1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.7.1.el5 (RHEL 5),
6 2.6.18-128.7.1.el5 (OEL 5),
7 2.6.22.14 vanilla (kernel.org).
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.21 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.6.sun1
12 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
13 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
14 removed cwd "./" (refer to Bugzilla 14399).
15 * File join has been disabled in this release, refer to Bugzilla 16929.
17 Severity : enhancement
19 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
22 Severity : enhancement
24 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
26 Severity : enhancement
28 Description: correctly handle big reply message.
29 Details : send LNet event if reply is bigger then buffer and adjust this buffer
34 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
36 Severity : enhancement
38 Description: Add support for OEL5.
42 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
46 Description: client could not reconnect to OST because of an active request.
47 Details : The client now retries to reconnect to the same server, if a
48 connect request failed with EBUSY or -EAGAIN. Test case 35b
49 conf-sanity.sh has been created.
53 Description: Don't allow make backward step on assignin osc next id.
54 Details : race between allocation next id and ll_sync thread can be cause
55 of set wrong osc next id and can be kill valid ost objects.
59 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
63 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
67 Description: Error handling in osc_statfs_interpret() has been improved.
68 Details : Check in osc_statfs_interpret() for EBADR.
72 Description: Do not send statfs() requests to OSTs disabled by administrator.
73 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
75 Severity : enhancement
77 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
79 Severity : enhancement
81 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
83 Severity : enhancement
85 Description: Enhancement for lfs(1) command to use numeric uid/gid.
89 Description: Improved cleanup procedure for conf-sanity test_46a.
93 Description: Do not start bulk transfer, if client tried to reconnect.
94 Details : Do not start bulk transfer, if client tried to reconnect,
95 but failed due to active rpcs. Check for exp_abort_active_req
100 Description: Prevent inconsistences between linux and lustre mount structures.
101 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
102 Make the sleep interruptible.
104 Severity : enhancement
106 Description: provide server to client comms path
107 Details : server to client communications path via new LDLM_SET_INFO rpc
111 Description: truncate starts GFP_FS allocation under transaction causing deadlock
112 Details : ldiskfs_truncate calls grab_cache_page which may start page
113 allocation under an open transaction. This may lead to
114 calling prune_icache with consequent lustre reentrance.
118 Description: send correctly lsm on open replay
119 Details : MDS is trust to LSM size on replay open, but client can set wrong size
124 Description: enable client interface failover
125 Details : When a child reconnects from another NID, properly update export
126 nid hash position and ldlm reverse import.
128 Severity : enhancement
130 Description: Communicate OST degraded/readonly state via statfs to MDS
131 Details : Flags in the statfs returned from OSTs indicate whether the
132 OST is in a degraded RAID state, or if the filesystem has
133 turned read-only after a filesystem error is detected.
138 Description: don't panic if EPROTO was hit when reading symlink
139 Details : correctly handling request reference in error cases.
141 Severity : enhancement
143 Description: Add LustreNetLink, a kernel-userspace communcation path.
145 Severity : enhancement
147 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
152 Description: access to llog context before init.
153 Details : move handling CATALOGS file at osc layer and forbid access to llog
157 Frequency : rare, only when client is writing to OST on local node
159 Description: Avoid deadlock for local client writes
160 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
161 memory freeing context. This allows OST threads to set the
162 PF_MEMALLOC flag on task structures in order to allocate memory
163 from reserved pools and complete IO.
164 Use GFP_HIGHUSER for OST allocations for non-local client writes,
165 so that the OST threads generate memory pressure and allow
166 inactive pages to be reclaimed.
168 Severity : enhancement
170 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
175 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
176 Details : move ldlm namespace creation in setup phase to avoid grab
177 _lprocfs_lock with cli_sem held.
181 Description: Temporarily disable grant shrink.
182 Details : Disable the feature for debugging.
185 Frequency : only when formatting test filesystems
187 Description: Unable to run several mkfs.lustre on loop devices at the same time
188 Details : mkfs.lustre returns error 256 on the concurrent loop devices
189 formatting. The solution is to proper handle the error.
191 Severity : enhancement
193 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
195 Severity : enhancement
197 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
202 Description: resolve race between obd_disconnect and class_disconnect_exports
203 Details : if obd_disconnect will be called to already disconnected export he
204 forget release one reference and osc module can't unloaded.
206 Severity : enhancement
208 Description: Allow tuning service thread via /proc
209 Details : For each service a new
210 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
211 created that can be used to set min/max thread counts, and get the
212 current number of running threads.
217 Descriptoin: don't return error if have partially created objects for file.
218 Details : lov_update_create_set uses set->set_success as index for created
219 objects, so if some requests will be failed, they will have hole
220 at end of array and qos_shrink_lsm for allocate correct lsm.
222 Severity : enhancement
224 Description: Update OFED support to 1.4.1
229 Description: Reduce small size read RPC
230 Details : Set read-ahead limit for every file and only do read-ahead when
231 available read-ahead pages are bigger than 1M to avoid small size
234 Severity : enhancement
236 Description: move AT tunable parameters for more consistent usage
237 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
239 Severity : enhancement
241 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
242 Details : allow skip disconnected ost for send statfs request and hide error
248 Description: fix racy locking of mballoc block bitmaps causing BUG
249 Details : The locking of the mballoc buddy bitmap and the in-memory
250 block bitmap was using two different spin locks in some
251 cases. This made it possible to incorrectly access the
252 mballoc bitmap while another process was modifying it,
253 causing a sanity assertion to fail. While no on-disk corruption
254 was reported, there was some risk of this happening.
257 Frequency : rare, on llog test 6
259 Descriptoin: don't allow connect to already connected import
260 Details : allowing connect to already connected import is hide connecting problem.
265 Descriptoin: allow kill process which wait statahead result
266 Details : for some reasons 'ls' can stick in waiting result from statahead,
267 in this case need way for kill this process.
269 Severity : enhancement
271 Description: Add state history info file, enhance import info file
272 Details : Track import connection state changes in a new osc/mdc proc file;
273 add overview-type data to the osc/mdc import proc file.
275 Severity : enhancement
277 Description: MDS create should not wait for statfs RPC while holding DLM lock.
279 Severity : enhancement
281 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
286 Descriptoin: OST grant too much space to client even there are not enough space.
287 Details : Client will shrink its grant cache to OST if there are no write
288 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
289 this grant cache if there are already not enough avaible space
290 (left_space < total_clients * 32M).
293 Frequency : start MDS on uncleanly shutdowned MDS device
295 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
296 Details : stay in waiting mds<>ost recovery finished produce random bugs
297 due race between two ll_sync thread for one lov target. send
298 ACTIVATE event only if connect realy finished and import have
302 Frequency : rare, connect and disconnect target at same time
304 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
305 Details : don't call obd_disconnect under lov_lock. this long time
306 operation and can block ptlrpcd which answer to connect request.
311 Descriptoin: don't lose wakeup for imp_recovery_waitq
312 Details : recover_import_no_retry or invalidate_import and import_close can
313 both sleep on imp_recovery_waitq, but we was send only one wakeup
317 Frequency : always with long access acl
319 Descriptoin: mds can't pack reply with long acl.
320 Details : mds don't control size of acl but they limited by reint/getattr
323 Severity : enhancement
325 Description: Update to SLES10 kernel-2.6.16.60-0.33.
327 Severity : enhancement
329 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
332 Frequency : start MDS on uncleanly shutdowned MDS device
334 Descriptoin: aborting recovery hang on MDS
335 Details : don't throttle destroy RPCs for the MDT.
338 Frequency : on remount
340 Description: external journal device not working after the remount
341 Details : clear dev_rdonly flag for external journal devices in
347 Description: shutdown vs evict race
348 Details : client_disconnect_export vs connect request race.
349 if client will evicted at this time - we start invalidate
350 thread without referece to import and import can be freed
354 Frequency : rare, need acl's on inode.
356 Description: client can't handle ost additional correctly
357 Details : if ost was added after client connected to mds client can have
358 hit lnet_try_match_md ... to big messages to wide striped files.
359 in this case need teach client to handle config events about add
360 lov target and update client max ea size at that event.
362 Severity : enhancement
364 Description: Changelogs
365 Details : Changelogs are a lightweight mechanism to track filesystem
366 metadata and namespace changes. The changelog is recorded
367 permanently on the MDTs, and is periodically "consumed" / purged
368 when records are no longer needed.
370 Severity : enhancement
372 Description: compact fld format with extents
373 Details : Store range of seq rather than every seq in FLD. Seq
374 controller update FLD rather than clients. In Case of CMD, mdt0
375 has FLD, all other metadata server act as non persistent proxy
376 for FLD queries and cache fld entries in fld cache.
381 Description: don't skip ost target if they assigned to file
382 Details : Drop slow OSCs if we can, but not for requested start idx.
383 This means "if OSC is slow and it is not the requested
384 start OST, then it can be skipped, otherwise skip it only
385 if it is inactive/recovering/out-of-space.
389 Description: more cleanup in mds_lov
390 Details : not send LOV EA under replay, we can't know about they size at this
391 time. Don't allow client connect to mds before any ost connected,
392 for avoid problems with LOV EA size and returning EIO to client.
394 Severity : enhancement
396 Description: Interoperability at server side (Disk interoperability)
398 Severity : enhancement
400 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
402 Severity : enhancement
404 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
406 Severity : enhancement
408 Description: New client IO stack (CLIO).
410 Severity : enhancement
412 Description: Commit on sharing. Eliminate inter-client dependencies between
413 uncommitted transactions by doing transaction commits.
414 Thereby clients may recovery independently.
417 Frequency : Create a symlink file with a very long name
419 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
420 Details : If there is no extra space in the request for early cancels,
421 ldlm_req_handles_avail() returns 0 instead of a negative value.
423 Severity : enhancement
425 Description: Add /proc entry for import status
426 Details : The mdc, osc, and mgc import directories now have
427 an import directory that contains useful import data for debugging
430 Severity : enhancement
432 Description: Re-disable certain /proc logging
433 Details : Enable and disable client's offset_stats, extents_stats and
434 extents_stats_per_process stats logging on the fly.
437 Frequency : Only on FC kernels 2.6.22+
439 Description: oops in statahead
440 Details : Do not drop reference count for the dentry from VFS when lookup,
441 VFS will do that by itself.
443 Severity : enhancement
445 Description: Generic /proc file permissions
446 Details : Set /Proc file permissions in a more generic way to enable non-
447 root users operate on some /proc files.
451 Description: Hitting mdc_commit_close() ASSERTION
452 Details : Properly handle request reference release in
453 ll_release_openhandle().
457 Description: quota recovery deadlock during mds failover
458 Details : This patch includes att18982, att18236, att18237 in bz14840.
460 1. fix osts hang when mds does failover with quotaon
461 2. prevent watchdog storm when osts threads wait for the
466 Frequency : only patchless client
467 Description: add workaround for race between add/remove dentry from hash
469 Severity : enhancement
471 Description: Allow OST glimpses to return PW locks
475 Description: LBUG when llog conf file is full
476 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
480 Description: Prevent import from entering FULL state when server in recovery
484 Description: service mount cannot take device name with ":"
485 Details : Only when device name contains ":/" will mount treat it as
491 Description: replace ptlrpcd with the statahead thread to interpret the async
492 statahead RPC callback
496 Frequency : on recovery
497 Description: I/O failures after umount during fail back
498 Details : if client reconnected to restarted server we need join to recovery
499 instead of find server handler is changed and process self
500 eviction with cancel all locks.
