. lquota_pending_commit();
2312 "2." has changed uid and gid in this case, then caused
2313 lquota_chkquota() and lquota_pending_commit() to use
2314 different uid and gid. That is the root of the bug.
2318 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2319 Details : Detect and print out the 'stripe count = -1' case properly.
2323 Description: Out or order replies might be lost on replay
2324 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2325 request with tid smaller then one currently being added, add it
2326 to the start, not end of the list.
2328 --------------------------------------------------------------------------------
2330 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2332 * Support for kernels:
2333 2.6.5-7.283 (SLES 9),
2334 2.6.9-55.EL (RHEL 4),
2335 2.6.16.46-0.14 (SLES 10),
2336 2.6.18.8 vanilla (kernel.org)
2337 * Client support for unpatched kernels:
2338 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2339 2.6.16 - 2.6.22 vanilla (kernel.org)
2340 * Due to recently discovered recovery problems, we do not recommend
2341 using patchless RHEL 4 clients with this or any earlier release.
2342 * Recommended e2fsprogs version: 1.39.cfs8
2343 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2344 * Starting with this release, the ldiskfs backing filesystem required
2345 by Lustre is now in its own package, lustre-ldiskfs. This package
2346 should be installed. It is versioned separately from Lustre and
2347 may be released separately in future.
2351 Description: block reactivating mgc import until all deactivates complete
2352 Details : Fix race when failing back MDT/MGS to itself (testing)
2355 Frequency : at statup only
2357 Description: mds_lov_synchronize race leads to various problems
2358 Details : simultaneous MDT->OST connections at startup can cause the
2359 sync to abort, leaving the OSC in a bad state.
2361 Severity : enhancement
2363 Description: add optional extra BUILD_VERSION info
2364 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2365 to override the lustre version.
2368 Frequency : 2.6.18 servers only
2370 Description: ll_kern_mount() doesn't release the module reference
2371 Details : The ldiskfs module reference count never drops down to 0
2372 because ll_kern_mount() doesn't release the module reference.
2377 Description: server LBUG when using old ost_num_threads parameter
2378 Details : Accept the old ost_num_threads parameter but warn that it
2379 is deprecated, and fix an off-by-one error that caused an LBUG.
2384 Description: Transient SCSI error results in persistent IO issue
2385 Details : iobuf->dr_error is not reinitialized to 0 between two
2389 Frequency : sometimes when underlying device returns I/O errors
2391 Description: OSTs not going read-only during write failures
2392 Details : OSTs are not remounted read-only when the journal commit threads
2393 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2394 instead of the ext3 wrappers.
2398 Description: poor connect scaling with increasing client count
2399 Details : Don't run filter_grant_sanity_check for more than 100 exports
2400 to improve scaling for large numbers of clients.
2403 Frequency : SLES10 only
2405 Description: sanity-quota.sh quotacheck failed: rc = -22
2406 Details : Quotas cannot be enabled on SLES10.
2409 Frequency : liblustre clients only
2411 Description: getdirentries does not give error when run on compute nodes
2412 Details : getdirentries does not fail when the size specified as an argument
2413 is too small to contain at least one entry
2415 Severity : enhancement
2417 Description: Add LNET router traceability for debug purposes
2418 Details : If a checksum failure occurs with a router as part of the
2419 IO path, the NID of the last router that forwarded the bulk data
2420 is printed so it can be identified.
2425 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2426 Details : Due to a race condition, liblustre clients were occasionally
2427 evicted incorrectly.
2429 Severity : enhancement
2431 Description: lfs setstripe use optional parameters instead of postional
2434 Severity : enhancement
2436 Description: Nanosecond timestamp support for ldiskfs
2437 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2438 resolution timestamps. There is not yet support for this at
2439 the Lustre filesystem level.
2442 Frequency : during server recovery
2444 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2445 Details : request with rq_no_resend flag not awake l_wait_event if they get
2449 Frequency : nfs export on patchless client
2451 Description: connectathon hang when test nfs export over patchless client
2452 Details : Disconnected dentry cannot be found with lookup, so we do not need
2453 to unhash it or make it invalid
2456 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2457 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2458 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2463 Description: open req refcounting wrong on reconnect
2464 Details : If reconnect happened between getting open reply from server and
2465 call to mdc_set_replay_data in ll_file_open, we will schedule
2466 replay for unreferenced request that we are about to free.
2467 Subsequent close will crash in variety of ways.
2468 Check that request is still eligible for replay in
2469 mdc_set_replay_data().
2474 Description: disable writes to filesystem when reading health_check file
2475 Details : the default for reading the health_check proc file has changed
2476 to NOT do a journal transaction and write to disk, because this
2477 can cause reads of the /proc file to hang and block HA state
2478 checking on a healthy but otherwise heavily loaded system. It
2479 is possible to return to the previous behaviour during configure
2480 with --enable-health-write.
2482 Severity : enhancement
2484 Description: 64-bit inode version
2485 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2486 to the inode. This will be required for version-based recovery.
2491 Description: MDS fails to start if a duplicate client export is detected
2492 Details : in some rare cases it was possible for a client to connect to
2493 an MDS multiple times. Upon recovery the MDS would detect this
2494 and fail during startup. Handle this more gracefully.
2496 Severity : enhancement
2498 Description: Add -o localflock option to simulate old noflock
2500 Details : This will achieve local-only flock/fcntl locks
2506 Description: log_commit_thread vs filter_destroy race leads to crash
2507 Details : Take import reference before releasing llog record semaphore
2512 Description: Wrong request locking in request set processing
2513 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2514 imp_delayed_list, in this place should be used imp_lock.
2517 Frequency : when reconnection
2519 Description: Grant Leak when osc reconnect to OST
2520 Details : When osc reconnect ost, OST(filter) should check whether it
2521 should grant more space to client by comparing fed_grant and
2522 cl_avail_grant, and return the granted space to client instead
2523 of "new granted" space, because client will call osc_init_grant
2524 to update the client grant space info.
2527 Frequency : when client reconnect to OST
2529 Description: Grant Leak when osc do resend and replay bulk write
2530 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2531 bulk write request, because the grant info while be sync between
2532 OSC and OST when reconnect, and we should ignore the grant info
2533 these of resend/replay write req.
2538 Description: Grant space more than avaiable left space sometimes.
2539 Details : When then OST is about to be full, if two bulk writing from
2540 different clients came to OST. Accord the avaliable space of the
2541 OST, the first req should be permitted, and the second one
2542 should be denied by ENOSPC. But if the seconde arrived before
2543 the first one is commited. The OST might wrongly permit second
2544 writing, which will cause grant space > avaiable space.
2547 Frequency : when client is evicted
2549 Description: Grant might be wrongly erased when osc is evicted by OST
2550 Details : when the import is evicted by server, it will fork another
2551 thread ptlrpc_invalidate_import_thread to invalidate the
2552 import, where the grant will be set to 0. While the original
2553 thread will update the grant it got when connecting. So if
2554 the former happened latter, the grant will be wrongly errased
2555 because of this race.
2560 Description: Checking Stale with correct fid
2561 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2562 to check whether it is stale, because sometimes, we want the
2563 enqueue happened anyway, and op_data.fid2 will not be initialized.
2565 Severity : enhancement
2567 Description: update patchless client
2568 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2571 Frequency : only with 2.4 kernel
2573 Description: random memory corruption
2574 Details : size of struct ll_inode_info is to big for union inode.u and this
2575 can be cause of random memory corruption.
2580 Description: Memory leak in recovery
2581 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2582 It should also check obd_fail before fsfilt_start, otherwise if
2583 fsfilt_start return -EROFS,(failover mds during mds recovery).
2584 then the req will return with repmsg->transno = 0 and rc = EROFS.
2585 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2586 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2587 be freed no matter whether fsfilt_commit success or not.
2590 Frequency : only with huge count clients
2592 Description: Prevents from taking the superblock lock in llap_from_page for
2594 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2595 for llap_from_page call in ll_removepage prevents from taking the
2596 superblock lock for a soon died page.
2601 Description: Not check open intent error before release open handle
2602 Details : in some rare cases, the open intent error is not checked before
2603 release open handle, which may cause
2604 ASSERTION(open_req->rq_transno != 0), because it tries to release
2605 the failed open handle.
2610 Description: Set cat log bitmap only after create log success.
2611 Details : in some rare cases, the cat log bitmap is set too early. and it
2612 should be set only after create log success.
2616 Description: Accessing a block bevice can re-enable I/O when Lustre is
2617 tearing down a device.
2618 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2623 Description: service threads may hog cpus when there are a lot of requests
2624 Details : Insert cond_resched to give other threads a chance to use some CPU
2629 Description: the cat log was not initialized in recovery
2630 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2631 unlink log on mds will not be initialized until mds post
2632 recovery. And also in mds post recovery, the unlink log will
2633 initialization will be done asynchronausly, so there will be race
2634 between add unlink log and unlink log initialization.
2638 Description: brw_stats were being printed incorrectly
2639 Details : brw_stats were being printed as log2 but all of them were not
2640 recorded as log2. Also remove some code duplication arising from
2641 filter_tally_{read,write}.
2645 Frequency : rare, only in recovery.
2646 Description: ASSERTION(req->rq_type != LI_POISON) failed
2647 Details : imp_lock should be held while iterating over imp_sending_list for
2648 prevent destroy request after get timeout in ptlrpc_queue_wait.
2652 Description: replay-single.sh test 52 fails
2653 Details : A lock's skiplist need to be cleanup when it being unlinked
2654 from its resource list.
2658 Description: Short directio read returns full requested size rather than
2660 Details : Direct I/O operations should return actual amount of bytes
2661 transferred rather than requested size.
2663 Severity : enhancement
2665 Description: metadata RPC reduction (e.g. for rm performance)
2666 Details : decrease the amount of synchronous RPC between clients and servers
2667 by canceling conflicing lock before the operation on the client
2668 and packing thier handles into the main operation RPC to server.
2670 Severity : enhancement
2672 Description: Async OSC create to avoid the blocking unnecessarily.
2673 Details : If a OST has no remain object, system will block on the creating
2674 when need to create a new object on this OST. Now, ways use
2675 pre-created objects when available, instead of blocking on an
2676 empty osc while others are not empty. If we must block, we block
2677 for the shortest possible period of time.
2681 Description: improve handling recoverable errors
2682 Details : if request processig with error which can be recoverable on server
2683 request should be resend, otherwise page released from cache and
2686 Severity : enhancement
2688 Description: refine locking for avoid write wrong info into lov_objid file
2689 Details : fix possible races with add new target and write/update data in
2692 --------------------------------------------------------------------------------
2694 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2699 Frequency : on some architectures
2701 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2702 Details : Uninitialized flags sometimes cause configuration commands to
2706 Frequency : patchless clients only
2708 Description: missing __iget() symbol export
2709 Details : The __iget() symbol export is missing. To avoid the need for
2710 this on patchless clients the deathrow inode reaper is turned
2711 off, and we depend on the VM to clean up old inodes. This
2712 dependency was during via the fix for bug 12181.
2714 --------------------------------------------------------------------------------
2716 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2718 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2719 INTEROPERATE with older versions automatically. In many cases a
2720 special upgrade step is needed. Please read the
2721 user documentation before upgrading any part of a live system.
2722 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2723 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2724 * WARNING: Lustre configuration and startup changes are required with
2725 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2729 Severity : enhancement
2731 Description: MountConf
2732 Details : Lustre configuration is now managed via mkfs and mount
2733 commands instead of lmc and lconf. New obd types (MGS, MGC)
2734 are added for dynamic configuration management. See
2735 https://mail.clusterfs.com/wikis/lustre/MountConf for
2738 Severity : enhancement
2740 Description: dynamic OST addition
2741 Details : OSTs can now be added to a live filesystem
2743 Severity : enhancement
2745 Description: startup order invariance
2746 Details : MDTs and OSTs can be started in any order. Clients only
2747 require the MDT to complete startup.
2749 Severity : enhancement
2751 Description: parallel, asynchronous orphan cleanup
2752 Details : orphan cleanup is now performed in separate threads for each
2753 OST, allowing parallel non-blocking operation.
2755 Severity : enhancement
2757 Description: optimized stripe assignment
2758 Details : stripe assignments are now made based on ost space available,
2759 ost previous usage, and OSS previous usage, in order to try
2760 to optimize storage space and networking resources.
2762 Severity : enhancement
2764 Description: Permanently set tunables
2765 Details : All writable /proc/fs/lustre tunables can now be permanently
2766 set on a per-server basis, at mkfs time or on a live system.
2768 Severity : enhancement
2770 Description: Lustre message v2
2771 Details : Add lustre message format v2.
2773 Severity : enhancement
2775 Description: client OST exclusion list
2776 Details : Clients can be started with a list of OSTs that should be
2777 declared "inactive" for known non-responsive OSTs.
2781 Description: SPEC SFS validation failure on NFS v2 over lustre.
2782 Details : Changes the blocksize for regular files to be 2x RPC size,
2783 and not depend on stripe size.
2785 Severity : enhancement
2787 Description: Multiple MD RPCs in flight.
2788 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2789 To avoid overly-overloading MDS, introduce a limit on number of
2790 MDS RPCs in flight for a single client and add /proc controls
2791 to adjust this limit.
2793 Severity : enhancement
2795 Description: client read/write statistics
2796 Details : Add client read/write call usage stats for performance
2797 analysis of user processes.
2798 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2799 file access. extents_stats shows chunk size distribution.
2800 extents_stats_per_process show chunk size distribution per
2803 Severity : enhancement
2805 Description: mds statistics
2806 Details : Add detailed mds operations statistics in
2807 /proc/fs/lustre/mds/*/stats.
2811 Description: Failure of copying files with lustre special EAs.
2812 Details : Client side always return success for setxattr call for lustre
2813 special xattr (currently only "trusted.lov").
2818 Description: Refcount LNET uuids
2819 Details : The global LNET uuid list grew linearly with every startup;
2820 refcount repeated list entries instead of always adding to
2823 Severity : enhancement
2825 Description: Dynamic service threads
2826 Details : Within a small range, start extra service threads
2827 automatically when the request queue builds up.
2830 Frequency : mixed-endian client/server environments
2832 Description: mixed-endian crashes
2833 Details : The new msg_v2 system had some failures in mixed-endian
2836 Severity : enhancement
2838 Description: Easy OST removal
2839 Details : OSTs can be permanently deactivated with e.g. 'lctl
2840 conf_param lustre-OST0001.osc.active=0'
2842 Severity : enhancement
2844 Description: MGS proc entries
2845 Details : Added basic proc entries for the MGS showing what filesystems
2848 Severity : enhancement
2850 Description: provide MGS failover
2851 Details : Added config lock reacquisition after MGS server failover.
2853 Severity : enhancement
2855 Description: add Linux 2.4 support
2856 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2860 Description: a large application tries to do I/O to the same resource and dies
2861 in the middle of it.
2862 Details : Check the req->rq_arrival time after the call to
2863 ost_brw_lock_get(), but before we do anything about
2864 processing it & sending the BULK transfer request. This
2865 should help move old stale pending locks off the queue as
2866 quickly as obd_timeout.