502 Severity : enhancement
504 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
506 Severity : enhancement
508 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
510 Severity : enhancement
512 Description: Upcall on Lustre log has been dumped
513 Details : Allow for a user mode script to be called once a Lustre log has
514 been dumped. It passes the filename of the dumped log to the
515 script, the location of the script can be specified via
516 /proc/sys/lnet/debug_log_upcall.
521 Description: avoid idr_remove called for id which is not allocated.
522 Details : Move assigment s_dev for clustered nfs to end of initialization,
523 to avoid problem with error handling.
528 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
529 Details : When connection is reused this not moved from CONN_UNUSED_HASH
530 into CONN_USED_HASH and this prodice warning when put connection
531 again in unused hash.
534 Severity : enhancement
536 Description: File striping can now be set to use an arbitrary pool of OSTs.
538 Severity : enhancement
540 Description: Export bytes_read/bytes_write count on OSC/OST.
544 Description: Early reply size mismatch, MGC loses connection
545 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
546 the connect flags are properly negotiated.
551 Description: quotas are not honored with O_DIRECT
552 Details : all writes with the flag O_DIRECT will use grants which leads to
553 this problem. Now using OBD_BRW_SYNC to guard this.
557 Description: add quota statistics
558 Details : 1. sort out quota proc entries and proc code.
559 2. add quota statistics
561 Severity : enhancement
563 Description: enable quota support for HEAD.
567 Description: Properly propagate oinfo flags from lov to osc for statfs
568 Details : restore missing copy oi_flags to lov requests.
570 Severity : enhancement
572 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
573 : l_getgroups(8), lst(8), routerstat(8)
574 Details : included man pages for llobdstat(8), llstat(8),
575 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
577 Severity : enhancement
579 Description: configure's --enable-quota should check the
580 : kernel .config for CONFIG_QUOTA
581 Details : configure is terminated if --enable-quota is passed but
582 : no quota support is in kernel
586 Description: Remove portals compatibility
587 Details : Remove portals compatibility, not interoperable with releases
592 Description: Resolve device initialization race
593 Details : Prevent proc handler from accessing devices added to the
594 obd_devs array but yet be intialized.
596 Severity : enhancement
598 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
600 Severity : enhancement
602 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
607 Description: Using wrong pointer in osc_brw_prep_request
608 Details : Access to array[-1] can produce panic if kernel compiled with
609 CONFIG_PAGE_ALLOC enabled
613 Description: Client runs out of low memory
614 Details : Consider only lowmem when counting initial number of llap pages
618 Description: *optional* service tags registration
619 Details : if the "service tags" package is installed on a Lustre node
620 When the filesystem is mounted, a local-node service tag will
621 be created. See http://inventory.sun.com/ for more information
622 about the Service Tags asset management system.
626 Description: Kernel BUG tries to release flock
627 Details : Lustre does not destroy flock lock before last reference goes
628 away. So always drop flock locks when client is evicted and
629 perform unlock regardless of successfulness of speaking to MDS.
633 Description: add refcount protection for osc callbacks, avoid panic on shutdown
637 Description: sanity test 65a fails if stripecount of -1 is set
638 Details : handle -1 striping on filesystem in ll_dirstripe_verify
643 Description: ASSERTION(CheckWriteback(page,cmd)) failed
644 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
647 Severity : enhancement
649 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
653 Description: do not process already freed flock
654 Details : flock can possibly be freed by another thread before it reaches
655 to ldlm_flock_completion_ast.
659 Description: LBUG during stress test
660 Details : Need properly lock accesses the flock deadlock detection list.
664 Description: oops in page fault handler
665 Details : kernel page fault handler can return two special 'pages' in error
666 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
670 Description: timeout with invalidate import.
671 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
672 should be handled by ptlrpcd. This produce long age waiting and
673 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
675 Severity : enhancement
677 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
681 Description: enable MGS and MDT services start separately
682 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
683 a co-located MGS without starting the MGS, which is a complement
684 to 'nosvc' mount option.
688 Frequency : after recovery
689 Description: precreate to many object's after del orphan.
690 Details : del orphan st in oscc last_id == next_id and this triger growing
691 count of precreated objects. Set flag LOW to skip increase count
692 of precreated objects.
696 Frequency : rare, on clear nid stats
697 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
698 Details : when clean nid stats sometimes try destroy live entry,
699 and this produce panic in free.
703 Description: Stack overflow during MDS log replay
704 ease stack pressure by using a thread dealing llog_process.
708 Description: wait until IO finished before start new when do lock cancel.
709 Details : VM protocol want old IO finished before start new, in this case
710 need wait until PG_writeback is cleared until check dirty flag and
711 call writepages in lock cancel callback.
713 Severity : enhancement
715 Description: using special macro for print time and cleanup in includes.
719 Description: mds_mfd_close() ASSERTION(rc == 0)
720 Details : In mds_mfd_close(), we need protect inode's writecount change
721 within its orphan write semaphore to prevent possible races.
725 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
729 Frequency : rare, on shutdown ost
730 Description: don't hit live lock with umount ost.
731 Details : shrink_dcache_parent can be in long loop with destroy dentries,
732 use shrink_dcache_sb instead.
736 Description: don't panic with use echo client
737 Details : echo client pass NULL as client nid pointer and this produce null
742 Description: fix build on ppc32
743 Details : compile code with -m64 flag produce wrong object file for ppc32.
747 Description: add message levels for liblustreapi
751 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
752 Details : Move the 'good_osts' check before the 'total_bavail' check. This
753 will result in an -EAGAIN and in the exit call path we call
754 alloc_rr() which will with increasing aggressiveness attempt to
755 aquire precreated objects on the minimum number of required OSCs.
759 Description: Use old size assignment to avoid deadlock
760 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
761 the scheduling while holding a spinlock. We do not need locking
762 for size in ll_update_inode() because size is only updated from
763 the MDS for directories or files without objects, so there is no
764 other place to do the update, and concurrent access to such inodes
765 are protected by the inode lock.
769 Description: resolve "_IOWR redefined" build error on SLES10
773 Description: dump the memory debugging after all modules are unloaded to
774 suppress false negative in conf_sanity test 39
776 Severity : enhancement
778 Description: build kernel-ib packages for OFED 1.3 in our release cycle
783 Description: fix SLES kernel versioning
784 Details : the kernel version for our SLES 10 kernel did not include a "-"
785 before the "smp" at the end. while this was not a problem in
786 general, it did mean that software trying to use the kernel
787 version to try to detect a vendor specific kernel would fail.
788 this was most evident by the OFED build scripts.
792 Description: Don't update lov_desc members until making sure they are valid
793 Details : When updating lov_desc members via proc fs, need fix their
794 validities before doing the real update.
798 Description: don't put request into delay list while invalidate in flight.
799 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
800 import in flight. this produce timeout for invalidate and sometimes
801 can cause stale data.
805 Frequency : on ppc only
806 Description: not convert ost objects for directory because it's not exist.
807 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
811 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
813 Severity : enhancement
815 Description: Update to support 2.6.22.14 vanilla kernel.
819 Frequency : rare, on recovery
820 Description: read procfs can produce deadlock in some situation
821 Details : Holding lprocfs lock which send rpc can produce block for destroy
822 obd objects and this also block reconnect with -EALREADY. This
823 isn't fix all lprocfs bugs - but make it rare.
825 Severity : enhancement
827 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
830 Frequency : frequent on X2 node
832 Description: mdc_set_open_replay_data LBUG
833 Details : Set replay data for requests that are eligible for replay.
837 Description: lustre_mgs: operation 101 on unconnected MGS
838 Details : When MGC is disconnected from MGS long enough, MGS will evict the
839 MGC, and late on MGC cannot successfully connect to MGS and a lot
840 of the error messages complaining that MGS is not connected.
844 Frequency : on network error
845 Description: panic with double free request if network error
846 Details : mdc_finish_enqueue is finish request if any network error ocuring,
847 but it's true only for synchronus enqueue, for async enqueue
848 (via ptlrpcd) this incorrect and ptlrpcd want finish request
851 Severity : enhancement
853 Description: client-side metadata stat-ahead during readdir(directory readahead)
854 Details : perform client-side metadata stat-ahead when the client detects
855 readdir and sequential stat of dir entries therein
858 Frequency : on start mds
860 Description: Implement get_info(last_id) in obdfilter.
863 Frequency : occasional
865 Description: Correctly check stale fid, not start epoch if ost not support SOM
866 Details : open with flag O_CREATE need set old fid in op_fid3 because
867 op_fid2 was overwritten with new generated fid, but mds can answer
868 with one of these two fids and both is not stale. Setattr
869 incorrectly started an epoch and assume will be called
870 done_writing, but without SOM done_writing ever being called.
873 Frequency : rare, depends on device drivers and load
875 Description: MDS or OSS nodes crash due to stack overflow
876 Details : Code changes in 1.8.0 increased the stack usage of some functions.
877 In some cases, in conjunction with device drivers that use a lot
878 of stack the MDS (or possibly OSS) service threads could overflow
879 the stack. One change which was identified to consume additional
880 stack has been reworked to avoid the extra stack usage.
883 Frequency : occasional
885 Description: Do not fail import if osc_interpret_create gets -EAGAIN
886 Details : If osc_interpret_create got -EAGAIN it immediately exits and
887 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
888 oscc_has_objects and see OSC has no objests and call
889 oscc_internal_create to resend create request.
891 Severity : enhancement
893 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
895 Severity : enhancement
897 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
900 Frequency : very rare
902 Description: Fix replay if there is an un-replied request and open
903 Details : In some cases, older replay request will revert the
904 mcd->mcd_last_xid on MDS which is used to record the client's
907 Severity : enhancement
909 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
911 Severity : enhancement
913 Description: Add rhel5 support to HEAD.
915 Serverity : enhancement
917 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
922 Description: Don't allow skipping OSTs if index has been specified.
923 Details : Don't allow skipping OSTs if index has been specified, make
924 locking in internal create lots better.
928 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
929 Details : access to bitfield in structure is always rounded to long
930 and this produce problem with not atomic change any bit.
934 Description: Lustre make rpms failed.
935 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
936 builds Lustre rpms from tarball.
939 Frequency : rare on shutdown ost
941 Description: If llog cancel was not send before clean_exports phase, this can
942 produce deadlock in llog code.
943 Details : If llog thread has last reference to obd and call class_import_put
944 this produce deadlock because llog_cleanup_commit_master wait when
945 last llog_commit_thread exited, but this never success because was
946 called from llog_commit_thread.
950 Description: allow userland application know is lost one of stripes.
951 Details : fill lvb_blocks with error code on ost and return it to
952 application if error flag found.
956 Description: NULL lov_tgts causing MDS oops
957 Details : more safe checks for NULL lov_tgts for avoid oops.
959 Severity : enhancement
961 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
965 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
966 Details : If an OST is down the MDS will hang indefinitely in
967 obd_statfs_rqset() waiting for the statfs data. While for
968 MDS QOS usage of statfs, it should not stuck in waiting.
970 Severity : enhancement
972 Description: remote_acl support
973 Details : Support ACL-based permission check for remote user.
974 Support setfacl/getfacl for remote user with the utils
975 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
976 the system "{s,g}etfacl" utils.
978 Severity : enhancement
980 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
982 Severity : enhancement
984 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
988 Description: make mgs_setparam() handle fsname containing dash
989 Details : fsname containing a dash does not work with lctl conf_param
991 Severity : enhancement
993 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
995 Severity : enhancement
997 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1001 Frequency : rare, at shutdown
1002 Description: access already free / zero obd_namespace.
1003 Details : if client_disconnect_export was called without force flag set,
1004 and exist connect request in flight, this can produce access to
1005 NULL pointer (or already free pointer) when connect_interpret
1006 store ocd flags in obd_namespace.