2869 Frequency : when an incorrect nid is specified during startup
2871 Description: ptlrpc connect to non-existant node causes kernel crash
2872 Details : LNET can't be re-entered from an event callback, which
2873 happened when we expire a message after the export has been
2874 cleaned up. Instead, hand the zombie cleanup off to another
2877 Severity : enhancement
2879 Description: plain/inodebits lock performance improvement
2880 Details : Grouping plain/inodebits in granted list by their request modes
2881 and bits policy, thus improving the performance of search through
2885 Frequency : only if OST filesystem is corrupted
2887 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2888 Details : for a short time RPCs with bulk IO are in the replay list,
2889 but replay of bulk IOs is unimplemented. If the OST filesystem
2890 is corrupted due to disk cache incoherency and then replay is
2891 started it is possible to trip an assertion. Avoid putting
2892 committed RPCs into the replay list at all to avoid this issue.
2895 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2897 Description: System hang on startup
2898 Details : This bug allowed the liblustre (e.g. catamount) client to
2899 return to the app before handling all startup RPCs. This
2900 could leave the node unresponsive to lustre network traffic
2901 and manifested as a server ptllnd timeout.
2903 Severity : enhancement
2905 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2906 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2907 boolean to control whether to print peer debug info when a
2908 client's RPC times out.
2911 Frequency : only for kernels with patches from Lustre below 1.4.3
2913 Description: Remove old rdonly API
2914 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2917 Frequency : only for devices with external journals
2919 Description: Set external device read-only also
2920 Details : During a commanded failover stop, we set the disk device
2921 read-only while the server shuts down. We now also set any
2922 external journal device read-only at the same time.
2925 Frequency : when upgrading from 1.4 while trying to change parameters
2927 Description: The wrong (new) MDC name was used when setting parameters for
2928 upgraded MDT's. Also allows changing of OSC (and MDC)
2929 parameters if --writeconf is specified at tunefs upgrade time.
2932 Frequency : when setting specific ost indicies
2934 Description: QOS code breaks on skipped indicies
2935 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2936 created with --index need not be sequential.
2938 Severity : enhancement
2940 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2941 Details : The uninit_groups feature works in conjunction with the kernel
2942 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2943 up the pass1 processing of e2fsck. This is a read-only feature
2944 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2945 mount filesystems that have had this feature enabled.
2947 Severity : enhancement
2949 Description: Improve multi-block allocation algorithm to avoid fragmentation
2950 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2951 allocation locality and avoid filesystem fragmentation.
2953 ------------------------------------------------------------------------------
2955 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2957 * Support for kernels:
2958 2.4.21-47.0.1.EL (RHEL 3)
2959 2.6.5-7.283 (SLES 9)
2960 2.6.9-42.0.10.EL (RHEL 4)
2961 2.6.12.6 vanilla (kernel.org)
2962 2.6.16.27-0.9 (SLES 10)
2963 * Recommended e2fsprogs version: 1.39.cfs5
2965 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2969 Frequency : occasional, depends on client load and configuration
2970 Bugzilla : 12181, 12203
2971 Description: data loss for recently-modified files
2973 Details : In some cases it is possible that recently written or created
2974 files may not be written to disk in a timely manner (this should
2975 normally be within 30s unless client IO load is very high).
2976 The problem appears as zero-length files or files that are a
2977 multiple of 1MB in size after a client crash or client eviction
2978 that are missing data at the end of the file.
2980 This problem is more likely to be hit on clients where files are
2981 repeatedly created and unlinked in the same directory, clients
2982 have a large amount of RAM, have many CPUs, the filesystem has
2983 many OSTs, the clients are rebooted frequently, and/or the files
2984 are not accessed by other nodes after being written.
2986 The presence of the problem can be detected by looking at
2987 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2988 smaller than the second (nr_unused) then dirty files will not
2989 be flushed automatically to disk. "sync; sleep 10" should be
2990 run several times on the node before unmounting it to update
2991 Lustre (this is also safe to run on nodes without this problem).
2993 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2994 that can cause this same problem, so customers using that kernel
2995 also need to update the kernel in addition to Lustre. In order
2996 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2998 It is normal that files written just before a client crash (less
2999 than 30s) may not yet have been flushed to disk, even for local
3003 Frequency : frequent on thin XT3 nodes
3005 Description: UUID collision on thin XT3 Linux nodes
3006 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3007 randomly, since we relied on an insufficiently-seeded PRNG.
3012 Description: OSS hangs after "All ost request buffers busy"
3013 Details : A deadlock between quota and journal operations caused OSS
3014 hangs after printing "All ost request buffers busy."
3017 Frequency : always on liblustre builds
3019 Description: Cleanup compiler warnings on liblustre
3022 Frequency : always on liblustre builds on XT3
3024 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3028 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3029 > 32000 subdirectories
3030 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3031 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3032 subdirectory count crosses 32000. This will aid e2fsck to
3033 correctly handle more than 32000 subdirectories.
3036 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3038 Description: System hang on startup
3039 Details : This bug allowed the liblustre (e.g. catamount) client to
3040 return to the app before handling all startup RPCs. This
3041 could leave the node unresponsive to lustre network traffic
3042 and manifested as a server ptllnd timeout.
3044 Severity : enhancement
3046 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3047 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3048 boolean to control whether to print peer debug info when a
3049 client's RPC times out.
3054 Description: make O_SYNC working on 2.6 kernels
3055 Details : 2.6 kernels use different method for mark pages for write,
3056 so need add a code to lustre for O_SYNC work.
3061 Description: Failure to close file and release space on NFS
3062 Details : Put inode details into lock acquired in ll_intent_file_open.
3063 Use mdc_intent_lock in ll_intent_open to properly
3064 detect all kind of errors unhandled by mdc_enqueue.
3069 Description: proc file read during shutdown sometimes raced obd removal,
3071 Details : Add lock to prevent obd access after proc file removal.
3074 Frequency : Only for files larger than 4GB on 32-bit clients.
3076 Description: improperly doing page alignment of locks
3077 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3078 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3083 Description: RPCs being resent when they shouldn't be
3084 Details : Some RPCs that should not be resent are being resent. This
3085 can cause inconsistencies in the RPC state machine. Do not
3086 resend such requests.
3089 Frequency : rare, only with NFS export
3091 Description: Crash on NFS re-export node
3092 Details : under very unusual load conditions an assertion is hit in
3093 ll_intent_file_open()
3096 Frequency : only if OST filesystem is corrupted
3098 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3099 Details : for a short time RPCs with bulk IO are in the replay list,
3100 but replay of bulk IOs is unimplemented. If the OST filesystem
3101 is corrupted due to disk cache incoherency and then replay is
3102 started it is possible to trip an assertion. Avoid putting
3103 committed RPCs into the replay list at all to avoid this issue.
3108 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3109 Details : Large single O_DIRECT read and write calls can fail to allocate
3110 a sufficiently large buffer to process the request. In case of
3111 allocation failure the allocation is retried with a smaller
3112 buffer and broken into smaller requests.
3114 Severity : enhancement
3116 Description: Add -o localflock option to simulate old noflock behaviour.
3117 Details : This will achieve local-only flock/fcntl locks coherentness.
3122 Description: versioning check is incomplete
3123 Details : Checking the version difference of client vs. server, report
3124 error if the gap is too big.
3129 Description: add support PG_writeback bit
3130 Details : add support for PG_writeback bit for Lustre, for more carefull
3131 work with page cache in 2.6 kernel. This also fix some deadlocks
3132 and remove hack for work O_SYNC with 2.6 kernel.
3134 Severity : enhancement
3136 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3137 Details : The uninit_groups feature works in conjunction with the kernel
3138 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3139 up the pass1 processing of e2fsck. This is a read-only feature
3140 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3141 mount filesystems that have had this feature enabled.
3143 Severity : enhancement
3145 Description: Improve multi-block allocation algorithm to avoid fragmentation
3146 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3147 allocation locality and avoid filesystem fragmentation.
3149 ------------------------------------------------------------------------------
3151 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3153 * Support for kernels:
3154 2.6.9-42.0.3.EL (RHEL 4)
3155 2.6.5-7.276 (SLES 9)
3156 2.4.21-47.0.1.EL (RHEL 3)
3157 2.6.12.6 vanilla (kernel.org)
3158 2.6.16.21-0.8 (SLES10)
3159 * Recommended e2fsprogs version: 1.39.cfs2-0
3161 * The backwards-compatible /proc/sys/portals symlink has been removed
3162 in this release. Before upgrading, please ensure that you change
3163 any configuration scripts or /etc/sysctl.conf files that access
3164 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3165 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3166 in advance of the upgrade on any system running Lustre 1.4.6 or
3167 newer, since /proc/sys/lnet was added in that version.
3168 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3172 Frequency : only when quota is used
3174 Description: avoid scanning export list for quota master
3175 Details : Change the algorithms to avoid scanning export list in order
3176 to improve the efficiency.
3179 Frequency : MDS failover only, very rarely
3181 Description: "went back in time" messages on mds failover
3182 Details : The greatest transno may be lost when the current operation
3183 finishes with an error (transno==0) and the client's last_rcvd
3184 record is over-written. Save the greatest transno in the
3185 mds_last_transno for this case.
3188 Frequency : always for specific kernels and striping counts
3190 Description: client may get "Matching packet too big" without ACL support
3191 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3192 when trying to access files in certain configurations. The
3193 clients should in fact be denied when mounting because they do
3194 not understand ACLs.
3197 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3199 Description: many clients connecting with IO in progress causes connect timeouts
3200 Details : Avoid synchronous journal commits to avoid delays caused by many
3201 clients connecting/disconnecting when bulk IO is in progress.
3202 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3203 OST_IO_PORTAL to avoid delays behind potentially many pending
3207 Frequency : occasionally with multiple writers to a single file
3209 Description: shared writes to file may result in wrong size reported by stat()
3210 Details : Allow growing of kms when extent lock is cancelled
3213 Frequency : always with random mmap IO to multi-striped file
3215 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3216 Details : Only if the hole is at the end of OST object so that kms is too
3217 small. Fix is to increase kms accordingly in ll_nopage.
3220 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3222 Description: writes to a missing object would leak memory on the OST
3223 Details : If there is an inconsistency between the MDS and OST filesystems,
3224 such that the MDS references an object that doesn't exist, writes
3225 to that object will leak memory due to incorrect cleanup in the
3226 error handling path, eventually running out of memory on the OST.
3231 Description: Creating too long symlink causes lustre errors
3232 Details : Check symlink and name lengths before sending requests to MDS.
3235 Frequency : only if flock is enabled (not on by default)
3237 Description: posix locks not released on fd closure on 2.6.9+
3238 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3239 kernels, this caused such locks not to be released on fd close and
3240 then assertions on fs unmount about still used locks.
3243 Frequency : MDS failover only, very rarely
3245 Description: clients may get ASSERTION(granted_lock != NULL)
3246 Details : When request was taking a long time, and a client was resending
3247 a getattr by name lock request. The were multiple lock requests
3248 with the same client lock handle and
3249 mds_getattr_name->fixup_handle_for_resent_request found one of the
3250 lock handles but later failed with ASSERTION(granted_lock != NULL).
3255 Description: handle->h_buffer_credits > 0, assertion failure
3256 Details : h_buffer_credits is zero after truncate, causing assertion
3257 failure. This patch extends the transaction or creates a new
3261 Frequency : NFS re-export or patchless client
3262 Bugzilla : 11179, 10796
3263 Description: Crash on NFS re-export node (__d_move)
3264 Details : We do not want to hash the dentry if we don't have a lock.
3265 But if this dentry is later used in d_move, we'd hit uninitialised
3266 list head d_hash, so we just do this to init d_hash field but
3267 leave dentry unhashed.
3270 Frequency : NFS re-export or patchless client
3272 Description: NFS exports has problem with symbolic link
3273 Details : lustre client didn't properly install dentry when re-exported
3274 to NFS or running patchless client.
3277 Frequency : NFS re-export or patchless client
3279 Description: Various nfs/patchless fixes.
3280 Details : fixes reuse disconected alias for lookup process - this fixes
3281 warning "find_exported_dentry: npd != pd",
3282 fix permission error with open files at nfs.
3283 fix apply umask when do revalidate.
3286 Frequency : occasional
3288 Description: Crash on NFS re-export node
3289 Details : calling clear_page() on the wrong pointer triggered oops in
3290 generic_mapping_read().
3293 Frequency : rarely, using O_DIRECT IO
3295 Description: unaligned directio crashes client with LASSERT
3296 Details : check for unaligned buffers before trying any requests.
3299 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3301 Description: stale data returned from RAID cache
3302 Details : If only a small amount of IO is done to the RAID device before
3303 reading it again it is possible to get stale data from the RAID
3304 cache instead of reading it from disk.
3307 Frequency : always for sles10 kernel
3309 Description: sles10 support
3310 Details : ll_follow_link: compile fixes and using of nd_set_link
3311 under newer kernels.
3314 Frequency : depends on arch, kernel and compiler version, always on sles10
3317 Description: recursive or deep enough symlinks cause stack overflow
3318 Details : getting rid of large stack-allocated variable in
3322 Frequency : depends on hardware
3324 Description: lustre write performance loss in the SLES10 kernel
3325 Details : the performance loss is caused by using of write barriers in the
3326 ext3 code. The SLES10 kernel turns barrier support on by
3327 default. The fix is to undo that change for ldiskfs.
3329 ------------------------------------------------------------------------------
3331 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3333 * Support for kernels:
3334 2.6.9-42.0.3EL (RHEL 4)
3335 2.6.5-7.276 (SLES 9)
3336 2.4.21-47.0.1.EL (RHEL 3)
3337 2.6.12.6 vanilla (kernel.org)
3341 Frequency : quota enabled and large files being deleted
3343 Description: releasing more than 4GB of quota at once hangs OST
3344 Details : If a user deletes more than 4GB of files on a single OST it
3345 will cause the OST to spin in an infinite loop. Release
3346 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3351 Description: statfs data retrieved from /proc may be stale or zero
3352 Details : When reading per-device statfs data from /proc, in the
3353 {kbytes,files}_{total,free,avail} files, it may appear
3354 as zero or be out of date.
3357 Frequency : systems with MD RAID1 external journal devices
3359 Description: lconf's call to blkid is confused by RAID1 journal devices
3360 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3361 of returning all block devices that match the journal UUID.
3364 Frequency : always, for aggregate stripe size over 4GB
3366 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3367 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3368 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3371 Frequency : always if "lfs find" used on a local file/directory
3373 Description: "lfs find" segfaults if used on a local file/directory
3374 Details : The case where a directory component was not specified wasn't
3375 handled correctly. Handle this properly.
3378 Frequency : always on ppc64
3380 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3381 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3382 perform endianness conversion.
3385 Frequency : rarely (truncate to non-zero file size after write under load)
3386 Bugzilla : 10730, 10687
3387 Description: Files padded with zeros to next 4K multiple
3388 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3389 it is possible that files that are truncated to a non-zero size
3390 immediately after being written are filled with zero bytes beyond
3391 the truncated size. No file data is lost.
3393 Severity : enhancement
3395 Description: Allow recovery/failover for liblustre clients.
3396 Details : liblustre clients were unaware of failover configurations until
3399 Severity : enhancement
3401 Description: user file locks should fail when not mounting with flock option
3402 Details : Set up an error-returning stub in ll_file_operations.lock field
3403 to prevent incorrect behaviour when client is mounted without
3404 flock option. Also, set up properly f_op->flock field for
3408 Frequency : always on ia64
3410 Description: "lfs df" loops on printing out MDS statfs information
3411 Details : The obd_ioctl_data was not initialized and in some systems
3412 this caused a failure during the ioctl that did not return
3413 an error. Initialize the struct and return an error on failure.