1010 Frequency : only at startup
1011 Description: not alloc memory with spinlock held.
1012 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1013 if any spinlock held.
1015 Severity : enhancement
1017 Description: make lustre randomly fail allocating memory
1018 Details : Make lustre randomly failed allocating memory for testing purpose.
1020 Severity : enhancement
1022 Description: lost problems with lov objid file
1023 Details : Fixes some scability and access to not inited memory problems
1024 in work with lov objdid file.
1029 Description: lfs find does not continue on file error
1030 Details : Continue other files processing when a file/dir is absent.
1034 Description: Inconsistent usage of lustre_pack_reply()
1035 Details : Standardize the usage of lustre_pack_reply() such that it
1036 always generate a CERROR on failure.
1039 Frequency : occasional
1041 Description: MDS hang or stay in waiting lock
1042 Details : If client receive lock with CBPENDING flag ldlm need send lock
1043 cancel as separate rpc, to avoid situation when cancel request
1044 can't processed due all i/o threads stay in wait lock.
1048 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1052 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1056 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1060 Description: Update to RHEL4 latest kernel.
1062 Severity : enhancement
1064 Description: Build SLES10 patchless client fails
1065 Details : The configure was broken by run ./configure with
1066 --with-linux-obj=.... argument for patchless client. When the
1067 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1068 can't search header adequately. Use absolute path such as
1069 -I($LINUX)/include instead.
1073 Description: interrupt oig_wait produce painc on resend.
1074 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1075 private set, and this produce situation when rq_ptlrpcd_data not
1076 copyed to new allocated request and triggered LBUG on assert
1077 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1078 ptlrpcd set if request is changed and can be safety changed to use
1083 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1084 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1085 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1086 these macros and replace them with their difinitions instead.
1090 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1091 Details : Update which_patch & target file for SLES10 latest kernel.
1094 Description: Remove client patches from SLES 10 kernel.
1095 Details : This causes SLES 10 clients to behave as patchless clients
1096 even on a Lustre-patched (server) kernel.
1098 Severity : enhancement
1100 Description: self-adjustable client's lru lists
1101 Details : use adaptive algorithm for managing client cached locks lru
1102 lists according to current server load, other client's work
1103 pattern, memory activities, etc. Both, server and client
1104 side namespaces provide number of proc tunables for controlling
1107 Severity : enhancement
1109 Description: light-weight GSS support
1110 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1111 and reduce performance overhead.
1113 Severity : enhancement
1115 Description: Linux keyring support
1116 Details : Support using service of Linux keyring for Lustre GSS internal
1117 context refresh/cache mechanism.
1121 Description: Fix errors in lfs documentation
1122 Details : Fixes man pages
1126 Description: don't use GFP_* in generic Lustre code.
1127 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1131 Description: obdclass is limited by single OBD_ALLOC(idarray)
1132 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1136 Description: warnings with build patchless client with vanila 2.6.19 and up
1137 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1138 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1142 Frequency : Only for SLES
1143 Description: sanity_quota fail test_1
1144 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1145 file, which makes TSTID[2] inunique.
1149 Frequency : Only for SLES9
1150 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1151 Details : Update target/ChangeLog/which_patch .
1155 Frequency : Only for test_10 in sanity-quota.sh
1156 Description: a bug in quota test code
1157 Details : checking a test flag in a wrong place
1161 Frequency : Only for test_13 in sanity-quota.sh
1162 Description: a bug in quota test script
1163 Details : shouldn't check overfull allocated quota
1167 Description: jbd statistics
1168 Details : Port older jbd statistics patch for sles10
1170 Severity : enhancement
1172 Description: Add printing inode info into message about error in writepage.
1177 Description: reply_lock_interpret crash due to race with it and lock cancel.
1178 Details : Do not replay locks that are being cancelled. Do not reference
1179 locks by their address during replay, just by their handle.
1183 Frequency : When flocks are used.
1184 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1185 Details : Flock locks might destroy just granted lock if it could be merged
1186 with another existing flock, this is done in completion handler,
1187 so teach ldlm_cli_enquque_fini that this is a valid case for
1193 Description: Oops in read and write path when failing to allocate lock.
1194 Details : Check if lock allocation failed and return error back.
1198 Description: lstripe command fails for valid OST index
1199 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1200 instead of lov->desc.ld_active_tgt_count.
1204 Description: sanity.sh failed test 103
1205 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1210 Description: df doesn't work properly if diskfs blocksize != 4K
1211 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1215 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1216 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1218 Severity : enhancement
1220 Description: Adaptive timeouts
1221 Details : RPC timeouts adapt to changing server load and network
1222 conditions to reduce resend attempts and improve recovery time.
1226 Description: llapi_file_create() does not allow some changes
1227 Details : add llapi_file_open() that allows specifying the mode and
1228 open flags, and also returns an open file handle.
1232 Description: merge and cleanup kernel patches.
1233 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1237 Description: Add journal checksum support.(Kernel part)
1238 Details : The journal checksum feature adds two new flags i.e
1239 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1240 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1241 indicates that the commit block contains the checksum for
1242 the blocks described by the descriptor blocks. Now commit
1243 record can be sent to disk without waiting for descriptor
1244 blocks to be written to disk. This behavior is controlled
1245 using JBD2_FEATURE_ASYNC_COMMIT flag.
1249 Description: OSS needs mutliple precreate threads
1250 Details : Add ability to start more than one create thread per OSS.
1254 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1255 Details : execute lfs setstripe on client
1259 Description: mds_obd_create error creating tmp object
1260 Details : When the user sets quota on root, llog will be affected and can't
1261 create files and write files.
1264 Frequency : Always on ia64 patchless client, and possibly others.
1266 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1267 Details : This allows the patchless client to be loaded on architectures
1268 without this export.
1272 Description: RedHat Update kernel for RHEL5
1273 Details : Add the kernel config file for RHEL5.
1277 Description: RedHat Update kernel for RHEL5
1278 Details : Modify the kernel config file more closer RHEL5.
1282 Description: Build failure against Centos5 (RHEL5)
1283 Details : Define PAGE_SIZE when it isn't present.
1287 Description: test_7 on sanity-quota.sh can't be used on separate machines
1288 Details : add facet to handle it
1292 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1293 Details : with NFS, the anon dentry's parent was set to itself in
1294 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1295 corresponding dentry other than use rec->ur_name.
1297 Severity : enhancement
1299 Description: lfs setstripe enhancement
1300 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1304 Description: enable data checksumming by default
1305 Details : enable checksum by default, allow --disable-checksum
1306 configure option and "-o nochecksum" mount option. Checksums
1307 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1311 Description: lustre support for RHEL5
1312 Details : Add support for RHEL5.
1316 Description: Client eviction due to failover config
1317 Details : after a connection loss, the lustre client should attempt to
1318 reconnect to the last active server first before trying the
1319 other potential connections.
1323 Description: when mds and osts use different quota unit(32bit and 64bit),
1324 quota will be released repeatly.
1325 Details : void sending multiple quota reqs to mds, which will keep the
1326 status between the reqs.
1330 Description: osts not allocated evenly to files
1331 Details : change the condition to increase offset_idx
1334 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1336 Description: Data corruption for OSTs that are formatted larger than 2TB
1338 Details : When generating the bio request for lustre file writes the
1339 sector number would overflow a temporary variable before being
1340 used for the IO. The data reads correctly from Lustre (which
1341 will overflow in a similar manner) but other file data or
1342 filesystem metadata may be corrupted in some cases.
1346 Description: Tune the kernel for good SCSI performance.
1347 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1348 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1353 Description: rewrite ext2-derived code in obdclass/uuid.c
1354 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1355 from scratch preserving functionality.
1359 Description: "lfs find -obd UUID" prints directories
1360 Details : "lfs find -obd UUID" will return all directory names instead
1361 of just file names. It is incorrect because the directories
1362 do not reside on the OSTs.
1366 Description: use i_size_read and i_size_write in 2.6 port
1367 Details : replace inode->i_size access with i_size_read/write()
1370 Frequency : when removing large files
1372 Description: scheduling issue during removal of large Lustre files
1373 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1374 It causes scheduling issues when removing large files (17TB in the
1378 Frequency : only with liblustre clients on XT3
1380 Description: evictions taking too long
1381 Details : allow llrd to evict clients directly on OSTs
1384 Frequency : only on ppc
1386 Description: /proc/fs/lustre/devices broken on ppc
1387 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1388 We should make sure the type of 'index' is loff_t and then cast
1389 explicitly as needed below. Do not assign an explicitly cast
1393 Frequency : only for rhel5
1395 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1396 Details : Modify the target file & which_kernel.
1398 Severity : enhancement
1400 Description: omit set fsid for export NFS
1401 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1402 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1406 Frequency : Always, for kernels after 2.6.16
1407 Description: Fix warning idr_remove called for id=.. which is not allocated.
1408 Details : Last kernels save old s_dev before kill super and not allow
1409 to restore from callback - restore it before call kill_anon_super.
1413 Description: buffer overruns could theoretically occur
1414 Details : llapi_semantic_traverse() modifies the "path" argument by
1415 appending values to the end of the origin string, and a
1416 overrun may occur. Adding buffer overrun check in liblustreapi.
1420 Description: Fix error on 'ls .' at the top of the Lustre mount.
1421 Details : Don't revalidate dentry if it is a root dentry.
1425 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1426 Details : Modify vm-tunables-rhel4.patch.
1430 Description: Kernel config for 2.6.18-vanilla.
1431 Details : Modify targets/2.6-vanilla.target.in.
1432 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1433 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1434 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1435 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1439 Description: improve handling recoverable errors
1440 Details : if request processig with error which can be recoverable on server
1441 request should be resend, otherwise page released from cache and
1446 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1447 Details : Modify target file & which_patch.
1448 A flaw was found in the IA32 system call emulation provided
1449 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1450 value could be stored in the %RAX register, which could trigger an
1451 out-of-bounds system call table access. An untrusted local user
1452 could exploit this flaw to run code in the kernel
1453 (ie a root privilege escalation). (CVE-2007-4573).
1457 Description: O_DIRECT bypasses client statistics.
1458 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1459 read_ahead_stats were updated. copied the stats section from
1460 osc_send_oap_rpc() into async_internal().
1464 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1468 Description: change order of libsysio includes
1469 Details : '#include sysio.h' should always come before '#include xtio.h'
1473 Description: handle "serious error: objid * already exists" more gracefully
1474 Details : If LAST_ID value on disk is smaller than the objects existing in
1475 the O/0/d* directories, it indicates disk corruption and causes an
1476 LBUG(). If the object is 0-length, then we should use the existing
1477 object. This will help to avoid a full fsck in most cases.
1479 Severity : enhancement
1481 Description: adapt the lustre_config script to support the upgrade case
1482 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1483 server targets to 1.6 in parallel.
1487 Description: To avoid grant space > avaible space when the disk is almost
1488 full. Without this patch you might see the error "grant XXXX >
1489 available" or some LBUG about grant, when the disk is almost
1491 Details : In filter_check_grant, for non_grant cache write, we should
1492 check the left space by if (*left > ungranted + bytes), instead
1493 of (*left > ungranted), because only we are sure the left space
1494 is enough for another "bytes", then the ungrant space should be
1495 increase. In client, we should update cl_avail_grant only there
1496 is OBD_MD_FLGRANT in the reply.
1500 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1501 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1502 described in CVE-2007-4573. This problem could allow untrusted
1503 local users to gain root access.
1506 Frequency : when using O_DIRECT and quotas
1508 Description: Incorrect file ownership on O_DIRECT output files
1509 Details : block usage reported by 'lfs quota' does not take into account
1510 files that have been written with O_DIRECT.