3416 Frequency : SLES 9 only
3418 Description: Error of copying files with lustre special EAs as root
3419 Details : Client side always return success for setxattr call for lustre
3420 special xattr (currently only "trusted.lov").
3423 Frequency : rarely on clusters with both ia64+i386 clients
3425 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3426 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3427 clients are doing concurrent writes to the same file it is
3428 possible that the ia64 clients may LASSERT because the OST
3429 extent locks are not PAGE_SIZE aligned. Ensure that grown
3430 locks are always aligned on the request boundary.
3433 Frequency : specific use, occasional
3435 Description: Overwriting in use executable truncates on-disk binary image
3436 Details : If one node attempts to overwrite an executable in use by
3437 another node, we now correctly return ETXTBSY instead of
3438 truncating the file.
3440 Severity : enhancement
3442 Description: Async OSC create to avoid the blocking unnecessarily.
3443 Details : If a OST has no remain object, system will block on the creating
3444 when need to create a new object on this OST. Now, ways use
3445 pre-created objects when available, instead of blocking on an
3446 empty osc while others are not empty. If we must block, we block
3447 for the shortest possible period of time.
3452 Description: chmod on Lustre root is propagated to other clients
3453 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3454 invalid by the follow_mount time.
3457 Frequency : liblustre clients only
3459 Description: Race in 'instant cancel' lock handling could lead to such locks
3460 never to be granted in case of SMP MDS
3461 Details : Do not destroy not yet granted but cbpending locks in
3465 Frequency : replay/resend of open
3467 Description: non null lock assetion failure in mds_intent_policy
3468 Details : Trying to replay/resend lockless open requests resulted in
3469 mds_open() returning 0 with no lock. Now it sets a flag if
3470 a lock is going to be returned.
3472 Severity : enhancement
3474 Description: Checksum enhancements
3475 Details : New checksum enhancements allow for resending RPCs that failed
3478 Severity : enhancement
3480 Description: Tunables on number of dirty pages in cacche
3481 Details : Allow to set limit on number of dirty pages cached.
3486 Description: client crash on unmount - lock still has references
3487 Details : In some error handling cases it was possible to leak a lock
3488 reference on a client while accessing a file. This was not
3489 harmful to the client during operation, but would cause the
3490 client to crash when the filesystem is unmounted.
3493 Frequency : specific case, rare
3495 Description: ETXTBSY on mds though file not in use by client
3496 Details : ETXTBSY is no longer incorrectly returned when attempting to
3497 chmod or chown a directory that the user previously tried to
3498 execute or a currently-executing binary.
3501 Frequency : extremely rare except on liblustre-based clients
3503 Description: Lustre space not freed when files are deleted
3504 Details : Clean up open-unlinked files after client eviction. Previously
3505 the unlink was skipped and the files remained as orphans.
3510 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3511 Details : In some cases it was possible to send a blocking callback to
3512 a client doing a glimpse, even though that client didn't get
3513 a lock granted. When the glimpse lock is cancelled on the OST
3514 the freed lock is left on the waiting list and corrupted the list.
3517 Frequency : all core dumps
3519 Description: Broke core dumps to lustre
3520 Details : Negative dentry may be unhashed if parent does not have UPDATE
3521 lock, but some callers, e.g. do_coredump, expect dentry to be
3522 hashed after successful create, hash it in ll_create_it.
3524 ------------------------------------------------------------------------------
3526 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3528 * Support for kernels:
3529 2.6.9-42.0.2.EL (RHEL 4)
3530 2.6.5-7.276 (SLES 9)
3531 2.4.21-40.EL (RHEL 3)
3532 2.6.12.6 vanilla (kernel.org)
3536 Frequency : always on RHEL 3
3538 Description: Number of open files grows over time
3539 Details : The number of open files grows over time, whether or not
3540 Lustre is started. This was due to a filp leak introduced
3541 by one of our kernel patches.
3543 ------------------------------------------------------------------------------
3545 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3547 * Support for kernels:
3548 2.6.9-42.EL (RHEL 4)
3549 2.6.5-7.276 (SLES 9)
3550 2.4.21-40.EL (RHEL 3)
3551 2.6.12.6 vanilla (kernel.org)
3556 Bugzilla : 5719, 9635, 9792, 9684
3557 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3558 Details : If a server is under heavy load and cannot reply to new
3559 connection requests before the client resends the (re)connect,
3560 the connection handling code can behave badly if two service
3561 threads are concurrently handing separate (re)connections from
3562 the same client. Add better locking to the connection handling
3563 code, and ensure that only a single connection will be processed
3564 for a given client UUID, even if the lock is dropped.
3566 Severity : enhancement
3568 Description: add TCP zero-copy support to kernel
3569 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3570 sends if the hardware supports scatter-gather and checksumming.
3571 This allows socklnd to do client-write and server-read more
3572 efficiently and reduce CPU utilization from skbuf copying.
3575 Frequency : only if NFS exporting from client
3577 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3578 Details : If setting attributes on a file created under NFS that had
3579 never been opened it would be possible to oops the client
3580 if the file had no objects.
3583 Frequency : always for liblustre
3585 Description: liblustre client does MDS+OSTs setattr RPC for each write
3586 Details : When doing a write from a liblustre client, the client
3587 incorrectly issued an RPC to the MDS and each OST the file was
3588 striped over in order to update the timestamps. When writing
3589 with small chunks and many clients this could overwhelm the MDS
3590 with RPCs. In all cases it would slow down the write because
3591 these RPCs are unnecessary.
3593 Severity : enhancement
3595 Description: allow number of MDS service threads to be changed at module load
3596 Details : It is now possible to change the number of MDS service threads
3597 running. Adding "options mds mds_num_threads={N}" to the MDS's
3598 /etc/modprobe.conf will set the number of threads for the next
3599 time Lustre is restarted (assuming the "mds" module is also
3600 reloaded at that time). The default number of threads will
3601 stay the same, 32 for most systems.
3606 Description: OST crash if filesystem is unformatted or corrupt
3607 Details : If an OST is started on a device that has never been formatted
3608 or if the filesystem is corrupt and cannot even mount then the
3609 error handling cleanup routines would dereference a NULL pointer.
3614 Description: NULL pointer deref in llap_from_page.
3615 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3616 incorrect) mapping if the page was truncated/reclaimed while it was
3617 searched for. Check for this condition and skip such pages when
3618 doing readahead. Introduce extra check to llap_from_page() to
3619 verify page->mapping->host is non-NULL (so page is not anonymous).
3622 Frequency : Sometimes when using sys_sendfile
3624 Description: "page not covered by a lock" warnings from ll_readpage
3625 Details : sendfile called ll_readpage without right page locks present.
3626 Now we introduced ll_file_sendfile that does necessary locking
3627 around call to generic_file_sendfile() much like we do in
3631 Frequency : with certain MDS communication failures at client mount time
3633 Description: NULL pointer deref after failed client mount
3634 Details : a client connection request may delayed by the network layer
3635 and not be sent until after the PTLRPC layer has timed out the
3636 request. If the client fails the mount immediately it will try
3637 to clean up before the network times out the request. Add a
3638 reference from the request import to the obd device and delay
3639 the cleanup until the network drops the request.
3642 Frequency : occasionally during client (re)connect
3644 Description: assertion failure during client (re)connect
3645 Details : processing a client connection request may be delayed by the
3646 client or server longer than the client connect timeout. This
3647 causes the client to resend the connection request. If the
3648 original connection request is replied in this interval, the
3649 client may trip an assertion failure in ptlrpc_connect_interpret()
3650 which thought it would be the only running connect process.
3653 Frequency : only with obd_echo servers and clients that are rebooted
3655 Description: kernel BUG accessing uninitialized data structure
3656 Details : When running an obd_echo server it did not start the ping_evictor
3657 thread, and when a client was evicted an uninitialized data
3658 structure was accessed. Start the ping_evictor in the RPC
3659 service startup instead of the OBD startup.
3661 Severity : enhancement
3662 Bugzilla : 10193 (patchless)
3663 Description: Remove dependency on various unexported kernel interfaces.
3664 Details : No longer need reparent_to_init, exit_mm, exit_files,
3665 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3668 Frequency : rare (only users of deprecated and unsupported LDAP config)
3670 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3671 Details : LDAP apparently contains 'lustreName' attributes instead of
3672 'name'. A simple remapping of the name is sufficient.
3675 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3677 Description: waiting_locks_callback trips kernel BUG if client is evicted
3678 Details : Running with the dump_on_timeout debug flag turned on makes
3679 it possible that the waiting_locks_callback() can try to dump
3680 the Lustre kernel debug logs from an interrupt handler. Defer
3681 this log dumping to the expired_lock_main() thread.
3683 Severity : enhancement
3685 Description: Support NFS exporting on 2.6 kernels.
3686 Details : Implement non-rawops metadata methods for NFS server to use without
3687 changing NFS server code.
3690 Frequency : very rare (synthetic metadata workload only)
3692 Description: two racing renames might cause an MDS thread to deadlock
3693 Details : Running the "racer" program may cause one MDS thread to rename
3694 a file from being the source of a rename to being the target of
3695 a rename at exactly the same time that another thread is doing
3696 so, and the second thread has already enqueued these locks after
3697 doing a lookup of the target and is trying to relock them in
3698 order. Ensure that we don't try to re-lock the same resource.
3701 Frequency : only very large systems with liblustre clients
3703 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3704 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3705 all OSTs in parallel. This allows the eviction of stale liblustre
3706 clients to proceed much faster than if they were done in series,
3707 and also offers similar improvements for other set_info RPCs.
3712 Description: excessive CPU usage during initial read phase on client
3713 Details : During the initial read phase on a client, it would agressively
3714 retry readahead on the file, consuming too much CPU and impacting
3715 performance (since 1.4.5.8). Improve the readahead algorithm
3716 to avoid this, and also improve some other common cases (read
3717 of small files in particular, where "small" is files smaller than
3718 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3723 Description: MDS crash when receiving packet with unknown intent.
3724 Details : Do not LBUG in unknown intent case, just return -EFAULT
3726 Severity : enhancement
3727 Bugzilla : 9293, 9385
3728 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3729 Details : Do not serialize getattr (non-intent version) and statfs.
3732 Frequency : occasional, when OST network is overloaded/intermittent
3734 Description: client evicted by OST after bulk IO timeout
3735 Details : If a client sends a bulk IO request (read or write) the OST
3736 may evict the client if it is unresposive to its data GET/PUT
3737 request. This is incorrect if the network is overloaded (takes
3738 too long to transfer the RPC data) or dropped the OST GET/PUT
3739 request. There is no need to evict the client at all, since
3740 the pinger and/or lock callbacks will handle this, and the
3741 client can restart the bulk request.
3744 Frequency : Always when mmapping file with no objects
3746 Description: client crashes when mmapping file with no objects
3747 Details : Check that we actually have objects in a file before doing any
3748 operations on objects in ll_vm_open, ll_vm_close and
3754 Description: Request leak when working with deleted CWD
3755 Details : Introduce advanced request refcount tracking for requests
3756 referenced from lustre intent.
3758 Severity : Enhancement
3760 Description: Cache open file handles on client.
3761 Details : MDS now will return special lock along with openhandle, if
3762 requested and client is allowed to hold openhandle, even if unused,
3763 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3764 closing files for every read/write openration.
3766 Severity : Enhancement
3768 Description: Cache open negative dentries on client when possible.
3769 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3770 negative dentries on lock revocation.
3775 Description: Remounting a client read-only wasn't possible with a zconf mount
3776 Details : It wasn't possible to remount a client read-only with llmount.
3778 Severity : enhancement
3779 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3780 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3781 MPICH is included to add Lustre-specific ADIO interfaces.
3782 This is based closely on the UFS ADIO layer and only differs
3783 in file creation, in order to allow the OST striping to be set.
3784 This is user-contributed code and not supported by CFS.
3789 Description: extended inode attributes (immutable, append-only) work improperly
3790 when 2.4 and 2.6 kernels are used on client/server or vice versa
3791 Details : Introduce kernel-independent values for these flags.
3793 Severity : enhancement
3796 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3797 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3798 granularity. Now, Fractional MB granularity is be supported,
3799 this is very useful for low memory system.
3801 Severity : enhancement
3803 Description: Getattr by fid
3804 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3805 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3806 have locks and avoids getattr rpc requests for it on every stat.
3809 Frequency : Always, for filesystems larger than 2TB
3811 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3812 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3813 filesystems larger than 2TB to be mounted with mballoc enabled.
3816 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3818 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3819 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3820 filesystems larger than 2TB is mounted then the kernel will
3821 silently corrupt the start of the filesystem. CONFIG_LBD is
3822 enabled for all CFS-supported kernels, but the possibility of
3823 this happening with a modified kernel config exists.
3825 Severity : enhancement
3827 Description: add client O_DIRECT support for 2.6 kernels
3828 Details : It is now possible to do O_DIRECT reads and writes to files
3829 in the Lustre client mountpoint on 2.6 kernel clients.
3831 Severity : enhancement
3833 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3834 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3835 parallel, not waiting for response from every OST before sending
3836 a rpc to the next OST.
3841 Description: setattr vs write race when updating file timestamps
3842 Details : Client processes that update a file timestamp into the past
3843 right after writing to the file (e.g. tar) it is possible that
3844 the updated file modification time can be reset to the current
3845 time due to a race between processing the setattr and write RPC.
3847 Severity : enhancement
3849 Description: Bring 'lfs find' closer in line with regular Linux find.
3850 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3851 -print0 options and obtains all the needed info through the lustre
3854 Severity : enhancement
3856 Description: support up to 1024 configured devices on one node
3857 Details : change obd_dev array from statically allocated to dynamically
3858 allocated structs as they are first used to reduce memory usage
3863 Description: Flush dirty partially truncated pages during truncate
3864 Details : Immediatelly flush partially truncated pages in filter_setattr,
3865 this way we completely avoid having any pages in page cache on OST
3866 and can retire ugly workarounds during writes to flush such pages.
3871 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3872 Details : protect inode from truncation within vfs_unlink() context
3873 just take a reference before calling vfs_unlink() and release it
3874 when parent's i_sem is free.
3879 Description: last_id value checked outside lock on OST caused LASSERT failure
3880 Details : If there were multiple MDS->OST object precreate requests in
3881 flight, it was possible that the OST's last object id was checked
3882 outside a lock and incorrectly tripped an assertion. Move checks
3883 inside locks, and discard old precreate requests.
3886 Frequency : always, if extents are used on OSTs
3888 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3889 Details : OSTs using the extents format would not zero the high 16 bits of
3890 the index physical block number. This is not a problem for any
3891 OST filesystems smaller than 16TB, and no kernels support ext3
3892 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3893 new/modified files) and can be fixed for existing filesystems
3894 with e2fsprogs-1.39-cfs1.
3899 Description: import connection selection may be incorrect if timer wraps
3900 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3901 import connections to be ignored if the 32-bit jiffies counter
3902 wraps. Use a 64-bit jiffies counter.
3905 Frequency : very large clusters immediately after boot
3907 Description: LNET request buffers exhausted under heavy short-term load
3908 Details : If a large number of client requests are generated on a service
3909 that has previously never seen so many requests it is possible
3910 that the request buffer growth cannot keep up with the spike in
3911 demand. Instead of dropping incoming requests, they are held in
3912 the LND until the RPC service can accept more requests.
3915 Frequency : Sometimes during replay
3917 Description: Assertion failure in ll_local_open after replay.
3918 Details : If replay happened on an open request reply before we were able
3919 to set replay handler, reply will become not swabbed tripping the
3920 assertion in ll_local_open. Now we set the handler right after
3921 recognising of open request
3924 Frequency : very rare
3926 Description: kernel reports "badness in vsnprintf"
3927 Details : Reading from the "recovery_status" /proc file in small chunks
3928 may cause a negative length in lprocfs_obd_rd_recovery_status()
3929 call to vsnprintf() (which is otherwise harmless). Exit early
3930 if there is no more space in the output buffer.
3932 Severity : enhancement
3934 Description: clear OBD RPC statistics by writing to them
3935 Details : It is now possible to clear the OBD RPC statistics by writing
3936 to the "stats" file.
3941 Description: Client mtime is not the same on different clients after utimes
3942 Details : In some cases, the client was using the utimes() syscall on
3943 a file cached on another node. The clients now validate the
3944 ctime from the MDS + OSTs to determine which one is right.
3949 Description: Inability to activate failout mode
3950 Details : lconf script incorrectly assumed that in pythong string's numeric
3951 value is used in comparisons.
3954 Frequency : always with multiple stripes per file
3956 Description: Inefficient object allocation for mutli-stripe files
3957 Details : When selecting which OSTs to stripe files over, for files with
3958 a stripe count that divides evenly into the number of OSTs,
3959 the MDS is always picking the same starting OST for each file.
3960 Return the OST selection heuristic to the original design.
3965 Description: mount failures may take full timeout to return an error
3966 Details : Under some heavy load conditions it is possible that a
3967 failed mount can wait for the full obd_timeout interval,
3968 possibly several minutes, before reporting an error.
3969 Instead return an error as soon as the status is known.
3971 Frequency : quota enabled and large files being deleted
3973 Description: releasing more than 4GB of quota at once hangs OST
3974 Details : If a user deletes more than 4GB of files on a single OST it
3975 will cause the OST to spin in an infinite loop. Release
3976 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3981 Description: statfs data retrieved from /proc may be stale or zero
3982 Details : When reading per-device statfs data from /proc, in the
3983 {kbytes,files}_{total,free,avail} files, it may appear
3984 as zero or be out of date.
3987 Frequency : systems with MD RAID1 external journal devices
3989 Description: lconf's call to blkid is confused by RAID1 journal devices
3990 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3991 of returning all block devices that match the journal UUID.
3994 Frequency : always, for aggregate stripe size over 4GB
3996 Description: assertion fails when trying to use 4GB stripe size
3997 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3998 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4001 Frequency : always on ppc64
4003 Description: the first write on an ext3 filesystem with mballoc got stuck
4004 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4005 perform endianness conversion.
4007 ------------------------------------------------------------------------------
4009 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4011 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4012 INTEROPERATE with older versions automatically. Please read the
4013 user documentation before upgrading any part of a live system.
4014 * WARNING: Lustre networking configuration changes are required with
4015 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4018 * Support for kernels:
4019 2.6.9-22.0.2.EL (RHEL 4)
4020 2.6.5-7.244 (SLES 9)
4021 2.6.12.6 vanilla (kernel.org)
4024 Severity : enhancement
4025 Bugzilla : 7981/8208
4026 Description: Introduced Lustre Networking (LNET)
4027 Details : LNET is new networking infrastructure for Lustre, it includes
4028 a reorganized network configuration mode (see the user
4029 documentation for full details) as well as support for routing
4030 between different network fabrics. Lustre Networking Devices
4031 (LNDs) for the supported network fabrics have also been
4032 created for this new infrastructure.
4034 Severity : enhancement
4035 Description: Introduced Access control lists
4036 Details : clients can set ACLs on files and directories in order to have
4037 more fine-grained permissions than the standard Unix UGO+RWX.
4038 The MDS must be started with the "-o acl" mount option.
4040 Severity : enhancement
4041 Description: Introduced filesystem quotas
4042 Details : Administrators may now establish per-user quotas on the
4045 Severity : enhancement
4047 Description: Configuration change for the XT3
4048 The PTLLND is now used to run Lustre over Portals on the XT3
4049 The configure option(s) --with-cray-portals are no longer used.
4050 Rather --with-portals=<path-to-portals-includes> is used to
4051 enable building on the XT3. In addition to enable XT3 specific
4052 features the option --enable-cray-xt3 must be used.
4057 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4058 Details : A race between allocating a new llcd and re-getting the llcd_lock
4059 allowed another thread to grab newly-allocated llcd.
4061 Severity : enhancement
4063 Description: 2.6 OST async journal commit and locking fix to improve performance
4064 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4065 2.6 kernels are now async as they already were in 2.4 kernels so
4066 that they can commit concurrently with the network bulk transfer.
4067 For block-allocated files the filter allocation semaphore is held
4068 to avoid filesystem fragmentation during allocation. BKL lock
4069 removed for 2.6 xattr operations where it is no longer needed.
4074 Description: lconf incorrectly determined whether two IP networks could talk
4075 Details : In some more complicated routing and multiple-network
4076 configurations, lconf will avoid trying to make a network
4077 connection to a disjoint part of the IP space. It was doing the
4078 math incorrectly for one set of cases.
4083 Description: Fix for potential infinite loop processing records in an llog.
4084 Details : If an llog record is corrupted/zeroed, it is possible to loop
4085 forever in llog_process(). Validate the llog record length
4086 and skip the remainder of the block on error.
4089 Frequency : occasional (liblustre only)
4091 Description: liblustre could not open files whose last component is a symlink
4092 Details : sysio_path_walk() would incorrectly pass the open intent to
4093 intermediate path components.
4096 Frequency : rare (liblustre only with non-standard tuning)
4097 Bugzilla : 7201 (7350)
4098 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4099 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4100 after it has been released, only do this for referenced locks
4102 Severity : enhancement
4104 Description: specifying an (invalid) directory default stripe_size of -1
4105 would reset the directory default striping
4106 Details : stripe_size -1 was used internally to signal directory stripe
4107 removal, now use "all default" to signal dir stripe removal
4108 as a directory striping of "all default" is not useful
4111 Frequency : common for large clusters running liblustre clients
4113 Description: doing an ls when liblustre clients are running is slow
4114 Details : sending a glimpse AST to a liblustre client waits for every AST
4115 to time out, as liblustre clients will not respond. Since they
4116 cannot cache data we refresh the OST lock LVB from disk instead.
4118 Severity : enhancement
4120 Description: doing an ls at the same time as file IO can be slow
4121 Details : enqueue and other "small" requests can be blocked behind many
4122 large IO requests. Create a new OST IO portal for non-IO
4123 requests so they can be processed faster.
4126 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4128 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4129 created file with mode 000
4130 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4131 allow an ftruncate() as a non-root user to a file with mode 000.
4132 Lustre now respects this flag to disable mode checks when
4133 truncating a file owned by the user
4136 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4139 Description: Revoking locks from clients that went dead or catatonic might take
4141 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4142 cancellation of such locks instant on servers without waiting for
4143 any reply from clients. Clients drops these locks when cancel
4144 notification from server is received without replying.
4147 Frequency : liblustre-only, when liblustre client dies or becomes busy
4149 Description: Doing ls on Linux clients can take a long time with active
4151 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4152 granting such locks to it in the first place if possible. Locks
4153 are taken by proxy on the OST during the read or write and
4154 dropped immediately afterward. Add connect flags handling, do
4155 not grant locks to liblustre clients for glimpse ASTs.
4157 Severity : enhancement
4159 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4160 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4161 sequential reads and behaves badly with applications doing random
4162 reads. Improve it by reading ahead at least read region, and
4163 avoiding excessive large RPC for small reads.
4165 Severity : enhancement
4167 Description: Creating more than 1000 files for a single job may cause a load
4168 imbalance on the OSTs if there are also a large number of OSTs.
4169 Details : qos_prep_create() uses an OST index reseed value that is an
4170 even multiple of the number of available OSTs so that if the
4171 reseed happens in the middle of the object allocation it will
4172 still utilize the OSTs as uniformly as possible.
4177 Description: OST or MDS may oops in ping_evictor_main()
4178 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4179 but doesn't restart at beginning of obd_exports_timed list
4182 Severity : enhancement
4184 Description: improve by-nid export eviction on the MDS and OST
4185 Details : allow multiple exports with the same NID to be evicted at one
4186 time without re-searching the exports list.
4189 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4191 Description: MDS may oops in groups_free()
4192 Details : in rare race conditions a newly allocated group_info struct is
4193 freed again, and this can be NULL. The 2.4 compatibility code
4194 for groups_free() checked for a NULL pointer, but 2.6 did not.
4197 Frequency : common for liblustre clients doing little filesystem IO
4198 Bugzilla : 9352, 7313
4199 Description: server may evict liblustre clients accessing contended locks
4200 Details : if a client is granted a lock or receives a completion AST
4201 with a blocking AST already set it would not reply to the AST
4202 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4205 Frequency : lfs setstripe, only systems with more than 160 OSTs
4207 Description: unable to set striping with a starting offset beyond OST 160
4208 Details : llapi_create_file() incorrectly limited the starting stripe
4209 index to the maximum single-file stripe count.
4212 Frequency : LDAP users only
4214 Description: lconf did not handle in-kernel recovery with LDAP properly
4215 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4217 Severity : enhancement
4219 Description: bind OST threads to NUMA nodes to improve performance
4220 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4221 node and do their allocations there to localize memory access
4223 Severity : enhancement
4225 Description: llmount can determine client NID directly from Myrinet (GM)
4226 Details : the client NID code from gmnalnid was moved directly into
4227 llmount, removing the need to use this or specifying the
4228 client NID explicitly when mounting GM clients with zeroconf
4231 Frequency : if client is started with down MDS
4233 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4234 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4235 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4236 ptlrpc_import_delayed_req() because we didn't check if the
4237 request was interrupted, and we also didn't break out of the
4238 event loop if there was no timeout
4243 Description: data loss during non-page-aligned writes to a single file from
4244 both multiple nodes and multiple threads on one node at same time
4245 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4246 inconsistency led to false short-reads, that were cached and later
4247 used by ->prepare_write() to fill in partially written page,
4248 leading to data loss.
4251 Frequency : always, if lconf --abort_recovery used
4253 Description: lconf --abort_recovery fails with 'Operation not supported'
4254 Details : lconf was attempting to abort recovery on the MDT device and not
4257 Severity : enhancement
4259 Description: remove cleanup logs
4260 Details : replace lconf-generated cleanup logs with lustre internal
4261 cleanup routines. Eliminates the need for client-cleanup and
4264 Severity : enhancement
4266 Description: add support for EAs (user and system) on lustre filesystems
4267 Details : it is now possible to store extended attributes in the Lustre
4268 client filesystem, and with the user_xattr mount option it
4269 is possible to allow users to store EAs on their files also
4271 Severity : enhancement
4273 Description: Add possibility (config option) to show minimal available OST free
4275 Details : When compiled with --enable-mindf configure option, statfs(2)
4276 (and so, df) will return least minimal free space available from
4277 all OSTs as amount of free space on FS, instead of summary of
4278 free spaces of all OSTs.
4280 Severity : enhancement
4282 Description: do not expand extent locks acquired on OST-side
4283 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4284 by server: they are not cached anyway.
4287 Frequency : when mmap is used/binaries executed from Lustre
4289 Description: Unmmap pages before throwing them away from read cache.
4290 Details : llap_shrink cache now attempts to unmap pages before discarding
4291 them (if unmapping failed - do not discard). SLES9 kernel has
4292 extra checks that trigger if this unmapping is not done first.
4297 Description: lconf didn't resolve symlinks before checking to see whether a
4298 given mountpoint was already in use
4301 Frequency : when migrating failover services
4302 Bugzilla : 6395, 9514
4303 Description: When migrating a subset of services from a node (e.g. failback
4304 from a failover service node) the remaining services would
4305 time out and evict clients.
4306 Details : lconf --force (implied by --failover) sets the global obd_timeout
4307 to 5 seconds in order to quickly disconnect, but this caused
4308 other RPCs to time out too quickly. Do not change the global
4309 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4311 Severity : enhancement
4312 Frequency : if MDS is started with down OST
4313 Bugzilla : 9439,5706
4314 Description: Allow startup/shutdown of an MDS without depending on the
4315 availability of the OSTs.
4316 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4317 Add appropriate locking and lov-osc refcounts for safe
4318 cleaning. Add osc abort_inflight calls in case the
4319 synchronize never started.
4322 Frequency : occasional (Cray XT3 only)
4324 Description: root not authorized to access files in CRAY_PORTALS environment
4325 Details : The client process capabilities were not honoured on the MDS in
4326 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4327 been accessed by an authorized user then root was able to access
4328 the file on the local client also. The root user capabilities
4329 are now allowed on the MDS, as this environment has secure UID.
4332 Frequency : occasional
4334 Description: ldiskfs "too long searching" message happens too often
4335 Details : A debugging message (otherwise harmless) prints too often on
4336 the OST console. This has been reduced to only happen when
4337 there are fragmentation problems on the filesystem.
4342 Description: Division by zero in statfs when all OSCs are inactive
4343 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4344 lov_statfs divides by value returned by lov_get_stripecnt().
4348 Bugzilla : 9489, 3273
4349 Description: First write from each client to each OST was only 4kB in size,
4350 to initialize client writeback cache, which caused sub-optimal
4351 RPCs and poor layout on disk for the first writen file.
4352 Details : Clients now request an initial cache grant at (re)connect time
4353 and so that they can start streaming writes to the cache right
4354 away and always do full-sized RPCs if there is enough data.
4355 If the OST is rebooted the client also re-establishes its grant
4356 so that client cached writes will be honoured under the grant.
4361 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4362 Details : Now I/O RPCs go to different portal number and (presumably) fast
4363 lock requests (and glimses) and other RPCs get their own service
4364 threads pool that should be able to service those RPCs
4367 Severity : enhancement
4369 Description: Ability to exchange lustre version between client and servers and
4370 issue warnings at client side if client is too old. Also for
4371 liblustre clients there is ability to refuse connection of too old
4373 Details : New 'version' field is added to connect data structure that is
4374 filled with version info. That info is later checked by server and
4378 Frequency : rare, liblustre only.
4379 Bugzilla : 9296, 9581
4380 Description: Two simultaneous writes from liblustre at offset within same page
4381 might proceed at the same time overwriting eachother with stale
4383 Details : I/O lock withing llu_file_prwv was released too early, before data
4384 actually was hitting the wire. Extended lock-holding time until
4385 server acknowledges receiving data.
4388 Frequency : extremely rare. Never observed in practice.
4390 Description: avoid generating lustre_handle cookie of 0.
4391 Details : class_handle_hash() generates handle cookies by incrementing
4392 global counter, and can hit 0 occasionaly (this is unlikely, but
4393 not impossible, because initial value of cookie counter is
4394 selected randonly). Value of 0 is used as a sentinel meaning
4395 "unassigned handle" --- avoid it. Also coalesce two critical
4396 sections in this function into one.
4398 Severity : enhancement
4400 Description: allow liblustre clients to delegate truncate locking to OST
4401 Details : To avoid overhead of locking, liblustre client instructs OST to
4402 take extent lock in ost_punch() on client's behalf. New connection
4403 flag is added to handle backward compatibility.