1515 Description: touch file failed when fs is not full
1516 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1517 otherwise we can get ENOSP while fs is not full.
1521 Description: parallel lock callbacks
1522 Details : Instead of sending blocking and completion callbacks as separated
1523 requests, adding them to a set and sending in parallel.
1526 Frequency : only for Cray XT3
1527 Bugzilla : 12829/13455
1528 Description: Changing primary group doesn't change the group lustre assigns to
1530 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1531 overridden with the primary group provided by the group upcall,
1532 whereas the supplied fsgid can be trusted if it is in the list of
1533 supplementary groups returned by the group upcall.
1535 Severity : enhancement
1537 Description: Allow masking D_WARNING, D_ERROR messages from console
1538 Details : Console messages can now be disabled via lnet.printk.
1542 Description: User code with malformed file open parameter crashes client node
1543 Details : Before packing join_file req, all the related reference should be
1544 checked carefully in case some malformed flags cause fake
1545 join_file req on client.
1549 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1550 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1551 mode is equal to the granted mode. However, it should also check
1552 that LDLM_FL_CP_REQD is not set.
1556 Description: Heavy nfs access might result in deadlocks
1557 Details : After ELC code landed, it is now improper to enqueue any mds
1558 locks under och_sem, because enqueue might want to decide to
1559 cancel open locks for same inode we are holding och_sem for.
1563 Description: Client eviction while running blogbench
1564 Details : A lot of unlink operations with concurrent I/O can lead to a
1565 deadlock causing evictions. To address the problem, the number of
1566 oustanding OST_DESTROY requests is now throttled to
1567 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1568 callbacks are processed in priority.
1572 Description: enable ACLs on MDS by default
1573 Details : ACLs must be enabled on MDS by default.
1576 Frequency : PPC/PPC64 only
1578 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1579 Details : fix duplicated definitions between asm-ppc64/types.h and
1580 lustre_types.h on PPC.
1583 Frequency : PPC/PPC64 only
1585 Description: asm-ppc/segment.h does not exist
1586 Details : fix compile issue on PPC.
1590 Description: better handle error messages in extents code
1593 Frequency : RHEL4 only
1595 Description: mkfs is very slow on IA64/RHEL4
1596 Details : A performance regression has been discovered in the MPT Fusion
1597 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1598 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1599 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1601 Severity : enhancement
1603 Description: SNMP support enhancement
1604 Details : Adding total number of sampled request for an MDS node in snmp
1607 Severity : enhancement
1609 Description: Optimize ldlm waiting list processing for PR extent locks
1610 Details : When processing waiting list for read extent lock and meeting read
1611 lock that is same or wider to it that is not contended, skip
1612 processing rest of the list and immediatelly return current
1613 status of conflictness, since we are guaranteed there are no
1614 conflicting locks in the rest of the list.
1618 Description: Time out and refuse to reconnect
1619 Details : When the failover node is the primary node, it is possible
1620 to have two identical connections in imp_conn_list. We must
1621 compare not conn's pointers but NIDs, otherwise we can defeat
1622 connection throttling.
1626 Description: port llog fixes from b1_6 into HEAD
1627 Details : Port llog reference couting and some llog cleanups from b1_6
1628 (bug 10800) into HEAD, for protect from panic and access to already
1629 free llog structures.
1633 Description: Detect stride IO mode in read-ahead
1634 Details : When a client does stride read, read-ahead should detect that and
1635 read-ahead pages according to the detected stride pattern.
1639 Description: data checksumming impacts single node performance
1640 Details : add support for several checksum algorithm. Currently, only CRC32
1641 and Adler-32 are supported. The checksum type can be changed on
1642 the fly via /proc/fs/lustre/osc/*/checksum_type.
1646 Description: use adler32 for page checksums
1647 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1652 Description: build for x2 fails
1653 Details : fix compile issue on Cray systems.
1657 Description: Properly match for duplicate locks
1658 Details : Due to different lock order from skiplists code, we need to
1659 traverse entire list for now
1662 Frequency : only on PPC/SLES10
1664 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1665 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1666 be defined. Add a hack in mkfs_lustre.c to work around this compile
1671 Description: LASSERT on MDS when client holding flock lock dies
1672 Details : ldlm pool logic depends on number of granted locks equal to
1673 number of released locks which is not true for flock locks, so
1674 just exclude such locks from consideration.
1678 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1679 Details : Use fsfilt_sync() for both the whole filesystem sync and
1680 individual file sync to eliminate dangerous inode locking
1681 with I_LOCK that can lead to a deadlock.
1685 Description: performance in 1.6.3
1686 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1687 sure that 1MB requests are not fragmented by the block layer.
1689 Severity : enhancement
1691 Description: organize the server-side client stats on per-nid basis
1692 Details : Change the structure of stats under obdfilter and mds to
1700 The "uuid"s file would list the uuids of _active_ exports.
1701 And the clear entry is to clear all stats and stale nids.
1703 Severity : enhancement
1705 Description: eliminate client locks in face of contention
1706 Details : file contention detection and lockless i/o implementation
1707 for contended files.
1711 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1714 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1716 Description: MDS may lose file striping (and hence file data) in some cases
1717 Details : If there are additional extended attributes stored on the MDS,
1718 in particular ACLs, SELinux, or user attributes (if user_xattr
1719 is specified for the client mount options) then there is a risk
1720 of attribute loss. Additionally, the Lustre file striping
1721 needs to be larger than default (e.g. striped over all OSTs),
1722 and an additional attribute must be stored initially in the
1723 inode and then increase in size enough to be moved to the
1724 external attribute block (e.g. ACL growing in size) for file
1729 Description: skiplist implementation simplification
1730 Details : skiplists are used to group compatible locks on granted list
1731 that was implemented as tracking first and last lock of each
1732 lock group the patch changes that to using doubly linked lists
1736 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1737 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1738 because LASSERT on both the data supplied by a client, and the
1739 data on disk is dangerous and incorrect.
1741 Severity : enhancement
1743 Description: Slow truncate/writes to huge files at high offsets.
1744 Details : Directly associate cached pages to lock that protect those pages,
1745 this allows us to quickly find what pages to write and remove
1746 once lock callback is received.
1750 Description: more ldlm soft lockups
1751 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1752 starve other threads from the resource lock for a long time in
1753 case of long waiting queue, so change the debug level from
1754 D_OTHER to the less frequently used D_INFO.
1756 Severity : enhancement
1758 Description: add -gid, -group, -uid, -user options to lfs find
1762 Description: Hung threads in invalidate_inode_pages2_range
1763 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1764 once one is completed. As a result, some RPCs are stuck in the
1765 queue and are never sent.
1769 Description: filter threads hungs on waiting journal commit
1770 Details : Cleanup filter group llog code, then only filter group llog will
1771 be only created in the MDS/OST syncing process.
1775 Description: Procfs and llog threads access destoryed import sometimes.
1776 Details : Sync the import destoryed process with procfs and llog threads by
1777 the import refcount and semaphore.
1779 Severity : enhancement
1781 Description: openlock cache of b1_6 port to HEAD
1786 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1787 Details : If insertion of an extent fails, then discard the inode
1788 preallocation and free data blocks else it can lead to duplicate
1793 Description: don't always update ctime in ext3_xattr_set_handle()
1794 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1795 In some cases the ctime should not be updated, for example for
1796 2.0->1.8 compatibility it is necessary to delete an xattr and it
1797 should not update the ctime.
1801 Bugzilla : 15713/16362
1802 Description: Assertion in iopen_connect_dentry in 1.6.3
1803 Details : looking up an inode via iopen with the wrong generation number can
1804 populate the dcache with a disconneced dentry while the inode
1805 number is in the process of being reallocated. This causes an
1806 assertion failure in iopen since the inode's dentry list contains
1807 both a connected and disconnected dentry.
1811 Description: assertion failure in ldlm_handle2lock()
1812 Details : fix a race between class_handle_unhash() and class_handle2object()
1813 introduced in lustre 1.6.5 by bug 13622.
1818 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1819 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1824 Description: oops in sd_iostats_seq_show()
1825 Details : unloading/reloading the scsi low level driver triggers a kernel
1826 bug when trying to access the sd iostat file.
1831 Description: Kernel panics during QLogic driver reload
1832 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1833 patch, causing memory corruption.
1838 Description: journal_dev option does not work in b1_6
1839 Details : pass mount option during pre-mount.
1841 Severity : enhancement
1843 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1844 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1845 extent information of a file. It can be used to map logical blocks
1846 in a file to physical blocks in the block device.
1850 Description: LDLM soft lockups - improvement
1851 Details : It is be possible to send the lock handle along with each read
1852 or write request because the client is already doing a lock match
1853 itself so there isn't any reason the OST should have to re-do that
1857 Frequency : only on Cray X2
1859 Description: X2 build failures
1860 Details : fix build failures on Cray X2.
1864 Description: xid & resent requests
1865 Details : Initialize RPC XID from clock at startup (randomly if clock is
1868 Severity : enhancement
1870 Description: Add lustre_start utility to start or stop multiple Lustre servers
1875 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1876 Details : In case of memory pressure, list_del() can be called twice on
1877 req->rq_history_list, causing a kernel oops.
1881 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1882 Details : fix stack overflow in the distributed lock manager by defering
1883 export eviction after a failed AST to the elt thread instead of
1884 handling it in the dlm interpret routine.
1888 Description: Convert some comments to new format.
1889 Details : Update documenting comments to match doxygen conventions.
1893 Description: Grammar fixes.
1894 Details : A couple of trivial spelling fixes.
1898 Description: OSD_COUNTERS-mandatory
1899 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1900 to simplify the code.
1904 Description: simplify cmm_device freeing logic.
1905 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1909 Description: Add lockdep support to dt_object_operations locking interface.
1910 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1911 parameter indicating lock ordering. Update mdd code to use new
1916 Description: Introduce failloc constants for lockless IO tests.
1917 Details : Add two new failloc constants to test lockless IO. Only one of
1918 them in implemented---another is checked in yet to be landed
1923 Description: Add lockdep support for inode mutex.
1924 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1925 in the situations where Lustre has to lock more than one inode
1930 Description: Add optional invariants checking support.
1931 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1932 configure switch. This macro is to be used for consistency and
1933 sanity checks that are too expensive to be left in `production'
1938 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1939 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1944 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1945 Details : Use spin_lock_nested() in (the only) situation where more than
1946 one ldlm_lock is locked simultaneously. Also, fix possible
1947 dead-lock in ldlm_lock_change_resource() by enforcing particular
1952 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1953 Details : Instead of specifying each ldlm_lock call-back through separate
1954 parameter, wrap them into struct ldlm_callback_suite.
1958 Description: Kill join_lru obd method and its callers.
1959 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1960 in memory---a requirement implemented through ->o_join_lru() obd
1961 method in HEAD. Remove this method and its users.
1965 Description: Add asynchronous ldlm ENQUEUE completion handler.
1966 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1967 case that stock ldlm_completion_ast() cannot handle as it waits
1968 until lock is granted. Introduce new ldlm_completion_ast_async()
1969 for this. Also comment ldlm_completion_ast().
1973 Description: ldlm_error <-> errno conversion.
1974 Details : Add functions to map (rather arbitrary) between LDLM error codes
1975 and standard errno values. CLIO needs this to prevent LDLM specific
1976 constants from escaping ldlm and osc.
1980 Description: Kill unused ldlm_handle2lock_ns() function.
1981 Details : Kill unused ldlm_handle2lock_ns() function.
1985 Description: Add lu_ref support to ldlm_lock
1986 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1987 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1988 and ldlm_resource. LDLM interface has to be changed, because code that
1989 releases a reference on a lock, has to "know" what reference this is.
1990 In the most frequent case
1992 lock = ldlm_handle2lock(handle);
1994 LDLM_LOCK_PUT(lock);
1996 no changes are required. When any other reference (received _not_ from
1997 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1998 instead of LDLM_LOCK_PUT().
2000 Arguably, changes are pervasive, and interface requires some discipline
2001 for proper use. On the other hand, it was very instrumental in finding
2002 a few leaked lock references.
2006 Description: Add ldlm_lock_addref_try().
2007 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2008 attempts to addref a lock that might be being canceled
2013 Description: Add ldlm_weigh_callback().
2014 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2015 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2016 hard-coded `number of pages' logic.
2020 Description: Add lockdep annotations to llog code.
2021 Details : Use appropriately tagged _nested() locking calls in the places
2022 where llog takes more than one ->lgh_lock lock.
2026 Description: Add loi_kms_set().
2027 Details : Wrap kms updates into a helper function.
2031 Description: Constify instances of struct lsm_operations.
2032 Details : Constify instances of struct lsm_operations.
2036 Description: lu_conf support.
2037 Details : On a server, a file system object is uniquely identified
2038 by a fid, which is sufficient to locate and load all object
2039 state (inode). On a client, on the other hand, more data are
2040 necessary instantiate an object. Change lu_object_find() and
2041 friends to take additional `lu_conf' argument describing object.
2042 Typically this includes layout information.
2046 Description: lu_context fixes.
2047 Details : Introduce new lu_context functions that are needed on the client
2048 side, where some system threads (ptlrpcd) are shared by multiple
2049 modules, and so cannot be stopped during module shutdown.
2053 Description: Add start and stop methods to lu_device_type_operations.
2054 Details : Introduce two new methods in lu_device_type_operations, that are
2055 invoked when first instance of a given type is created and last one
2056 is destroyed respectively. This is need by CLIO.
2060 Description: Add lu_ref support to struct lu_device.
2061 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2062 track leaked references.
2066 Description: Introduce lu_kmem_descr.
2067 Details : lu_kmem_descr and its companion interface allow to create
2068 and destroy a number of kmem caches at once.
2072 Description: Fix lu_object finalization race.
2073 Details : Fix a race between lu_object_find() finding an object and its
2074 concurrent finalization. This race is (most likely) not possible
2075 on the server, but might happen on the client.
2079 Description: Introduce lu_ref interface.
2080 Details : lu_ref is a debugging module allowing to track references to
2081 a given object. It is quite cpu expensive, and has to be
2082 explicitly enabled with --enable-lu_ref. See usage description
2087 Description: Factor lu_site procfs stats into a separate function.
2088 Details : Separate lu_site stats printing code into a separate function
2089 that can be reused on a client.
2093 Description: Constify instances of struct {lu,dt,md}_device_operations.
2094 Details : Constify instances of struct {lu,dt,md}_device_operations.
2098 Description: Introduce struct md_site and move meta-data specific parts of
2099 struct lu_site here.
2100 Details : Move md-specific fields out of struct lu_site into special struct
2101 md_site, so that lu_site can be used on a client.
2105 Description: Kill mdd_lov_destroy().
2106 Details : Remove unused mdd code.
2110 Description: Add st_block checking to multistat.c.
2111 Details : Add st_block checking to multistat.c.
2115 Description: Add lu_ref support to struct obd_device.
2116 Details : Add lu_ref tracking to obd_device.
2120 Description: Kill obd_set_fail_loc().
2121 Details : Remove unused code.
2125 Description: Add special type for ptlrpc_request interpret functions.
2126 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2127 there. Actual usage will be in CLIO.
2131 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2132 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2133 doesn't work with lockdep.
2137 Description: Add rwv.c test program.
2138 Details : New testing program exercising readv(2) and writev(2) (Qian).
2142 Description: Add sendfile.c test program.
2143 Details : New testing program exercising sendfile(2) (Jay).
2147 Description: Ratelimit a message that can be very frequent.
2148 Details : Ratelimit a memory allocation failure message that can
2153 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2154 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2158 Description: Liblustre build fixes.
2159 Details : Liblustre build fixes.
2163 Description: libcfs: add cfs_{need,cond}_resched() interface.
2164 Details : libcfs: add cfs_{need,cond}_resched() definition and
2165 implementations for Linux, NT, and liblustre.
2167 Severity : enhancement
2169 Description: More exported tunables for mballoc
2170 Details : Add support for tunable preallocation window and new tunables for
2171 large/small requests
2175 Description: Detect corruption of block bitmap and checking for preallocations
2176 Details : Checks validity of on-disk block bitmap. Also it does better
2177 checking of number of applied preallocations. When corruption is
2178 found, it turns filesystem readonly to prevent further corruptions.
2182 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2183 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2184 window to 0. If it does hit cache miss, and read-pattern is still
2185 stride-io mode, does not reset the stride window, but also does not
2186 increase the stride window length in this case.
2190 Frequency : only for big-endian servers
2191 Description: Check if system is big-endian while mounting fs with extents feature
2192 Details : Mounting a filesystem with extents feature will fail on big-endian
2193 systems since ext3-based ldiskfs is not supported on big-endian
2194 systems. This can be over-riden with "bigendian_extents" mount option.
2196 Severity : enhancement
2198 Description: The root squash functionality
2199 Details : A security feature, which is to prevent users from being able
2200 to mount lustre on their desktop, run as root, and delete
2201 all of the files in the filesystem. The goal is accomplished by
2202 remapping user id (UID) and group id (GID) of the root user to
2203 a UID and GID specified by the system administartor via Lustre
2204 configuration management server (MGS). The functionality also
2205 allows to specify sets of clients for which the remapping does
2210 Description: Excessive recovery window
2211 Details : With AT enabled, the recovery window can be excessively long (6000+
2212 seconds). To address this problem, we no longer use
2213 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2214 timeout no longer depends on the service time, it is set to
2215 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2216 time via pb_service_time.
2220 Description: Watchdog triggered on MDS failover
2221 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2222 the OSTs know that the MDS "UUID" can be reused for the same export
2223 from a different NID, so we do not need to wait for the export to be
2227 Frequency : rare, only if using MMP with Linux RAID
2229 Description: MMP doesn't work with Linux RAID
2230 Details : While using HA for Lustre servers with Linux RAID, it is possible
2231 that MMP will not detect multiple mounts. To make this work we
2232 need to unplug the device queue in RAID when the MMP block is being
2233 written. Also while reading the MMP block, we should read it from
2234 disk and not the cached one.
2236 Severity : enhancement
2238 Description: open file using fid
2239 Details : A file can be opened using just its fid, like
2240 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2243 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2246 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2247 Details : add proper locking for creation of jbd_4k slab cache
2251 Description: MMP check in ext3_remount() fails without displaying any error
2252 Details : When multiple mount protection fails during remount, proper error
2255 Severity : enhancement
2257 Description: Allow stripe size to be up to 4G-64k
2258 Details : Fix math logic to allow large stripe sizes.
2262 Description: add check for >8TB ldiskfs filesystems
2263 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2264 Don't allow >8TB ldiskfs filesystems to be mounted without
2265 force_over_8tb mount option
2270 Descriptoin: OSS complains "too many blocks" and crashes
2271 Details : In filter_commitrw_write():
2272 1. lquota_chkquota();
2273 2. obdo_from_inode();
2274 3. lquota_pending_commit();
2275 "2." has changed uid and gid in this case, then caused
2276 lquota_chkquota() and lquota_pending_commit() to use
2277 different uid and gid. That is the root of the bug.
2281 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2282 Details : Detect and print out the 'stripe count = -1' case properly.
2284 --------------------------------------------------------------------------------
2286 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2288 * Support for kernels:
2289 2.6.5-7.283 (SLES 9),
2290 2.6.9-55.EL (RHEL 4),
2291 2.6.16.46-0.14 (SLES 10),
2292 2.6.18.8 vanilla (kernel.org)
2293 * Client support for unpatched kernels:
2294 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2295 2.6.16 - 2.6.22 vanilla (kernel.org)
2296 * Due to recently discovered recovery problems, we do not recommend
2297 using patchless RHEL 4 clients with this or any earlier release.
2298 * Recommended e2fsprogs version: 1.39.cfs8
2299 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2300 * Starting with this release, the ldiskfs backing filesystem required
2301 by Lustre is now in its own package, lustre-ldiskfs. This package
2302 should be installed. It is versioned separately from Lustre and
2303 may be released separately in future.
2307 Description: block reactivating mgc import until all deactivates complete
2308 Details : Fix race when failing back MDT/MGS to itself (testing)
2311 Frequency : at statup only
2313 Description: mds_lov_synchronize race leads to various problems
2314 Details : simultaneous MDT->OST connections at startup can cause the
2315 sync to abort, leaving the OSC in a bad state.
2317 Severity : enhancement
2319 Description: add optional extra BUILD_VERSION info
2320 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2321 to override the lustre version.
2324 Frequency : 2.6.18 servers only
2326 Description: ll_kern_mount() doesn't release the module reference
2327 Details : The ldiskfs module reference count never drops down to 0
2328 because ll_kern_mount() doesn't release the module reference.
2333 Description: server LBUG when using old ost_num_threads parameter
2334 Details : Accept the old ost_num_threads parameter but warn that it
2335 is deprecated, and fix an off-by-one error that caused an LBUG.
2340 Description: Transient SCSI error results in persistent IO issue
2341 Details : iobuf->dr_error is not reinitialized to 0 between two
2345 Frequency : sometimes when underlying device returns I/O errors
2347 Description: OSTs not going read-only during write failures
2348 Details : OSTs are not remounted read-only when the journal commit threads
2349 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2350 instead of the ext3 wrappers.
2354 Description: poor connect scaling with increasing client count
2355 Details : Don't run filter_grant_sanity_check for more than 100 exports
2356 to improve scaling for large numbers of clients.
2359 Frequency : SLES10 only
2361 Description: sanity-quota.sh quotacheck failed: rc = -22
2362 Details : Quotas cannot be enabled on SLES10.
2365 Frequency : liblustre clients only
2367 Description: getdirentries does not give error when run on compute nodes
2368 Details : getdirentries does not fail when the size specified as an argument
2369 is too small to contain at least one entry
2371 Severity : enhancement
2373 Description: Add LNET router traceability for debug purposes
2374 Details : If a checksum failure occurs with a router as part of the
2375 IO path, the NID of the last router that forwarded the bulk data
2376 is printed so it can be identified.
2381 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2382 Details : Due to a race condition, liblustre clients were occasionally
2383 evicted incorrectly.
2385 Severity : enhancement
2387 Description: lfs setstripe use optional parameters instead of postional
2390 Severity : enhancement
2392 Description: Nanosecond timestamp support for ldiskfs
2393 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2394 resolution timestamps. There is not yet support for this at
2395 the Lustre filesystem level.
2398 Frequency : during server recovery
2400 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2401 Details : request with rq_no_resend flag not awake l_wait_event if they get
2405 Frequency : nfs export on patchless client
2407 Description: connectathon hang when test nfs export over patchless client
2408 Details : Disconnected dentry cannot be found with lookup, so we do not need
2409 to unhash it or make it invalid
2412 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2413 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2414 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2419 Description: open req refcounting wrong on reconnect
2420 Details : If reconnect happened between getting open reply from server and
2421 call to mdc_set_replay_data in ll_file_open, we will schedule
2422 replay for unreferenced request that we are about to free.