4405 Severity : enhancement
4406 Bugzilla : 4928, 7341, 9758
4407 Description: allow number of OST service threads to be specified
4408 Details : a module parameter allows the number of OST service threads
4409 to be specified via "options ost ost_num_threads={N}" in the
4410 OSS's /etc/modules.conf or /etc/modprobe.conf.
4414 Bugzilla : 6146, 9635, 9895
4415 Description: servers crash with bad pointer in target_handle_connect()
4416 Details : In rare cases when a client is reconnecting it was possible that
4417 the connection request was the last reference for that export.
4418 We would temporarily drop the export reference and get a new
4419 one, but this may have been the last reference and the export
4420 was just destroyed. Get new reference before dropping old one.
4422 Severity : enhancement
4423 Frequency : if client is started with failover MDS
4425 Description: Allow multiple MDS hostnames in the mount command
4426 Details : Try to read the configuration from all specified MDS
4427 hostnames during a client mount in case the "primary"
4430 Severity : enhancement
4432 Description: Stop sending data to evicted clients as soon as possible.
4433 Details : Check if the client we are about to send or are sending data to
4434 was evicted already. (Check is done every second of waiting,
4435 for which l_wait_event interface was extended to allow checking
4436 of exit condition at specified intervals).
4439 Frequency : rare, normally only when NFS exporting is done from client
4441 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4443 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4446 Frequency : always, if srand() is called before liblustre initialization
4448 Description: Liblustre uses system PRNG disturbing its usage by user application
4449 Details : Introduce internal to lustre fast and high-quality PRNG for
4450 lustre usage and make liblustre and some other places in generic
4451 lustre code to use it.
4453 Severity : enhancement
4454 Bugzilla : 9477, 9557, 9870
4455 Description: Verify that the MDS configuration logs are updated when xml is
4456 Details : Check if the .xml configuration logs are newer than the config
4457 logs stored on the MDS and report an error if this is the case.
4458 Request --write-conf, or allow starting with --old_conf.
4460 Severity : enhancement
4462 Description: Handle symlinks in the path when checking if Lustre is mounted.
4463 Details : Resolve intermediate symlinks when checking if a client has
4464 mounted a filesystem to avoid duplicate client mounts.
4469 Description: lconf can hit an error exception but still return success.
4470 Details : The lconf command catches the Command error exception at the top
4471 level script context and will exit with the associated exit
4472 status, but doesn't ensure that this exit status is non-zero.
4477 Description: failure of ptlrpc thread startup can cause oops
4478 Details : Starting a ptlrpc service thread can fail if there are a large
4479 number of threads or the server memory is very fragmented.
4480 Handle this without oopsing.
4483 Frequency : always, only if liblustre and non-default acceptor port was used
4485 Description: liblustre cannot connect to servers with non-default acceptor port
4486 Details : tcpnal_set_default_params() was not called and was therefore
4487 ignoring the environment varaible TCPNAL_PORT, as well as other
4488 TCPNAL_ environment variables
4493 Description: two objects could be created on the same OST for a single file
4494 Details : If an OST is down, in some cases it was possible to create two
4495 objects on a single OST for a single file. No problems other
4496 than potential performance impact and spurious error messages.
4500 Bugzilla : 5681, 9562
4501 Description: Client may oops in ll_unhash_aliases
4502 Details : Client dcache may become inconsistent in race condition.
4503 In some cases "getcwd" can fail if the current directory is
4509 Description: Inode refcounting problems in NFS export code
4510 Details : link_raw functions used to call d_instantiate without obtaining
4511 extra inode reference first.
4515 Bugzilla : 9942, 9903
4516 Description: Referencing freed requests leading to crash, memleaks with NFS.
4517 Details : We used to require that call to ll_revalidate_it was always
4518 followed by ll_lookup_it. Also with revalidate_special() it is
4519 possible to call ll_revalidate_it() twice for the same dentry
4520 even if first occurence returned success. This fix changes semantic
4521 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4522 reference on a request referred from the intent.
4523 ll_intent_release() then releases such a request.
4526 Frequency : rare, normally benchmark loads only
4528 Description: unlinked inodes were kept in memory on the client
4529 Details : If a client is repeatedly creating and unlinking files it
4530 can accumulate a lot of stale inodes in the inode slab cache.
4531 If there is no other client load running this can cause the
4532 client node to run out of memory. Instead flush old inodes
4533 from client cache that have the same inode number as a new inode.
4536 Frequency : SLES9 2.6.5 kernel and long filenames only
4537 Bugzilla : 9969, 10379
4538 Description: utime reports stale NFS file handle
4539 Details : SLES9 uses out-of-dentry names in some cases, which confused
4540 the lustre dentry revalidation. Change it to always use the
4544 Frequency : rare, unless heavy write-truncate concurrency is continuous
4545 Bugzilla : 4180, 6984, 7171, 9963, 9331
4546 Description: OST becomes very slow and/or deadlocked during object unlink
4547 Details : filter_destroy() was holding onto the parent directory lock
4548 while truncating+unlinking objects. For very large objects this
4549 may block other threads for a long time and slow overall OST
4550 responsiveness. It may also be possible to get a lock ordering
4551 deadlock in this case, or run out of journal credits because of
4552 the combined truncate+unlink. Solution is to do object truncate
4553 first in one transaction without parent lock, and then do the
4554 final unlink in a new transaction with the parent lock. This
4555 reduces the lock hold time dramatically.
4558 Frequency : rare, 2.4 kernels only
4560 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4561 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4562 before unmounting their filesystem in order to avoid deadlock.
4563 The kernel_locked() function in 2.4 kernels only checks whether
4564 the kernel lock is held, not whether it is this process that is
4565 holding it as 2.6 kernels do.
4570 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4571 Details : The client ptlrpc code may be trying to reconnect to a down
4572 server before a previous connection attempt has timed out.
4573 Increase the reconnect interval to be longer than the connection
4574 timeout interval to avoid sending duplicate connections to
4578 Frequency : echo_client brw_test command
4580 Description: fix echo_client to work with OST preallocated code
4581 Details : OST preallocation code (5137) didn't take echo_client IO path
4582 into account: echo_client calls filter methods outside of any
4583 OST thread and, hence, there is no per-thread preallocated
4584 pages and buffers to use. Solution: hijack pga pages for IO. As
4585 a byproduct, this avoids unnecessary data copying.
4589 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4590 Description: Client can oops in mdc_commit_close() after open replay
4591 Details : It was possible for the MDS to return an open request with no
4592 transaction number in mds_finish_transno() if the client was
4593 evicted, but without actually returning an error. Clients
4594 would later try to replay that open and may trip an assertion
4595 Simplify the client close codepath, and always return an error
4596 from the MDS in case the open is not successful.
4599 Frequency : rare, 2.6 OSTs only
4601 Description: OST may deadlock under high load on fragmented files
4602 Details : If there was a heavy load and highly-fragmented OST filesystems
4603 it was possible to have all the OST threads deadlock waiting on
4604 allocation of biovecs, because the biovecs were not released
4605 until the entire RPC IO was completed. Instead, release biovecs
4606 as soon as they are complete to ensure forward IO progress.
4608 Severity : enhancement
4610 Description: Support for specifying external journal device at mount
4611 Details : If an OST or MDS device is formatted with an external journal
4612 device, this device major/minor is stored in the ext3 superblock
4613 and may not be valid for failover. Allow detecting and
4614 specifying the external journal at mount time.
4619 Description: Mounting an MDS with pending unlinked files may cause oops
4620 Details : target_finish_recovery() calls mds_postrecov() which returned
4621 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4622 considers this an error and immediately begins cleaning up the
4623 lov, just after starting the mds_lov process
4625 Severity : enhancement
4627 Description: Implement 'lfs df' to report actual free space on per-OST basis
4628 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4629 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4630 usage of objects; '-h' to report in human readable format.
4632 ------------------------------------------------------------------------------
4634 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4641 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4642 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4643 at startup during orphan unlinking. Instead, skip buddy update
4644 before setup, as it will be regenerated after recovery is complete.
4649 Description: If an OST is inactive, its locks might reference stale inodes.
4650 Details : lov_change_cbdata() must iterate over all namespaces, even if
4651 they are inactive to clear inode references from the lock.
4653 Severity : enhancement
4654 Frequency : occasional, if non-standard max_dirty_mb used
4656 Description: Client will block write RPCs if not enough grant
4657 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4658 then the client will block writes while waiting for another RPC
4659 to complete instead of consuming its dirty limit. With change
4660 we get improved performance when max_dirty_mb is small.
4662 Severity : enhancement
4663 Bugzilla : 3389, 6253
4664 Description: Add support for supplementary groups on the MDS.
4665 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4666 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4667 lookups for user supplementary groups into a cache.
4671 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4672 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4673 for liblustre O_EXCL is enforced by MDS.
4677 Description: readdir never returns NULL in liblustre.
4678 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4679 dentry in struct dirent.
4683 Frequency : liblustre only, depends on application IO pattern
4684 Description: liblustre clients evicted if not contacting servers
4685 Details : Don't put liblustre clients into the ping_evictor list, so
4686 they will not be evicted by the pinger ever.
4688 Severity : enhancement
4690 Description: Add ability to evict clients by NID from MDS.
4691 Details : By echoing "nid:$NID" string into
4692 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4693 $NID would be instantly evicted from this MDS and from all active
4694 OSTs connected to it.
4698 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4699 Details : lookup_it_finish() used to query file size from OSTs that was not
4704 Description: service threads change working directory to that of init
4705 Details : Starting lustre service threads may pin the working directory
4706 of the parent thread, making that filesystem busy. Threads
4707 now change to the working directory of init to avoid this.
4711 Frequency : during shutdown only
4712 Description: shutdown with a failed MDS or OST can cause unmount to hang
4713 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4718 Frequency : 2.6 only
4719 Description: chmod/chown may include an extra supplementary group
4720 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4721 supplementary group and if none is specified this is used.
4724 Bugzilla : 5479 (6816)
4726 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4727 Details : If lookup is in progress on a file that is unlinked we might try
4728 to revalidate the inode and fail in revalidate after lookup is
4729 complete and ll_file_open() enqueues the open again but
4730 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4733 Frequency : always, if lconf --abort_recovery used
4735 Description: lconf --abort_recovery fails with 'Operation not supported'
4736 Details : lconf was attempting to abort recovery on the MDT device and not
4739 ------------------------------------------------------------------------------
4741 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4746 Frequency : rare (only unsupported configurations with a node running as an
4748 Bugzilla : 6514, 5137
4749 Description: Mounting a Lustre file system on a node running as an OST could
4751 Details : OSTs now preallocates memory needed to write out data at
4752 startup, instead of when needed, to avoid having to
4753 allocate memory in possibly low memory situations.
4754 Specifically, if the file system is mounted on on OST,
4755 memory pressure could force it to try to write out data,
4756 which it needed to allocate memory to do. Due to the low
4757 memory, it would be unable to do so and the node would
4758 become unresponsive.
4760 Severity : enhancement
4762 Description: Addition of lconf --service command line option
4763 Details : lconf now accepts a '--service <arg>' option, which is
4764 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4766 Severity : enhancement
4768 Description: Failover mode is now the default for OSTs.
4769 Details : By default, OSTs will now run in failover mode. To return to
4770 the old behaviour, add '--failout' to the lmc line for OSTs.
4772 Severity : enhancement
4774 Description: Health checks are now provided for MDS and OSTs
4775 Details : Additional detailed health check information on MSD and OSTs
4776 is now provided through the procfs health_check value.
4779 Frequency : occasional, depends on IO load
4781 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4782 after numerous create/delete cycles
4783 Details : The ext3 inode allocation policy would not allocate new inodes
4784 very well on the OSTs because there are no new directories
4785 being created. Instead we look for groups with free space if
4786 the parent directories are nearly full.
4791 Description: Network or server problems during mount may cause partially
4792 mounted clients instead of returning an error.
4793 Details : The config llog parsing code may overwrite the error return
4794 code during mount error handling, returning success instead
4800 Description: MDS can fail to allocate large reply buffers
4801 Details : After long uptimes the MDS can fail to allocate large reply
4802 buffers (e.g. zconf client mount config records) due to memory
4803 fragmentation or consumption by the buffer cache. Preallocate
4804 some large reply buffers so that these replies can be sent even
4805 under memory pressure.
4809 Frequency : rare (liblustre)
4810 Description: fsx running with liblustre complained that using truncate() to
4811 extend the file doesn't work. This patch corrects that issue.
4812 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4813 ATTR_SIZE and lsm use in llu_setattr_raw.
4817 Frequency : rare, only 2.6 kernels
4818 Description: Unusual file access patterns on the MDS may result in inode
4819 data being lost in very rare circumstances.
4820 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4821 code in the 2.6 kernel caused the inode and/or EA data not to
4822 be read from disk, causing single-file corruption.
4826 Frequency : rare, only 2.6 filesystems using extents
4827 Description: Heavy concurrent write and delete load may cause data corruption.
4828 Details : It was possible under high-load situations to have an extent
4829 metadata block in the block device cache from a just-unlinked
4830 file overwrite a newly-allocated data block. We now unmap any
4831 metadata buffers that alias just-allocated data blocks.
4835 Frequency : filesystems with default stripe_count larger than 77
4836 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4837 Details : lconf specifies an inode size of 4096 bytes when the default
4838 stripe_count is larger than 77. This conflicts with the default
4839 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4840 this case to avoid pinning too much memory for large EAs.
4842 ------------------------------------------------------------------------------
4844 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4849 Frequency : rare (extremely heavy IO load with hundreds of clients)
4851 Description: Client is evicted, gets IO error writing to file
4852 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4853 caused clients to be evicted for AST timeouts. The fixes in
4854 bug 5192 mean we no longer need to have such short AST timeouts
4855 so ldlm_timeout has been increased.
4858 Frequency : occasional during --force or --failover shutdown under load
4859 Bugzilla : 5949, 4834
4860 Description: Server oops/LBUG if stopped with --force or --failover under load
4861 Details : a collection of import/export refcount and cleanup ordering
4862 issues fixed for safer force cleanup
4865 Frequency : only filesystems larger than 120 OSTs
4866 Bugzilla : 5990, 6223
4867 Description: lfs getstripe would oops on a very large filesystem
4868 Details : lov_getconfig used kfree on vmalloc'd memory
4871 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4872 Bugzilla : 6242, 6243
4873 Description: reading from files that had been truncated to a non-zero size
4874 but never opened returned no data
4875 Details : ll_file_read() reads zeros from no-object files to EOF
4880 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4881 Details : There's little harm in aborting MDS/OSS recovery and letting it
4882 try again, so I removed the LASSERT and return an error instead.
4884 Severity : enhancement
4886 Description: New debugging infrastructure for tracking down data corruption
4887 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4888 (b) cover more of the client-side code path, and (c) try to narrow
4889 down where problems occurred
4893 Bugzilla : 3819, 4364, 4397, 6313
4894 Description: Racing close and eviction MDS could cause assertion in mds_close
4895 Details : It was possible to get multiple mfd references during close and
4896 client eviction, leading to one thread referencing a freed mfd.