2423 Subsequent close will crash in variety of ways.
2424 Check that request is still eligible for replay in
2425 mdc_set_replay_data().
2430 Description: disable writes to filesystem when reading health_check file
2431 Details : the default for reading the health_check proc file has changed
2432 to NOT do a journal transaction and write to disk, because this
2433 can cause reads of the /proc file to hang and block HA state
2434 checking on a healthy but otherwise heavily loaded system. It
2435 is possible to return to the previous behaviour during configure
2436 with --enable-health-write.
2438 Severity : enhancement
2440 Description: 64-bit inode version
2441 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2442 to the inode. This will be required for version-based recovery.
2447 Description: MDS fails to start if a duplicate client export is detected
2448 Details : in some rare cases it was possible for a client to connect to
2449 an MDS multiple times. Upon recovery the MDS would detect this
2450 and fail during startup. Handle this more gracefully.
2452 Severity : enhancement
2454 Description: Add -o localflock option to simulate old noflock
2456 Details : This will achieve local-only flock/fcntl locks
2462 Description: log_commit_thread vs filter_destroy race leads to crash
2463 Details : Take import reference before releasing llog record semaphore
2468 Description: Wrong request locking in request set processing
2469 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2470 imp_delayed_list, in this place should be used imp_lock.
2473 Frequency : when reconnection
2475 Description: Grant Leak when osc reconnect to OST
2476 Details : When osc reconnect ost, OST(filter) should check whether it
2477 should grant more space to client by comparing fed_grant and
2478 cl_avail_grant, and return the granted space to client instead
2479 of "new granted" space, because client will call osc_init_grant
2480 to update the client grant space info.
2483 Frequency : when client reconnect to OST
2485 Description: Grant Leak when osc do resend and replay bulk write
2486 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2487 bulk write request, because the grant info while be sync between
2488 OSC and OST when reconnect, and we should ignore the grant info
2489 these of resend/replay write req.
2494 Description: Grant space more than avaiable left space sometimes.
2495 Details : When then OST is about to be full, if two bulk writing from
2496 different clients came to OST. Accord the avaliable space of the
2497 OST, the first req should be permitted, and the second one
2498 should be denied by ENOSPC. But if the seconde arrived before
2499 the first one is commited. The OST might wrongly permit second
2500 writing, which will cause grant space > avaiable space.
2503 Frequency : when client is evicted
2505 Description: Grant might be wrongly erased when osc is evicted by OST
2506 Details : when the import is evicted by server, it will fork another
2507 thread ptlrpc_invalidate_import_thread to invalidate the
2508 import, where the grant will be set to 0. While the original
2509 thread will update the grant it got when connecting. So if
2510 the former happened latter, the grant will be wrongly errased
2511 because of this race.
2516 Description: Checking Stale with correct fid
2517 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2518 to check whether it is stale, because sometimes, we want the
2519 enqueue happened anyway, and op_data.fid2 will not be initialized.
2521 Severity : enhancement
2523 Description: update patchless client
2524 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2527 Frequency : only with 2.4 kernel
2529 Description: random memory corruption
2530 Details : size of struct ll_inode_info is to big for union inode.u and this
2531 can be cause of random memory corruption.
2536 Description: Memory leak in recovery
2537 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2538 It should also check obd_fail before fsfilt_start, otherwise if
2539 fsfilt_start return -EROFS,(failover mds during mds recovery).
2540 then the req will return with repmsg->transno = 0 and rc = EROFS.
2541 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2542 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2543 be freed no matter whether fsfilt_commit success or not.
2546 Frequency : only with huge count clients
2548 Description: Prevents from taking the superblock lock in llap_from_page for
2550 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2551 for llap_from_page call in ll_removepage prevents from taking the
2552 superblock lock for a soon died page.
2557 Description: Not check open intent error before release open handle
2558 Details : in some rare cases, the open intent error is not checked before
2559 release open handle, which may cause
2560 ASSERTION(open_req->rq_transno != 0), because it tries to release
2561 the failed open handle.
2566 Description: Set cat log bitmap only after create log success.
2567 Details : in some rare cases, the cat log bitmap is set too early. and it
2568 should be set only after create log success.
2572 Description: Accessing a block bevice can re-enable I/O when Lustre is
2573 tearing down a device.
2574 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2579 Description: service threads may hog cpus when there are a lot of requests
2580 Details : Insert cond_resched to give other threads a chance to use some CPU
2585 Description: the cat log was not initialized in recovery
2586 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2587 unlink log on mds will not be initialized until mds post
2588 recovery. And also in mds post recovery, the unlink log will
2589 initialization will be done asynchronausly, so there will be race
2590 between add unlink log and unlink log initialization.
2594 Description: brw_stats were being printed incorrectly
2595 Details : brw_stats were being printed as log2 but all of them were not
2596 recorded as log2. Also remove some code duplication arising from
2597 filter_tally_{read,write}.
2601 Frequency : rare, only in recovery.
2602 Description: ASSERTION(req->rq_type != LI_POISON) failed
2603 Details : imp_lock should be held while iterating over imp_sending_list for
2604 prevent destroy request after get timeout in ptlrpc_queue_wait.
2608 Description: replay-single.sh test 52 fails
2609 Details : A lock's skiplist need to be cleanup when it being unlinked
2610 from its resource list.
2614 Description: Short directio read returns full requested size rather than
2616 Details : Direct I/O operations should return actual amount of bytes
2617 transferred rather than requested size.
2619 Severity : enhancement
2621 Description: metadata RPC reduction (e.g. for rm performance)
2622 Details : decrease the amount of synchronous RPC between clients and servers
2623 by canceling conflicing lock before the operation on the client
2624 and packing thier handles into the main operation RPC to server.
2626 Severity : enhancement
2628 Description: Async OSC create to avoid the blocking unnecessarily.
2629 Details : If a OST has no remain object, system will block on the creating
2630 when need to create a new object on this OST. Now, ways use
2631 pre-created objects when available, instead of blocking on an
2632 empty osc while others are not empty. If we must block, we block
2633 for the shortest possible period of time.
2637 Description: improve handling recoverable errors
2638 Details : if request processig with error which can be recoverable on server
2639 request should be resend, otherwise page released from cache and
2642 Severity : enhancement
2644 Description: refine locking for avoid write wrong info into lov_objid file
2645 Details : fix possible races with add new target and write/update data in
2648 --------------------------------------------------------------------------------
2650 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2655 Frequency : on some architectures
2657 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2658 Details : Uninitialized flags sometimes cause configuration commands to
2662 Frequency : patchless clients only
2664 Description: missing __iget() symbol export
2665 Details : The __iget() symbol export is missing. To avoid the need for
2666 this on patchless clients the deathrow inode reaper is turned
2667 off, and we depend on the VM to clean up old inodes. This
2668 dependency was during via the fix for bug 12181.
2670 --------------------------------------------------------------------------------
2672 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2674 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2675 INTEROPERATE with older versions automatically. In many cases a
2676 special upgrade step is needed. Please read the
2677 user documentation before upgrading any part of a live system.
2678 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2679 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2680 * WARNING: Lustre configuration and startup changes are required with
2681 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2685 Severity : enhancement
2687 Description: MountConf
2688 Details : Lustre configuration is now managed via mkfs and mount
2689 commands instead of lmc and lconf. New obd types (MGS, MGC)
2690 are added for dynamic configuration management. See
2691 https://mail.clusterfs.com/wikis/lustre/MountConf for
2694 Severity : enhancement
2696 Description: dynamic OST addition
2697 Details : OSTs can now be added to a live filesystem
2699 Severity : enhancement
2701 Description: startup order invariance
2702 Details : MDTs and OSTs can be started in any order. Clients only
2703 require the MDT to complete startup.
2705 Severity : enhancement
2707 Description: parallel, asynchronous orphan cleanup
2708 Details : orphan cleanup is now performed in separate threads for each
2709 OST, allowing parallel non-blocking operation.
2711 Severity : enhancement
2713 Description: optimized stripe assignment
2714 Details : stripe assignments are now made based on ost space available,
2715 ost previous usage, and OSS previous usage, in order to try
2716 to optimize storage space and networking resources.
2718 Severity : enhancement
2720 Description: Permanently set tunables
2721 Details : All writable /proc/fs/lustre tunables can now be permanently
2722 set on a per-server basis, at mkfs time or on a live system.
2724 Severity : enhancement
2726 Description: Lustre message v2
2727 Details : Add lustre message format v2.
2729 Severity : enhancement
2731 Description: client OST exclusion list
2732 Details : Clients can be started with a list of OSTs that should be
2733 declared "inactive" for known non-responsive OSTs.
2737 Description: SPEC SFS validation failure on NFS v2 over lustre.
2738 Details : Changes the blocksize for regular files to be 2x RPC size,
2739 and not depend on stripe size.
2741 Severity : enhancement
2743 Description: Multiple MD RPCs in flight.
2744 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2745 To avoid overly-overloading MDS, introduce a limit on number of
2746 MDS RPCs in flight for a single client and add /proc controls
2747 to adjust this limit.
2749 Severity : enhancement
2751 Description: client read/write statistics
2752 Details : Add client read/write call usage stats for performance
2753 analysis of user processes.
2754 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2755 file access. extents_stats shows chunk size distribution.
2756 extents_stats_per_process show chunk size distribution per
2759 Severity : enhancement
2761 Description: mds statistics
2762 Details : Add detailed mds operations statistics in
2763 /proc/fs/lustre/mds/*/stats.
2767 Description: Failure of copying files with lustre special EAs.
2768 Details : Client side always return success for setxattr call for lustre
2769 special xattr (currently only "trusted.lov").
2774 Description: Refcount LNET uuids
2775 Details : The global LNET uuid list grew linearly with every startup;
2776 refcount repeated list entries instead of always adding to
2779 Severity : enhancement
2781 Description: Dynamic service threads
2782 Details : Within a small range, start extra service threads
2783 automatically when the request queue builds up.
2786 Frequency : mixed-endian client/server environments
2788 Description: mixed-endian crashes
2789 Details : The new msg_v2 system had some failures in mixed-endian
2792 Severity : enhancement
2794 Description: Easy OST removal
2795 Details : OSTs can be permanently deactivated with e.g. 'lctl
2796 conf_param lustre-OST0001.osc.active=0'
2798 Severity : enhancement
2800 Description: MGS proc entries
2801 Details : Added basic proc entries for the MGS showing what filesystems
2804 Severity : enhancement
2806 Description: provide MGS failover
2807 Details : Added config lock reacquisition after MGS server failover.
2809 Severity : enhancement
2811 Description: add Linux 2.4 support
2812 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2816 Description: a large application tries to do I/O to the same resource and dies
2817 in the middle of it.
2818 Details : Check the req->rq_arrival time after the call to
2819 ost_brw_lock_get(), but before we do anything about
2820 processing it & sending the BULK transfer request. This
2821 should help move old stale pending locks off the queue as
2822 quickly as obd_timeout.
2825 Frequency : when an incorrect nid is specified during startup
2827 Description: ptlrpc connect to non-existant node causes kernel crash
2828 Details : LNET can't be re-entered from an event callback, which
2829 happened when we expire a message after the export has been
2830 cleaned up. Instead, hand the zombie cleanup off to another
2833 Severity : enhancement
2835 Description: plain/inodebits lock performance improvement
2836 Details : Grouping plain/inodebits in granted list by their request modes
2837 and bits policy, thus improving the performance of search through
2841 Frequency : only if OST filesystem is corrupted
2843 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2844 Details : for a short time RPCs with bulk IO are in the replay list,
2845 but replay of bulk IOs is unimplemented. If the OST filesystem
2846 is corrupted due to disk cache incoherency and then replay is
2847 started it is possible to trip an assertion. Avoid putting
2848 committed RPCs into the replay list at all to avoid this issue.