4898 Severity: : enhancement
4899 Bugzilla : 3262, 6359
4900 Description: Attempts to reconnect to servers are now more aggressive.
4901 Details : This builds on the enhanced upcall-less recovery that was added
4902 in 1.4.2. When trying to reconnect to servers, clients will
4903 now try each server in the failover group every 10 seconds. By
4904 default, clients would previously try one server every 25 seconds.
4909 Description: After recovery, certain operations trigger a failed
4910 assertion on a client.
4911 Details : Failing over an mds, using lconf -d --failover, while a
4912 client was doing a readdir() call would cause the client to
4913 LBUG after recovery completed and the readdir() was resent.
4915 Severity : enhancement
4917 Description: Default groups are now added by lconf
4918 Details : You can now run lconf --group <servicename> without having to
4919 manually add groups with lmc.
4922 Frequency : occasional
4924 Description: Nodes with an elan id of 0 trigger a failed assertion
4927 Frequency : always when accessing e.g. tty/console device nodes
4929 Description: tty and some other devices nodes cannot be used on lustre
4930 Details : file's private_data field is used by device data and lustre
4931 values in there got lost. New field was added to struct file to
4932 store fs-specific private data.
4935 Frequency : when exporting Lustre via NFS
4937 Description: NFSD failed occasionally when looking up a path component
4938 Details : NFSD is looking up ".." which was broken in ext3 directories
4939 that had grown large enough to become hashed.
4942 Frequency : Clusters with multiple interfaces not on the same subnet
4944 Description: Nodes will repeatedly try to reconnect to an interface which it
4945 cannot reach and report an error to the log.
4946 Details : Extra peer list entries will be created by lconf with some peers
4947 unreachable. lconf now validates the peer before adding it.
4950 Frequency : Only if a default stripe is set on the filesystem root.
4952 Description: Setting a default stripe on the filesystem root prevented the
4953 filesystem from being remounted.
4954 Details : The client was sending extra request flags in the root getattr
4955 request and did not allocate a reply buffer for the dir EA.
4958 Frequency : occasional, higher if lots of files are accessed by one client
4959 Bugzilla : 6159, 6097
4960 Description: Client trips assertion regarding lsm mismatch/magic
4961 Details : While revalidating inodes the VFS looks up inodes with ifind()
4962 and in rare cases can find an inode that is being freed.
4963 The ll_test_inode() code will free the lsm during ifind()
4964 when it finds an existing inode and then the VFS later attaches
4965 this free lsm to a new inode.
4969 Bugzilla : 6422, 7030
4970 Description: MDS deadlock between mkdir and client eviction
4971 Details : Creating a new file via mkdir or mknod (starting a transaction
4972 and getting the ns lock) can deadlock with client eviction
4973 (gets ns lock and trying to finish a synchronous transaction).
4976 Frequency : occasional
4977 Description: While starting a server, the fsfilt_ext3 module could not be
4979 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4980 kernels. Previously, lconf would still use the ext3 name
4981 when trying to load modules. Now, it will correctly use
4982 ext3 on 2.4 and ldiskfs on 2.6.
4984 Severity : enhancement
4985 Description: The default stripe count has been changed to 1
4986 Details : The interpretation of the default stripe count (0, to lfs
4987 or lmc) has been changed to mean striping across a single
4988 OST, rather than all available. For general usage we have
4989 found a stripe count of 1 or 2 works best.
4991 Severity : enhancement
4992 Description: Add support for compiling against Cray portals.
4993 Details : Conditional compiling for some areas that are different
4997 Frequency : occasional
4998 Bugzilla : 6409, 6834
4999 Description: Creating files with an explicit stripe count may lead to
5000 a failed assertion on the MDS
5001 Details : If some OSTs are full or unavailable, creating files may
5002 trigger a failed assertion on the MDS. Now, Lustre will
5003 try to use other servers or return an error to the
5007 Frequency : occasional
5009 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5010 caused lock timeout during readahead (since 1.4.2).
5011 Details : Processes doing readahead might match a lock that hasn't been
5012 granted yet if there are overlapping and conflicting lock
5013 requests. The readahead process waits on ungranted lock
5014 (original lock is CBPENDING), while OST waits for that process
5015 to cancel CBPENDING read lock and eventually evicts client.
5017 Severity : enhancement
5019 Description: Initial enabling of flock support for clients
5020 Details : Implements fcntl advisory locking and file status functions.
5021 This feature is provided as an optional mount flag (default
5022 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5023 locking are implemented yet, and those that are are not guaranteed
5024 to be completely correct in production environments.
5025 mount -t lustre -o [flock|noflock] ...
5028 Frequency : occasional
5030 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5031 assertion in the ext3 JBD (journaling) layer.
5032 Details : The b_committed_data struct is protected by the big kernel lock
5033 in 2.4 kernels, serializing journal_commit_transaction() and
5034 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5035 access to this struct. In 2.6 kernels there is finer grained
5036 locking to improve SMP performance of the JBD layer.
5040 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5045 Description: merge and cleanup kernel patches.
5048 -----------------------------------------------------------------------------
5050 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5052 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5053 versions of Lustre. Please update all servers and clients to
5054 version 1.4.2 or later at the same time. You must also run
5055 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5056 to update the configuration logs.
5058 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5059 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5060 - swab llog records of type '0' so we get proper header size/idx (5861)
5061 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5062 - fix rename of one directory over another leaking an inode (5953)
5063 - avoid SetPageDirty on 2.6 (5981)
5064 - don't re-add just-being-destroyed locks to the waiting list (5653)
5065 - when creating new directories, inherit the parent's custom
5066 striping settings if present parent (3048)
5067 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5068 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5069 - don't hold client locks on temporary worklist from l_lru (5666)
5070 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5071 - automatically evict dead clients (5921)
5072 - Update file size properly in create+truncate+fstat case (6196)
5073 - Do not unhash mountpoint dentries, do not allow removal of
5075 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5076 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5077 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5078 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5079 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5080 - Correctly handle reads of files with no objects (6243)
5081 - lctl recover will also mark a device active if deactivate used (5933)
5083 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5084 - allow --write-conf on an MDS with different nettype than client (5619)
5085 - don't write config llogs to MDS for mounts not from that MDS (5617)
5086 - lconf should create multiple TCP connections from a client (5201)
5087 - init scripts are now turned off by default; run chkconfig --on
5088 lustre and chkconfig --on lustrefs to use them
5089 - upcalls are no longer needed for clients to recover to failover
5091 - add --abort-recovery option to lconf to abort recovery on device
5093 - add support for an arbitrary number of OSTs (3026)
5094 - Quota support protocol changes.
5095 - forward compatibility changes to wire structs (6007)
5096 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5097 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5098 - improved "lustre status" script
5099 - initialize blocksize for non-regular files (6062)
5100 - added --disable-server and --disable-client configure options (5782)
5101 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5102 - Vanilla 2.4.29 support
5103 - increase maximum number of obd devices to 520 (6242)
5104 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5105 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5106 SLES 9 2.6.5 kernels
5107 - limit stripes per file to 160 (the maximum EA size) (6093)
5109 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5112 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5113 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5114 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5115 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5116 - small ext3 extents cleanups and fixes (5733)
5117 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5118 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5119 - avoid needless client->OST connect, fix handle mismatch (5317)
5120 - fix DLM error path that led to out-of-sync client, long delays (5779)
5121 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5122 - fix several locking issues related to i_size (5492,5624,5654,5672)
5123 - don't move pending lock onto export if it is already evicted (5683)
5124 - fix kernel oops when creating .foo in unlinked directory (5548)
5125 - fix deadlock in obdfilter statistics vs. object create (5811)
5126 - use time_{before,after} to avoid timer jiffies wrap (5882)
5127 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5128 - Do not leak request if server was not able to process it (5154)
5129 - If mds_open unable to find parent dir, make that negative lookup(5154)
5130 - don't create new directories with extent-mapping (5909, 5936)
5132 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5133 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5134 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5135 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5136 - support for RHEL4 kernel on i686 (5773)
5137 - provide error messages when incompatible logs are encountered (5898)
5139 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5140 * version 1.4.0.10 (1.4.1 release candidate 1)
5142 - don't keep a lock reference when lock is not granted (4238)
5143 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5144 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5145 - drop import inflight refcount on signal_completed_replay error (5255)
5146 - unlock page after async write error during send (3677)
5147 - handle missing objects in filter_preprw_read properly (5265)
5148 - no transno return for symlink open, don't save no-trasno open (3440)
5149 - don't try to complete elan receive that already failed (4012)
5150 - free RPC server reply state on error (5406)
5151 - clean up thread from ptlrpc_start_thread() on error (5160)
5152 - readahead could read extra page into cache that wasn't ejected (5388)
5153 - prevent races in class_attach/setup/cleanup/detach (5260)
5154 - don't dereference de->d_inode after l_dput of de (5458)
5155 - use "int" for stripe value returned from lock_to_stripe (5544)
5156 - mballoc allocation and error-checking fixes in 2.6 (5504)
5157 - block device patches to fix I/O request sizes in 2.6 (5482)
5158 - look up hostnames for IB nals (5602)
5159 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5160 - don't start multiple acceptors for the same port (5277)
5161 - fix incorrect LASSERT in mds_getattr_name (5635)
5162 - export a proc file for general "ping" checking (5628)
5163 - fix "lfs check" to not block when the MDS is down (5628)
5165 - service request history (4965)
5166 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5167 - create an "evict_client" /proc entry on OSTs, like the MDS has
5168 - fix mount usage message, return errors per mount(8) (5168)
5169 - change grep [] to grep "[]" in tests so they work in more UMLs
5170 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5171 - remove ancient LOV_MAGIC_V0 EA support (5047)
5172 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5173 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5174 - allow readahead allocations to fail when low on memory (5383)
5175 - mmap locking landed again, after considerable improvement (2828)
5176 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5178 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5181 - send OST transaction number in read/write reply to free req (4966)
5182 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5183 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5184 - the watchdog thread now runs as interruptible (5246)
5185 - flock/lockf fixes (but it's still disabled, pending 5135)
5186 - don't use EXT3 constants in llite code (5094)
5187 - memory shortage at startup could cause assertion (5176)
5189 - reorganization of lov code
5190 - single portals codebase
5192 - add extents/mballoc support (5025)
5193 - direct I/O reads in the obdfilter (4048)
5194 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5196 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5199 - send OST transaction number in read/write reply to free req (4966)
5200 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5201 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5202 - the watchdog thread now runs as interruptible (5246)
5203 - handle missing objects in filter_preprw_read properly (5265)
5204 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5205 - drop import inflight refcount on signal_completed_replay error (5255)
5206 - unlock page after async write error during send (3677)
5207 - return original error code on reconstructed replies (3761)
5208 - no transno return for symlink open, don't save no-trasno open (3440)
5210 - add pid to ldlm debugging output (4922)
5211 - bump the watchdog timeouts -- we can't handle 30sec yet
5212 - extra debugging for orphan dentry/inode bug (5259)
5214 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5217 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5218 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5219 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5220 - fix NULL dereference in /proc/sys/portals/routes (4827)
5221 - allow failed mdc_close() operations to be interrupted (4561)
5222 - stop precreate on OST before MDS would time out on it (4778)
5223 - don't send partial-page writes before EOF from client (4410)
5224 - discard client grant for sub-page writes on large-page clients (4520)
5225 - don't free dentries not owned by NFS code, check generation (4806)
5226 - fix lsm leak if mds_create_objects() fails (4801)
5227 - limit debug_daemon file size, always print CERROR messages (4789)
5228 - use transno after validating reply (3892)
5229 - process timed out requests if import state changes (3754)
5230 - update mtime on OST during writes, return in glimpse (4829)
5231 - add mkfsoptions to LDAP (4679)
5232 - use ->max_readahead method instead of zapping global ra (5039)
5233 - don't interrupt __l_wait_event() during strace
5235 - add software watchdogs to catch hung threads quickly (4941)
5236 - make lustrefs init script start after nfs is mounted
5237 - fix CWARN/ERROR duplication (4930)
5238 - return async write errors to application if possible (2248)
5239 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5240 - print NAL number in %x format (4645)
5241 - update barely-supported suse-2.4.21-171 series (4842)
5242 - support for sles 9 %post scripts
5243 - support for building 2.6 kernel-source packages
5244 - support for sles km_* packages
5246 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5249 - ignore -ENOENT errors in osc_destroy (3639)
5250 - notify osc create thread that OSC is being cleaned up (4600)
5251 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5252 - reconstruct ost_handle() like mds_handle() (4657)
5253 - create a new thread to do import eviction to avoid deadlock (3969)
5254 - let lconf resolve symlinked-to devices (4629)
5255 - don't unlink "objects" from directory with default EA (4554)
5256 - hold socknal file ref over connect in case target is down (4394)
5257 - allow more than 32000 subdirectories in a single directory (3244)
5258 - fix blocks count for O_DIRECT writes (3751)
5259 - OST returns ENOSPC from object create when no space left (4539)
5260 - don't send truncate RPC if file size isn't changing (4410)
5261 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5262 - bind to privileged port in socknal and tcpnal (3689)
5264 - rate limit CERROR/CWARN console message to avoid overload (4519)
5265 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5266 - basic mmap support (3918)
5267 - kernel patch series update from b1_4 (4711)
5269 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5272 - avoid crash during MDS cleanup with OST shut down (2775)
5273 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5274 - don't use bad inodes on the MDS (3744)
5275 - dynamic object preallocation to improve recovery speed (4236)
5276 - don't hold spinlock over lock dumping or change debug flags (4401)
5277 - don't zero obd_dev when it is force cleaned (3651)
5278 - print grants to console if they go negative (4431)
5279 - "lctl deactivate" will stop automatic recovery attempts (3406)
5280 - look for existing locks in ldlm_handle_enqueue() (3764)
5281 - don't resolve lock handle twice in recovery avoiding race (4401)
5282 - revalidate should check working dir is a directory (4134)
5284 - don't always mark "slow" obdfilter messages as errors (4418)
5286 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5289 - don't close LustreDB during write_conf until it is done (3860)
5290 - fix typo in lconf for_each_profile (3821)
5291 - allow dumping logs from multiple threads at one time (3820)
5292 - don't allow multiple threads in OSC recovery (3812)
5293 - fix debug_size parameters (3864)
5294 - fix mds_postrecov to initialize import for llog ctxt (3121)
5295 - replace config semaphore with spinlock (3306)
5296 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5297 - don't allow enqueue to complete on a destroyed export (3822)
5298 - down write_lock before checking llog header bitmap (3825)
5299 - recover from lock replay timeout (3764)
5300 - up llog sem before sending rpc (3652)
5301 - reduce ns lock hold times when setting kms (3267)
5302 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5303 - fix NULL deref and obd_dev leak on setup error (3312)
5304 - replace some LBUG about llog ops with error handling (3841)
5305 - don't match INVALID dentries from d_lookup and spin (3784)
5306 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5307 - fix invalid assertion in ptlrpc_set_wait (3880)
5309 - add libwrap support for the TCP acceptor (3996)
5310 - add /proc/sys/portals/routes for non-root route listing (3994)
5311 - allow setting MDS UUID in .xml (2580)
5312 - print the stack of a process that LBUGs (4228)
5314 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5317 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5318 - make sure to unset replay flag from failed open requests (3440)
5319 - if default stripe count is 0, use OST count for inode size (3636)
5320 - update parent mtime/ctime on client for create/unlink (2611)
5321 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5322 - free recovery state on server during a forced cleanup (3571)
5323 - unregister_reply for resent reqs (3063)
5324 - loop back devices mounting and status check on 2.6 (3563)
5325 - fix resource-creation race that can provoke i_size == 0 (3513)
5326 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5327 - more debugging for page-accounting assertion (3746)
5328 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5329 - avoid deadlock after precreation failure (3758)
5330 - fix race and lock order deadlock in orphan handling (3450, 3750)
5331 - add validity checks when grabbing inodes from l_ast_data (3599)
5333 - add /proc/.../recovery_status to obdfilter (3428)
5334 - lightweight CDEBUG infrastructure, debug daemon (3668)
5335 - change default OSC RPC parameters to be better on small clusters
5336 - turn off OST read cache for files smaller than 32MB
5337 - install man pages and include them in rpms (3100)
5338 - add new init script for (un)mounting lustre filesystems (2593)
5339 - run chkconfig in %post for init scripts (3701)
5340 - drop scimac NAL (unmaintained)
5342 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5345 - clean kiobufs before and after use (3485)
5346 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5347 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5348 - change init script to fail more clearly if not run as root (1528)
5349 - allow clients to reconnect during replay (1742)
5350 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5351 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5352 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5353 - fix cond_resched (3554)
5354 - only evict unfinished clients after recovery (3515)
5355 - allow bulk resend, prevent data loss (3570)
5356 - dynamic ptlrpc request buffer allocation (2102)
5357 - don't allow unlinking open directory if it isn't empty (2904)
5358 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5359 - remove extraneous obd dereference causing LASSERT failure (3334)
5360 - don't use get_cycles() when creating temp. files on the mds (3156)