2851 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2853 Description: System hang on startup
2854 Details : This bug allowed the liblustre (e.g. catamount) client to
2855 return to the app before handling all startup RPCs. This
2856 could leave the node unresponsive to lustre network traffic
2857 and manifested as a server ptllnd timeout.
2859 Severity : enhancement
2861 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2862 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2863 boolean to control whether to print peer debug info when a
2864 client's RPC times out.
2867 Frequency : only for kernels with patches from Lustre below 1.4.3
2869 Description: Remove old rdonly API
2870 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2873 Frequency : only for devices with external journals
2875 Description: Set external device read-only also
2876 Details : During a commanded failover stop, we set the disk device
2877 read-only while the server shuts down. We now also set any
2878 external journal device read-only at the same time.
2881 Frequency : when upgrading from 1.4 while trying to change parameters
2883 Description: The wrong (new) MDC name was used when setting parameters for
2884 upgraded MDT's. Also allows changing of OSC (and MDC)
2885 parameters if --writeconf is specified at tunefs upgrade time.
2888 Frequency : when setting specific ost indicies
2890 Description: QOS code breaks on skipped indicies
2891 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2892 created with --index need not be sequential.
2894 Severity : enhancement
2896 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2897 Details : The uninit_groups feature works in conjunction with the kernel
2898 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2899 up the pass1 processing of e2fsck. This is a read-only feature
2900 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2901 mount filesystems that have had this feature enabled.
2903 Severity : enhancement
2905 Description: Improve multi-block allocation algorithm to avoid fragmentation
2906 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2907 allocation locality and avoid filesystem fragmentation.
2909 ------------------------------------------------------------------------------
2911 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2913 * Support for kernels:
2914 2.4.21-47.0.1.EL (RHEL 3)
2915 2.6.5-7.283 (SLES 9)
2916 2.6.9-42.0.10.EL (RHEL 4)
2917 2.6.12.6 vanilla (kernel.org)
2918 2.6.16.27-0.9 (SLES 10)
2919 * Recommended e2fsprogs version: 1.39.cfs5
2921 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2925 Frequency : occasional, depends on client load and configuration
2926 Bugzilla : 12181, 12203
2927 Description: data loss for recently-modified files
2929 Details : In some cases it is possible that recently written or created
2930 files may not be written to disk in a timely manner (this should
2931 normally be within 30s unless client IO load is very high).
2932 The problem appears as zero-length files or files that are a
2933 multiple of 1MB in size after a client crash or client eviction
2934 that are missing data at the end of the file.
2936 This problem is more likely to be hit on clients where files are
2937 repeatedly created and unlinked in the same directory, clients
2938 have a large amount of RAM, have many CPUs, the filesystem has
2939 many OSTs, the clients are rebooted frequently, and/or the files
2940 are not accessed by other nodes after being written.
2942 The presence of the problem can be detected by looking at
2943 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2944 smaller than the second (nr_unused) then dirty files will not
2945 be flushed automatically to disk. "sync; sleep 10" should be
2946 run several times on the node before unmounting it to update
2947 Lustre (this is also safe to run on nodes without this problem).
2949 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2950 that can cause this same problem, so customers using that kernel
2951 also need to update the kernel in addition to Lustre. In order
2952 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2954 It is normal that files written just before a client crash (less
2955 than 30s) may not yet have been flushed to disk, even for local
2959 Frequency : frequent on thin XT3 nodes
2961 Description: UUID collision on thin XT3 Linux nodes
2962 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2963 randomly, since we relied on an insufficiently-seeded PRNG.
2968 Description: OSS hangs after "All ost request buffers busy"
2969 Details : A deadlock between quota and journal operations caused OSS
2970 hangs after printing "All ost request buffers busy."
2973 Frequency : always on liblustre builds
2975 Description: Cleanup compiler warnings on liblustre
2978 Frequency : always on liblustre builds on XT3
2980 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2984 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2985 > 32000 subdirectories
2986 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2987 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2988 subdirectory count crosses 32000. This will aid e2fsck to
2989 correctly handle more than 32000 subdirectories.
2992 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2994 Description: System hang on startup
2995 Details : This bug allowed the liblustre (e.g. catamount) client to
2996 return to the app before handling all startup RPCs. This
2997 could leave the node unresponsive to lustre network traffic
2998 and manifested as a server ptllnd timeout.
3000 Severity : enhancement
3002 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3003 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3004 boolean to control whether to print peer debug info when a
3005 client's RPC times out.
3010 Description: make O_SYNC working on 2.6 kernels
3011 Details : 2.6 kernels use different method for mark pages for write,
3012 so need add a code to lustre for O_SYNC work.
3017 Description: Failure to close file and release space on NFS
3018 Details : Put inode details into lock acquired in ll_intent_file_open.
3019 Use mdc_intent_lock in ll_intent_open to properly
3020 detect all kind of errors unhandled by mdc_enqueue.
3025 Description: proc file read during shutdown sometimes raced obd removal,
3027 Details : Add lock to prevent obd access after proc file removal.
3030 Frequency : Only for files larger than 4GB on 32-bit clients.
3032 Description: improperly doing page alignment of locks
3033 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3034 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3039 Description: RPCs being resent when they shouldn't be
3040 Details : Some RPCs that should not be resent are being resent. This
3041 can cause inconsistencies in the RPC state machine. Do not
3042 resend such requests.
3045 Frequency : rare, only with NFS export
3047 Description: Crash on NFS re-export node
3048 Details : under very unusual load conditions an assertion is hit in
3049 ll_intent_file_open()
3052 Frequency : only if OST filesystem is corrupted
3054 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3055 Details : for a short time RPCs with bulk IO are in the replay list,
3056 but replay of bulk IOs is unimplemented. If the OST filesystem
3057 is corrupted due to disk cache incoherency and then replay is
3058 started it is possible to trip an assertion. Avoid putting
3059 committed RPCs into the replay list at all to avoid this issue.
3064 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3065 Details : Large single O_DIRECT read and write calls can fail to allocate
3066 a sufficiently large buffer to process the request. In case of
3067 allocation failure the allocation is retried with a smaller
3068 buffer and broken into smaller requests.
3070 Severity : enhancement
3072 Description: Add -o localflock option to simulate old noflock behaviour.
3073 Details : This will achieve local-only flock/fcntl locks coherentness.
3078 Description: versioning check is incomplete
3079 Details : Checking the version difference of client vs. server, report
3080 error if the gap is too big.
3085 Description: add support PG_writeback bit
3086 Details : add support for PG_writeback bit for Lustre, for more carefull
3087 work with page cache in 2.6 kernel. This also fix some deadlocks
3088 and remove hack for work O_SYNC with 2.6 kernel.
3090 Severity : enhancement
3092 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3093 Details : The uninit_groups feature works in conjunction with the kernel
3094 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3095 up the pass1 processing of e2fsck. This is a read-only feature
3096 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3097 mount filesystems that have had this feature enabled.
3099 Severity : enhancement
3101 Description: Improve multi-block allocation algorithm to avoid fragmentation
3102 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3103 allocation locality and avoid filesystem fragmentation.
3105 ------------------------------------------------------------------------------
3107 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3109 * Support for kernels:
3110 2.6.9-42.0.3.EL (RHEL 4)
3111 2.6.5-7.276 (SLES 9)
3112 2.4.21-47.0.1.EL (RHEL 3)
3113 2.6.12.6 vanilla (kernel.org)
3114 2.6.16.21-0.8 (SLES10)
3115 * Recommended e2fsprogs version: 1.39.cfs2-0
3117 * The backwards-compatible /proc/sys/portals symlink has been removed
3118 in this release. Before upgrading, please ensure that you change
3119 any configuration scripts or /etc/sysctl.conf files that access
3120 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3121 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3122 in advance of the upgrade on any system running Lustre 1.4.6 or
3123 newer, since /proc/sys/lnet was added in that version.
3124 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3128 Frequency : only when quota is used
3130 Description: avoid scanning export list for quota master
3131 Details : Change the algorithms to avoid scanning export list in order
3132 to improve the efficiency.
3135 Frequency : MDS failover only, very rarely
3137 Description: "went back in time" messages on mds failover
3138 Details : The greatest transno may be lost when the current operation
3139 finishes with an error (transno==0) and the client's last_rcvd
3140 record is over-written. Save the greatest transno in the
3141 mds_last_transno for this case.
3144 Frequency : always for specific kernels and striping counts
3146 Description: client may get "Matching packet too big" without ACL support
3147 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3148 when trying to access files in certain configurations. The
3149 clients should in fact be denied when mounting because they do
3150 not understand ACLs.
3153 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3155 Description: many clients connecting with IO in progress causes connect timeouts
3156 Details : Avoid synchronous journal commits to avoid delays caused by many
3157 clients connecting/disconnecting when bulk IO is in progress.
3158 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3159 OST_IO_PORTAL to avoid delays behind potentially many pending
3163 Frequency : occasionally with multiple writers to a single file
3165 Description: shared writes to file may result in wrong size reported by stat()
3166 Details : Allow growing of kms when extent lock is cancelled
3169 Frequency : always with random mmap IO to multi-striped file
3171 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3172 Details : Only if the hole is at the end of OST object so that kms is too
3173 small. Fix is to increase kms accordingly in ll_nopage.
3176 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3178 Description: writes to a missing object would leak memory on the OST
3179 Details : If there is an inconsistency between the MDS and OST filesystems,
3180 such that the MDS references an object that doesn't exist, writes
3181 to that object will leak memory due to incorrect cleanup in the
3182 error handling path, eventually running out of memory on the OST.
3187 Description: Creating too long symlink causes lustre errors
3188 Details : Check symlink and name lengths before sending requests to MDS.
3191 Frequency : only if flock is enabled (not on by default)
3193 Description: posix locks not released on fd closure on 2.6.9+
3194 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3195 kernels, this caused such locks not to be released on fd close and
3196 then assertions on fs unmount about still used locks.
3199 Frequency : MDS failover only, very rarely
3201 Description: clients may get ASSERTION(granted_lock != NULL)
3202 Details : When request was taking a long time, and a client was resending
3203 a getattr by name lock request. The were multiple lock requests
3204 with the same client lock handle and
3205 mds_getattr_name->fixup_handle_for_resent_request found one of the
3206 lock handles but later failed with ASSERTION(granted_lock != NULL).
3211 Description: handle->h_buffer_credits > 0, assertion failure
3212 Details : h_buffer_credits is zero after truncate, causing assertion
3213 failure. This patch extends the transaction or creates a new
3217 Frequency : NFS re-export or patchless client
3218 Bugzilla : 11179, 10796
3219 Description: Crash on NFS re-export node (__d_move)
3220 Details : We do not want to hash the dentry if we don't have a lock.
3221 But if this dentry is later used in d_move, we'd hit uninitialised
3222 list head d_hash, so we just do this to init d_hash field but
3223 leave dentry unhashed.
3226 Frequency : NFS re-export or patchless client
3228 Description: NFS exports has problem with symbolic link
3229 Details : lustre client didn't properly install dentry when re-exported
3230 to NFS or running patchless client.
3233 Frequency : NFS re-export or patchless client
3235 Description: Various nfs/patchless fixes.
3236 Details : fixes reuse disconected alias for lookup process - this fixes
3237 warning "find_exported_dentry: npd != pd",
3238 fix permission error with open files at nfs.
3239 fix apply umask when do revalidate.
3242 Frequency : occasional
3244 Description: Crash on NFS re-export node
3245 Details : calling clear_page() on the wrong pointer triggered oops in
3246 generic_mapping_read().