5361 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5362 - handle EEXIST for set-stripe, set proper directory name (3336)
5364 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5365 - fix ksocknal_fmb_callback() error messages (2918)
5367 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5370 - don't copy lvb into (possibly NULL) reply on error (2983)
5371 - don't deref dentry after dput, don't free lvb on error (2922)
5372 - use the kms to determine writeback rpc length (2947)
5373 - increment oti_logcookies when osc is inactive (2948)
5374 - update client's i_blocks count via lvb messages (2543)
5375 - handle intent open/close of special files properly (1557)
5376 - mount MDS with errors=remount-ro, like obdfilter (2009)
5377 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5378 - don't use cancelling-locks' kms values (2947)
5379 - use highest lock extent for kms, not last one (2925)
5380 - don't dereference ERR_PTR() dentry in error handling path (3107)
5381 - fix thread race in portals_debug_dumplog() (3122)
5382 - create lprocfs device entries at setup instead of at attach (1519)
5383 - common AST error handler, don't evict client on completion race (3145)
5384 - zero nameidata in detach_mnt in 2.6 (3118)
5385 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5386 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5387 - initialize RPC timeout timer earlier for 2.6 (3219)
5388 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5389 - print nal/nid for unknown nid (3258)
5390 - additional checks for oscc recovery before doing precreate (3284)
5391 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5392 - don't crash in mdc_close for bad permissions on open (3285)
5393 - zero i_rdev for non-device files (3147)
5394 - clear page->private before handing to FS, better assertion (3119)
5395 - tune the read pipeline (3236)
5396 - fix incorrect decref of invalidated dentry (2350)
5397 - provide read-ahead stats and refine rpc in flight stats (3328)
5398 - don't hold journal transaction open across create RPC (3313)
5399 - update atime on MDS at close time (3265)
5400 - close LDAP connection when recovering to avoid server load (3315)
5401 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5402 - don't leak open file on MDS after open resend (3325)
5403 - serialize filter_precreate and filter_destroy_precreated (3329)
5404 - loop device shouldn't call sync_dev() for nul device (3092)
5405 - clear page cache after eviction (2766)
5406 - resynchronize MDS->OST in background (2824)
5407 - refuse to mount the same filesystem twice on same mountpoint (3394)
5408 - allow llmount to create routes for mounting behind routers (3320)
5409 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5410 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5411 - fix rare problem with rename on htree directories (3417)
5413 - allow default OST striping configuration per directory (1414)
5414 - fix compilation for qswnal for 2.6 kernels (3125)
5415 - increase maximum number of MDS request buffers for large systems
5416 - change liblustreapi to be useful for external progs like lfsck (3098)
5417 - increase local configuration timeout for slow disks (3353)
5418 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5420 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5423 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5424 - don't overwrite extent policy data in reply if lock was blocked (2901)
5425 - drop filter export grants atomically with removal from device (2663)
5426 - del obd_self_export from work_list in class_disconnect_exports (2908)
5427 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5428 - swab reply message in mdc_close, other PPC fixes (2464)
5429 - fix destroying of named logs (2325)
5430 - overwrite old logs when running lconf --write_conf (2264)
5431 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5432 - fix race in target_handle_connect (2898)
5433 - mds_reint_create() should take same inode create lock (2926)
5434 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5435 - don't close files for self_export to avoid uninitialized obd (2936)
5436 - allow MDS with the same name as client node (2939)
5437 - hold dentry reference for closed log files for unlink (2325)
5438 - reserve space for all logs during transactions (2059)
5439 - don't evict page beyond end of stripe extent (2925)
5440 - don't oops on a deleted current working directory (2399)
5441 - handle hard links to targets without a parent properly (2517)
5442 - don't dereference NULL lock when racing during eviction (2867)
5443 - don't grow lock extents when lots of conflicting locks (2919)
5445 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5448 - account for cache space usage on clients to avoid data loss (974)
5449 - lfsck support in lustre kernel code (2349)
5450 - reduce journal credits needed for BRW writes (2370)
5451 - orphan handling to avoid losing space on client/server crashes
5452 - ptlrpcd can be blocked, stopping ALL progress (2477)
5453 - use lock value blocks to assist in proper KMS, faster stat (1021)
5454 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5455 - recovery for initial connections (2355)
5456 - fixes for mds_cleanup_orphans (1934)
5457 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5458 - block all file creations until orphan recovery completes (1901)
5459 - client remove rq_connection from request struct (2423)
5460 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5461 - recovery timer race (2670)
5462 - mdc_close recovey bug (2532)
5463 - ptlrpc cleanup bug (2710)
5464 - mds timeout on local locks (2588)
5465 - namespace lock held during RPCs (2431)
5466 - handle interrupted sync write properly (2503)
5467 - don't try to handle a message that hasn't been replied to (2699)
5468 - client assert failure during cleanup after abort recovery (2701)
5469 - leak mdc device after failed mount (2712)
5470 - ptlrpc_check_set allows timedout requests to complete (2714)
5471 - wait for inflight reqs when ptlrpcd finishes (2710)
5472 - make sure unregistered services are removed from the srv_list
5473 - reset bulk XID's when resending them (caught by 1138 test)
5474 - unregister_bulk after timeout
5475 - fix lconf error (2694)
5476 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5477 - readahead locks pages, leaves pending causing memory pressure (2673)
5478 - increase OST request buffers to 4096 on large machines (2729)
5479 - fix up permission of existing directories in simple_mkdir (2661)
5480 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5481 - don't assign transno to errored transactions (2742)
5482 - don't delete objects on OST if given a bogus objid from MDS (2751)
5483 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5484 - if rq_no_resend, then timeout request after recovery (2432)
5485 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5486 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5487 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5488 - MDS recovery shouldn't delete objects that it has given out (2730)
5489 - if enqueue arrives after completion, don't clobber LVB (2819)
5490 - don't unlock pages twice when trigger_group_io returns error (2814)
5491 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5492 - don't write pages to disk if there was an error (1450)
5493 - don't ping imports that have recovery disabled (2676)
5494 - take buffered bytes into account when balancing socknal conn (2817)
5495 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5496 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5497 - remove little-endian swabbing of llog records (1987)
5498 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5499 - retry reposting request buffers if they fail (1191)
5500 - grow extent at grant time to avoid granting a revoked lock (2809)
5501 - lock revoke doesn't evict page if covered by a second lock (2765)
5502 - disable VM readahead to avoid reading outside lock extents (2805)
5504 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5505 - updated kernel patches for hp-2.4.20 kernel (2681)
5507 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5510 - fix truncated write corruption (2366)
5511 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5513 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5514 - ensure len(uuid) < 37 in lmc (1171)
5515 - fix ia64 OOPS in llog_test (2255)
5516 - zero end of page at obdfilter for partial page writes (2648)
5517 - don't leave stale dentries around after renames (bug 2428)
5518 - fix timeouts when evicting a client with a single lock held (2642)
5519 - set deadline for the initial HELLO message to drain (2634)
5520 - print out dotted-quad IP addresses in the socknal (2302)
5522 - additional debugging for MDS client eviction problem (2443)
5523 - fix mkfsoptions support for osts (2603, 2604)
5525 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5528 - add series for the vanilla 2.6.0 kernel
5529 - add series for the vanilla 2.4.24 kernel
5530 - add series for a cray x86/64 UL kernel drop
5531 - fix xattr patches for the vanilla 2.4.19 series
5533 - generate true UUIDs in lmc (1171)
5534 - have portals stack dumping break in UML (2466)
5535 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5536 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5537 - wake callback waiting threads on client eviction (2460)
5538 - Add --ptldebug and --subsystem to lmc (1719)
5539 - update assertion to allow safe interrupt allocation
5540 - set rq_no_resend for cancel requests (2432)
5541 - recalculate ptlrpcd timeout after resend (2494)
5542 - call vfs_rmdir when removing pending directories (2368)
5543 - fix renaming a file to itself (2429)
5544 - lmc creates a default one-stripe lov (2454)
5545 - expand procfs space to handle large clusters (2326)
5546 - increase UML stack to avoid overflow
5547 - update lconf's list of debug and subsystem masks
5548 - fix lfs find --obd (2510)
5549 - /proc tunable for disabling filter read caching (2591)
5550 - stop rpm packages from altering slapd.conf (2301)
5551 - disable nagle in the socknal under 0conf (2578)
5552 - choose mds inode size based on stripe count (2572)
5553 - fix kernel-source rpm problems (2516)
5555 - add --disable-doc to avoid pdf generation (2421)
5556 - update documentation, tests, type-os, comments
5557 - avoid format warnings on ia64
5558 - remove the TOE NAL
5559 - tiny code cleanups by removing unused fields
5561 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5564 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5565 - remove the most blatant lies from BUILDING file (2371)
5566 - change default debug level to reasonable production setting
5567 - reduce client side cache size to reduce cache flush time
5568 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5569 - make TCP zerocopy and pinger support enabled by default (2476)
5570 - sync writes completed after process exits caused crashes (2319)
5571 - maintain correct mount count on the MDS (2356)
5572 - backout 1557, because 2316 wasn't really fixed
5573 - better file I/O statistics gathering in /proc
5574 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5575 - another kernel patch to fix zero-copy TCP function export
5576 - don't take duplicate lock when processing re-sent getattr (2420)
5577 - lctl uses obd_self_export instead of creating new conn (2353)
5578 - MDS/OST recovery case which requires object creation asserted (2425)
5579 - move lfs from /usr/sbin to /usr/bin in packages
5580 - fix race between mds_client_add and mds_client_free (2417)
5581 - use kmalloc instead of slabs in portals (2430)
5582 - don't create duplicate records when a failover MDS is present (2442)
5583 - remove unnecessary mount age check (2332)
5584 - don't remove directory inodes from locks prematurely (2451)
5585 - don't break if MDS service name is the same as hostname (2103)
5586 - fix races in client write RPC generation when cache full (2482)
5588 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5591 - remove now-unused request->rq_obd (278)
5592 - if an allocation fails, print out how much memory we've used (1933)
5593 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5594 - add the "configurable stack size" patch to most series files (1256)
5595 - ability to write large log records, for 100+ OST configs (2306)
5596 - fix NULL deref when filter_prep fails (2314)
5597 - fix operator precedence error in filter_sync
5598 - dynamic allocation of socknal TX descriptors (2315)
5599 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5600 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5601 - ensure that CWARN messages in Portals always get to the syslog
5602 - __init/__exit are not for prototype decls (ldlm_init/exit)
5603 - x86-64 compile warning fixes
5604 - fix gateway LMC keyword conflict (2318)
5605 - fix MDS lock inversions in getattr/reint paths (1844)
5606 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5607 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5608 - DLM race condition prevented some lock evictions (2328)
5609 - ENOMEM detection and retry on socknal sends (2230)
5610 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5611 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5612 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5613 - better error messages when a client is rejected during recovery (1505)
5614 - avoid cancelling locks which were never granted, after failure (2330)
5615 - fix i_sem/journal inversion in mds_client_add (2333)
5616 - fix truncate/getattr lock cycle deadlock (2334)
5617 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5618 - fix two rare exit paths which could leak an l_lock() ref (2321)
5619 - fencepost error in MDS/OST orphan recovery (2226)
5620 - make log record alignment 8 bytes (1988)
5621 - lstripe now fails when requested offset > ost_count (2237)
5622 - ensure that all kernel series have a complete list.h (1607)
5623 - fix crashes in special-file operations (2316)
5624 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5626 - allow configurable automake binary, for testing new versions
5627 - small update to the lfs documentation
5629 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5631 * fix negative export reference count in fsfilt_sync (2312)
5633 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5634 * release candidate 0.9.1
5636 - orphans are moved into the PENDING directory for possible recovery
5637 - replayed opens now open by fid for orphan/rename safety (1042)
5638 - last close of an orphan inode generates a transno (683)
5639 - chdir() and mount() now pin the directory entry (1020)
5640 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5641 - discard very old requests without processing them (1502)
5642 - remove some common, well-understood CERRORs (1505)
5643 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5644 - clear "grant" flags in OST replies until OST grant code lands (1644)
5645 - fix read performance by not clobbering i_blksize on client (1598)
5646 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5647 - make LRU size a /proc tunable, clears locks when reduced (707)
5648 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5649 - server locks take references on exports now (1558)
5650 - build fixes for 2.4.20-rh trees (1663)
5651 - return an error from lov_create if all OSCs are inactive (1751)
5652 - fix import levels when a reconnect happens without a timeout (1597)
5653 - exit early from mds_open if we get a lookup error (1749)
5654 - partial page read at EOF wouldn't wait for disk before sending (1642)
5655 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5656 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5657 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5658 - write/truncate lock inversion (1639)
5659 - Don't auto-load obdclass, portals modules during cleanup (1495)
5660 - fix timestamps from jumping to "now" (1763)
5661 - extra journal assertions (1648)
5662 - add an extra multiunlink test (1771)
5663 - fix read_record/write_record API (1776)
5664 - fix leak of offset_extent, possible incorrect i_size later (1772)
5665 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5666 - Debugging for the kqswnal_get_idle_tx problems (1820)
5667 - Allow recovery to be attempted multiple times (1536)
5668 - Write out MDS last_rcvd file after it is first created (1600)
5669 - Fix tx_descriptor leak in failed transmit situations (1827)
5670 - ext3 journaling fixes for assertion failure after IO error (1871)
5671 - class_export_put() on freed export after completion AST error (1896)
5672 - Fix revalidate looping in VFS (1322)
5673 - Don't access a freed export during MDS_REINT timeout (1521)
5674 - Add open-unlink recovery support on the MDS (1673,1764)
5675 - Return an error if no MDS data was read from last_rcvd (1946)
5676 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5677 - Don't setup a disk device that doesn't match exported UUID (317)
5678 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5679 - avoid committing NULL handle in force close
5680 - local.sh is now a one-stripe LOV configuration
5681 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5682 - fix ext3 htree duplicate directory entry corruption (1516)
5683 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5684 - update to new LOV EA format (2097)
5685 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5687 2003-06-15 Phil Schwan <phil@clusterfs.com>
5690 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5691 - per-import/export recovery handling (958, 931, 959)
5692 - multiple last-rcvd slots, for serving multiple FSes (949)
5693 - connections are again shared between multiple imp/exports (963, 964)
5694 - "umount -f" would hang if any requests needed to be sent (393, 978)
5695 - avoid pinning large req buffer by copying for queued messages (989)
5696 - add "uuid" to "lctl device" command to help upcalls (991)
5697 - "open" RPCs with transnos would confuse recovery counters (1037)
5698 - do proper endian conversion of all wire messages (288, 340, 891)
5699 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5700 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5701 - fix ptlrpc_connection leak in target_handle_connect (1174)
5702 - fix import refcounting bug in OST and MDS cleanup (1134)
5703 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5704 - fix very unlikely obd_types race condition (501)
5705 - remove osc_open hack for echo_client (1187)
5706 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5707 - a failure in read_inode2 leads to deadlock (1139)
5708 - cancel ack-locks as soon as transaction is committed (1072)
5709 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5710 - make sure to commitrw after any preprw to avoid deadlock (1162)
5711 - failing to execute a file in a lustre FS would lock inode (1203)
5712 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5713 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5714 - obd_uuid2tgt would walk off the end of the list (1255)
5715 - on IA64 the osc would give portals incorrect bulk size (1258)
5716 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5717 - fix lock inversion caused by new llite matching code (1282)
5718 - limit the number of dirty pages on a client to 10MB (1286)
5719 - timed out locks were not being corrected cancelled (1289)
5720 - fix O_DIRECT above 4GB on IA-32 (1292)
5721 * major user-visible changes
5722 - fail out/fail over policy now controlled by the upcall (993)
5724 - add OBD_PING to check server availability and failure (954)
5725 - lustre messages are now sent in sending host order (288, 340, 891)
5726 - add eadatalen to MDS getattr reply (340)
5727 - OST read replies may contain second buffer, with per-page status (593)
5729 2003-03-11 Phil Schwan <phil@clusterfs.com>
5732 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5733 - failed open()s could cause deadlock; fixed (867, 869)
5734 - stop cancelling OST locks when files are closed (481)
5735 - overlapping XID spaces caused network corruption (851, 853)
5736 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5737 - setattr_raw added, to do single-RPC, server-side setattrs
5738 - lmc/lconf syntax change for OST UUIDs
5739 - fix crashy race condition between ptlrpc_free_req and osc_close
5740 - don't use request in mdc_enqueue if we hit a timeout (889)
5741 - don't set the inode i_size for regular files from the MDS (896)
5742 - handle out of order completion AST (842)
5743 - don't LBUG if a lock request times out after receiving AST (913)
5744 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5745 - if a bad lock AST arrives, send an error instead of dropping entirely
5746 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5747 - fix leak in bulk IO when only partially completed (899, 900, 926)
5748 - fix O_DIRECT for ia64 (55)
5749 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5750 - C-z after timeout could hang a process forever; fixed (977)
5752 - client-side I/O cache (678, 924, 929, 941, 970)
5754 - READPAGE and SETATTRs which don't take server-side locks get
5757 2003-02-11 Phil Schwan <phil@clusterfs.com>
5760 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5761 - Fix fsfilt_extN_readpage() to read a full page of directory
5762 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5763 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5764 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5765 - Most intent operations were removed, in favour of a new RPC mode
5766 that does a single RPC to the server and bypasses most of the VFS
5767 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5768 - Aggressively cancel local locks on DLM servers
5769 - mds_reint_unlink sends EA to the client if it's the last nlink.
5770 client uses that EA to unlink OST objects.
5771 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5772 - recursive symlinks were fixed (439)
5773 - fixed NULL deref in DEBUG_REQ
5774 - filter_update_lastobjid no longer calls sync, which annoyed extN
5775 - fixed multi-client small-writes to a single file problem (445)
5776 - fixed mtime updates during file writes (607)
5777 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5778 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5779 - fixed Linux OST size reporting problem (444, 656)
5780 - OST now updates object mtime with writes or setattr (607, 619)
5781 - client verifies file size before zeroing page past EOF (445)
5782 - OST now writes last allocated objid to disk with allocation (108)
5783 - LOV on echo now works (409)
5785 - mds_reint_unlink sends a new buffer, with the EA included. this
5786 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5787 set if a regular file was being unlinked, and it was the last link
5788 - use PtlGet from the target for bulk writes (315)
5789 - OST now updates object mtime with writes or setattr (607, 619)
5790 - LDLM now has a grant-time callback to revalidate locked items, if
5792 - Many MDS operations were reorganized to combat race conditions
5794 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5795 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5797 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5800 - Fully reactivate OST imports after reconnection (512, others)
5801 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5802 - More graceful error handling for truncating on dead OST (515)
5803 - Don't error out unless we're actually accessing dead stripes (474)
5804 - Fix garbage sizes when stripes are missing (410)
5805 - LRU counters were broken, causing constant lock purge (433, 432)
5806 - garbage on read from stripes with failed OSTs (441)
5807 - mark OSCs as active before reconnecting during recovery (438)
5808 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5809 - lfind did not preserve OST order in output (443)
5810 - symlinks cause hung clients, incorrect data (439)
5811 - stop dereferencing request after dropping refcount (457)
5812 - don't LASSERT(spin_is_locked) on non-SMP (455)
5813 - fixes for many rename() bugs
5814 - fstat didn't correctly synchronize attributes (399)
5815 - server must handle lock cancellation during blocking AST prep (487)
5816 - bulk descriptors were free()d too soon (511)
5817 - fix paths in lconf, which would load incorrect modules (451, 507)
5818 - fix confusing lconf 'host not found' error message (386)
5819 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5820 - fix race condition in mdc_blocking_ast() for inode access (526)
5821 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5822 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5823 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5824 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5825 - fix lstripe in several ways (kernel side) (527)
5826 - fix request leak in ldlm_cli_enqueue (262)
5827 - incorrect OSC was marked inactive after OST failure
5828 - call mds_fs_cleanup before unmounting filesystem (524)
5829 - fix races between taking ns_lock and ldlm_lock_change_resource
5830 - fix races updating LOV export open file list
5831 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5832 - fix recovery NULL deref in ldlm_cli_cancel_unused
5833 - fix some DLM races by using new hash table for lock handles (419)
5834 - permit the client to specify desired inodes, at replay
5835 - duplicate requests when we queue them for replay reintegration
5836 - fix last_rcvd offset calculation
5837 - sync after each recovered transaction, so we always make progress
5838 - never, not always, ERESTART requests without transnos
5839 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5841 - skip replay if the MDS says that the client is already connected
5842 - don't check for a recovery-enabled export to match lctl's UUID
5843 - don't INC_USE_COUNT for phantom exports
5844 - don't crash when cleaning up phantom exports (567)
5845 - don't double-finish or set replay data for errored mdc_open requests
5846 - abort requests when they time out, so we don't get old replies
5847 - send/receive replies for AST messages again
5848 - if the client says that it doesn't have the lock, cancel it on the
5850 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5851 mark it as destroyed, it will all work out when decref is called
5852 - fix module use counts (22, 581)
5854 - ASTs now expect a reply (server cancels lock on error reply)
5856 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5859 - fix many simultaneous client startup (392)
5860 - fix dentry->d_it clobbering
5861 - credentials weren't being shipped for readdir/getattr operations
5862 - remove invalid assertions triggered during some concurrent MD
5864 - proper Lustre versions added (336, 389)
5865 - fix memory leak for create error case (398)
5866 - fix LOV locking bug that would get cli/srv out of sync
5867 - fix echo client over LOV (409)
5868 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5869 - fix double-O_EXCL intent crash (424)
5870 - avoid sending multiple lock CANCELs (352)
5872 - MDS can do multi-client recovery (modulo bugs in new code)
5874 - many updates, edits, cleanups
5876 2002-11-18 Phil Schwan <phil@clusterfs.com>
5879 - fix null d_it dereference (346)
5880 - fix full OST/dbench hang (333)
5881 - fix permission problem with file removal (286)
5882 - fix removal of OSCs from LOV when they fail
5883 - fix NULL deref during bulk timeout (214)
5884 - fix problems related to multiple filesystems on one MDS (241)
5885 - fixed serious subtle metadata locking bugs
5886 - free locks on clients when inodes are removed due to memory
5888 - fix inode pointer in lock data (285)
5889 - partial support for multiple MDS on a single host (241)
5890 - data locks weren't cancelled at clear_inode time (290, 311)
5891 - intent locks could lead to unbounded lock growth (205)
5892 - added a maximum lock count, an LRU list, and a flusher
5893 - fix multiple rename (365)
5894 - properly abstracted the echo client
5895 - OSC locked 1 byte too many; fixed
5896 - rewrote brw callback code:
5897 - fixed recovery bugs related to LOVs (306)
5898 - fixed too-many-pages-in-one-write crash (191)
5899 - fixed (again) crash in sync_io_timeout (214)
5900 - probably fixed callback-related race (385)
5902 - Add capability to MDS protocol
5903 - LDLM cancellations and callbacks on different portals
5905 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5908 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5909 - defer open object destruction to close time (601981, 138)
5910 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5911 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5912 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5913 - don't call obd_setattr in ll_file_release for destroyed objects
5914 * protocol change to lustre_msg: move |version| and add |flags|
5915 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5916 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5917 * added replay of create, unlink, link and rename operations during
5918 MDS failover; recovery should be much more robust now
5919 * remove failed OSCs from LOVs (only lov_create uses this so far)
5920 * the lustre-HOWTO was brought (more) up to date (582544)
5922 2002-10-23 Phil Schwan <phil@clusterfs.com>
5925 - in-use dentries weren't being reused properly (617851)
5926 - prevent multiple LDLM setup (599178)
5927 - fix LOV size calculations for truncate (617853)
5928 - fix client handling of MDS intent errors (POSIX)
5929 - fix permission bug in lovstripe.c test (624321)
5930 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5931 - truncate past end of file could corrupt data
5932 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5933 - a race in recovery could return ETIMEDOUT to apps (623947)
5934 - building outside the source directory was fixed
5935 * the lustre-HOWTO was brought (more) up to date (582544)
5936 * major progress was made on recovery functionality
5938 2002-10-10 Phil Schwan <phil@clusterfs.com>
5941 - recovery deadlock fix
5942 - rm -rf causes LBUG fix (617817)
5943 - file open by multiple tasks fix (618962)
5944 - directory permissions bugs (602707 and 620007)
5945 - journal_stop fixed with locking (611313)
5946 - O_APPEND failures resolved (618273, perhaps 614459)
5947 - lconf PATH fix (619770)
5948 - IA64 build fix (621450)
5949 - RPC buffer sizes scale with amount of memory
5951 2002-10-01 Phil Schwan <phil@clusterfs.com>
5954 - locks would be cancelled without throwing away data pages,
5955 resulting in inconsistent data (605627)
5956 - inode attributes were not always being refreshed (605627, 612449)
5957 - lconf now continues to cleanup after lctl reports an error
5958 - MDS now enforces user permissions (602707)
5959 - lprocfs cleanup fixed, but not yet enabled (614157)
5960 - fixed infinite server hang, should a client not respond to an AST
5961 - avoid going into recovery if user calls readlink() with a buffer
5962 that's too small (613941)
5963 - AST RPCs no longer require replies (614867) -- this may be changed
5964 - don't crash server if client sends an IOV that's too big (611336)
5965 - fixed lock conversion deadlock (611892)
5966 - fixed the following of symlinks (614622)
5967 * recovery: the server can remove locks from a client that dies, other
5968 clients can make progress
5969 * more extN patch fixes
5970 * compile-time configurable ptlrpc buffer allocations
5972 - collaborative read cache document
5973 - Lustre Lite Performance CDR document-in-progress
5975 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5978 - fix typo in patch-2.4.18
5980 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5983 - clear ptlrpc request each time in handle_incoming_request()
5984 - unlink of files now destroys the object on the OST
5986 2002-09-19 Peter Braam <braam@clusterfs.com>
5988 * add hard link support
5989 * change obdfile creation method
5990 * kernel patch changed
5992 2002-09-19 Peter Braam <braam@clusterfs.com>
5995 - stack overflow bug in extN fixed
5997 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5999 * documentation updates
6000 - add man pages for config tools
6001 - update tests/README to describe testing with new config tools
6002 - finish metadata API descriptions
6003 * bug fixes and cleanups
6004 - statfs workaround for 16TB limit
6005 - LOV stripe allocation improved, can stripe on subset of OSTs
6006 - LOV file size/IO offset was wrong for files > 4GB in size
6007 - object EA data was being dropped, caused files to be unreadable
6008 - memory overflow with non-LOV OST caused memory corruption
6009 - fixed regression tests to work with new config tools, obdfilter
6010 - fixed bug when directory size became larger than 1 block
6011 - fixed bug (for single client case) when PWD was deleted
6012 - invalidate local directory pages when doing intent-based ops
6013 - avoid LDLM oops when lock callback contained bad data
6015 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6017 * documentation updates
6018 * bug fixes and cleanups
6019 - configuration tools
6022 - 64-bit compile warnings
6023 - 64-bit internal statfs data
6025 * test_brw on persistent OST devices
6027 * lprocfs (disabled)
6029 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6031 * documentation updates
6032 * bug fixes and cleanups
6033 * configuration tools
6035 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6038 * many small fixes to 0_5_4
6039 * io/network handling
6040 * thinkos in MDS operations
6042 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6045 * crucial basic fixes to 0.5.3
6046 * IOR, Iozone work over Elan
6049 2002-08-07 Phil Schwan <phil@clusterfs.com>
6050 * version 0_5_3, our first alpha
6051 * we use the new Portals iovs
6052 * documentation updates
6053 * bug fixes and cleanups
6054 * small changes in the DLM wire protocol
6056 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6057 * version 0_5_1 with some initial stability,
6058 * locking on MD and file I/O.
6059 * documentation updates
6060 * several bug fixes since 0.5.0
6061 * small changes in wire protocol
6063 2002-07-18 Phil Schwan <phil@clusterfs.com>
6065 * delivered as Lustre Light Alpha
6066 * fixed a crash after handling invalid MDS requests
6067 * fixed directory pages for architectures with non-4k pages sizes
6069 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6070 * release version v0_4_4
6071 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6073 * Fixes a number of interruption problems with OST operations.
6074 * Update documentation for portals header changes
6075 * Move all wire protocol structs/defines to lustre_idl.h
6076 * Fixes symlink length bug.
6077 * Add tcpdump to repository.
6079 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6080 * release version v0_4_3
6081 * Fixes statfs for inodes on extN.
6082 * Fixes bug in runtests which would delete /etc/hosts.
6083 * Use 64-bit object IDs wherever possible (not into VFS though)
6084 Remove ost_get_info, which is unused by lustre, and out of date.
6086 2002-07-03 Peter Braam <braam@clusterfs.com>
6087 * release version v0_4_2 Fixes a lookup error (type not passed)
6088 * move forward to head of Portals
6089 * move forward to latest Lustre kernel
6091 2002-06-25 Peter Braam <braam@clusterfs.com>
6092 * release version v0_4_1. Hopefully stable on single node use.