3249 Frequency : rarely, using O_DIRECT IO
3251 Description: unaligned directio crashes client with LASSERT
3252 Details : check for unaligned buffers before trying any requests.
3255 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3257 Description: stale data returned from RAID cache
3258 Details : If only a small amount of IO is done to the RAID device before
3259 reading it again it is possible to get stale data from the RAID
3260 cache instead of reading it from disk.
3263 Frequency : always for sles10 kernel
3265 Description: sles10 support
3266 Details : ll_follow_link: compile fixes and using of nd_set_link
3267 under newer kernels.
3270 Frequency : depends on arch, kernel and compiler version, always on sles10
3273 Description: recursive or deep enough symlinks cause stack overflow
3274 Details : getting rid of large stack-allocated variable in
3278 Frequency : depends on hardware
3280 Description: lustre write performance loss in the SLES10 kernel
3281 Details : the performance loss is caused by using of write barriers in the
3282 ext3 code. The SLES10 kernel turns barrier support on by
3283 default. The fix is to undo that change for ldiskfs.
3285 ------------------------------------------------------------------------------
3287 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3289 * Support for kernels:
3290 2.6.9-42.0.3EL (RHEL 4)
3291 2.6.5-7.276 (SLES 9)
3292 2.4.21-47.0.1.EL (RHEL 3)
3293 2.6.12.6 vanilla (kernel.org)
3297 Frequency : quota enabled and large files being deleted
3299 Description: releasing more than 4GB of quota at once hangs OST
3300 Details : If a user deletes more than 4GB of files on a single OST it
3301 will cause the OST to spin in an infinite loop. Release
3302 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3307 Description: statfs data retrieved from /proc may be stale or zero
3308 Details : When reading per-device statfs data from /proc, in the
3309 {kbytes,files}_{total,free,avail} files, it may appear
3310 as zero or be out of date.
3313 Frequency : systems with MD RAID1 external journal devices
3315 Description: lconf's call to blkid is confused by RAID1 journal devices
3316 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3317 of returning all block devices that match the journal UUID.
3320 Frequency : always, for aggregate stripe size over 4GB
3322 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3323 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3324 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3327 Frequency : always if "lfs find" used on a local file/directory
3329 Description: "lfs find" segfaults if used on a local file/directory
3330 Details : The case where a directory component was not specified wasn't
3331 handled correctly. Handle this properly.
3334 Frequency : always on ppc64
3336 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3337 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3338 perform endianness conversion.
3341 Frequency : rarely (truncate to non-zero file size after write under load)
3342 Bugzilla : 10730, 10687
3343 Description: Files padded with zeros to next 4K multiple
3344 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3345 it is possible that files that are truncated to a non-zero size
3346 immediately after being written are filled with zero bytes beyond
3347 the truncated size. No file data is lost.
3349 Severity : enhancement
3351 Description: Allow recovery/failover for liblustre clients.
3352 Details : liblustre clients were unaware of failover configurations until
3355 Severity : enhancement
3357 Description: user file locks should fail when not mounting with flock option
3358 Details : Set up an error-returning stub in ll_file_operations.lock field
3359 to prevent incorrect behaviour when client is mounted without
3360 flock option. Also, set up properly f_op->flock field for
3364 Frequency : always on ia64
3366 Description: "lfs df" loops on printing out MDS statfs information
3367 Details : The obd_ioctl_data was not initialized and in some systems
3368 this caused a failure during the ioctl that did not return
3369 an error. Initialize the struct and return an error on failure.
3372 Frequency : SLES 9 only
3374 Description: Error of copying files with lustre special EAs as root
3375 Details : Client side always return success for setxattr call for lustre
3376 special xattr (currently only "trusted.lov").
3379 Frequency : rarely on clusters with both ia64+i386 clients
3381 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3382 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3383 clients are doing concurrent writes to the same file it is
3384 possible that the ia64 clients may LASSERT because the OST
3385 extent locks are not PAGE_SIZE aligned. Ensure that grown
3386 locks are always aligned on the request boundary.
3389 Frequency : specific use, occasional
3391 Description: Overwriting in use executable truncates on-disk binary image
3392 Details : If one node attempts to overwrite an executable in use by
3393 another node, we now correctly return ETXTBSY instead of
3394 truncating the file.
3396 Severity : enhancement
3398 Description: Async OSC create to avoid the blocking unnecessarily.
3399 Details : If a OST has no remain object, system will block on the creating
3400 when need to create a new object on this OST. Now, ways use
3401 pre-created objects when available, instead of blocking on an
3402 empty osc while others are not empty. If we must block, we block
3403 for the shortest possible period of time.
3408 Description: chmod on Lustre root is propagated to other clients
3409 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3410 invalid by the follow_mount time.
3413 Frequency : liblustre clients only
3415 Description: Race in 'instant cancel' lock handling could lead to such locks
3416 never to be granted in case of SMP MDS
3417 Details : Do not destroy not yet granted but cbpending locks in
3421 Frequency : replay/resend of open
3423 Description: non null lock assetion failure in mds_intent_policy
3424 Details : Trying to replay/resend lockless open requests resulted in
3425 mds_open() returning 0 with no lock. Now it sets a flag if
3426 a lock is going to be returned.
3428 Severity : enhancement
3430 Description: Checksum enhancements
3431 Details : New checksum enhancements allow for resending RPCs that failed
3434 Severity : enhancement
3436 Description: Tunables on number of dirty pages in cacche
3437 Details : Allow to set limit on number of dirty pages cached.
3442 Description: client crash on unmount - lock still has references
3443 Details : In some error handling cases it was possible to leak a lock
3444 reference on a client while accessing a file. This was not
3445 harmful to the client during operation, but would cause the
3446 client to crash when the filesystem is unmounted.
3449 Frequency : specific case, rare
3451 Description: ETXTBSY on mds though file not in use by client
3452 Details : ETXTBSY is no longer incorrectly returned when attempting to
3453 chmod or chown a directory that the user previously tried to
3454 execute or a currently-executing binary.
3457 Frequency : extremely rare except on liblustre-based clients
3459 Description: Lustre space not freed when files are deleted
3460 Details : Clean up open-unlinked files after client eviction. Previously
3461 the unlink was skipped and the files remained as orphans.
3466 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3467 Details : In some cases it was possible to send a blocking callback to
3468 a client doing a glimpse, even though that client didn't get
3469 a lock granted. When the glimpse lock is cancelled on the OST
3470 the freed lock is left on the waiting list and corrupted the list.
3473 Frequency : all core dumps
3475 Description: Broke core dumps to lustre
3476 Details : Negative dentry may be unhashed if parent does not have UPDATE
3477 lock, but some callers, e.g. do_coredump, expect dentry to be
3478 hashed after successful create, hash it in ll_create_it.
3480 ------------------------------------------------------------------------------
3482 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3484 * Support for kernels:
3485 2.6.9-42.0.2.EL (RHEL 4)
3486 2.6.5-7.276 (SLES 9)
3487 2.4.21-40.EL (RHEL 3)
3488 2.6.12.6 vanilla (kernel.org)
3492 Frequency : always on RHEL 3
3494 Description: Number of open files grows over time
3495 Details : The number of open files grows over time, whether or not
3496 Lustre is started. This was due to a filp leak introduced
3497 by one of our kernel patches.
3499 ------------------------------------------------------------------------------
3501 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3503 * Support for kernels:
3504 2.6.9-42.EL (RHEL 4)
3505 2.6.5-7.276 (SLES 9)
3506 2.4.21-40.EL (RHEL 3)
3507 2.6.12.6 vanilla (kernel.org)
3512 Bugzilla : 5719, 9635, 9792, 9684
3513 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3514 Details : If a server is under heavy load and cannot reply to new
3515 connection requests before the client resends the (re)connect,
3516 the connection handling code can behave badly if two service
3517 threads are concurrently handing separate (re)connections from
3518 the same client. Add better locking to the connection handling
3519 code, and ensure that only a single connection will be processed
3520 for a given client UUID, even if the lock is dropped.
3522 Severity : enhancement
3524 Description: add TCP zero-copy support to kernel
3525 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3526 sends if the hardware supports scatter-gather and checksumming.
3527 This allows socklnd to do client-write and server-read more
3528 efficiently and reduce CPU utilization from skbuf copying.
3531 Frequency : only if NFS exporting from client
3533 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3534 Details : If setting attributes on a file created under NFS that had
3535 never been opened it would be possible to oops the client
3536 if the file had no objects.
3539 Frequency : always for liblustre
3541 Description: liblustre client does MDS+OSTs setattr RPC for each write
3542 Details : When doing a write from a liblustre client, the client
3543 incorrectly issued an RPC to the MDS and each OST the file was
3544 striped over in order to update the timestamps. When writing
3545 with small chunks and many clients this could overwhelm the MDS
3546 with RPCs. In all cases it would slow down the write because
3547 these RPCs are unnecessary.
3549 Severity : enhancement
3551 Description: allow number of MDS service threads to be changed at module load
3552 Details : It is now possible to change the number of MDS service threads
3553 running. Adding "options mds mds_num_threads={N}" to the MDS's
3554 /etc/modprobe.conf will set the number of threads for the next
3555 time Lustre is restarted (assuming the "mds" module is also
3556 reloaded at that time). The default number of threads will
3557 stay the same, 32 for most systems.
3562 Description: OST crash if filesystem is unformatted or corrupt
3563 Details : If an OST is started on a device that has never been formatted
3564 or if the filesystem is corrupt and cannot even mount then the
3565 error handling cleanup routines would dereference a NULL pointer.
3570 Description: NULL pointer deref in llap_from_page.
3571 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3572 incorrect) mapping if the page was truncated/reclaimed while it was
3573 searched for. Check for this condition and skip such pages when
3574 doing readahead. Introduce extra check to llap_from_page() to
3575 verify page->mapping->host is non-NULL (so page is not anonymous).
3578 Frequency : Sometimes when using sys_sendfile
3580 Description: "page not covered by a lock" warnings from ll_readpage
3581 Details : sendfile called ll_readpage without right page locks present.
3582 Now we introduced ll_file_sendfile that does necessary locking
3583 around call to generic_file_sendfile() much like we do in
3587 Frequency : with certain MDS communication failures at client mount time
3589 Description: NULL pointer deref after failed client mount
3590 Details : a client connection request may delayed by the network layer
3591 and not be sent until after the PTLRPC layer has timed out the
3592 request. If the client fails the mount immediately it will try
3593 to clean up before the network times out the request. Add a
3594 reference from the request import to the obd device and delay
3595 the cleanup until the network drops the request.
3598 Frequency : occasionally during client (re)connect
3600 Description: assertion failure during client (re)connect
3601 Details : processing a client connection request may be delayed by the
3602 client or server longer than the client connect timeout. This
3603 causes the client to resend the connection request. If the
3604 original connection request is replied in this interval, the
3605 client may trip an assertion failure in ptlrpc_connect_interpret()
3606 which thought it would be the only running connect process.
3609 Frequency : only with obd_echo servers and clients that are rebooted
3611 Description: kernel BUG accessing uninitialized data structure
3612 Details : When running an obd_echo server it did not start the ping_evictor
3613 thread, and when a client was evicted an uninitialized data
3614 structure was accessed. Start the ping_evictor in the RPC
3615 service startup instead of the OBD startup.
3617 Severity : enhancement
3618 Bugzilla : 10193 (patchless)
3619 Description: Remove dependency on various unexported kernel interfaces.
3620 Details : No longer need reparent_to_init, exit_mm, exit_files,
3621 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3624 Frequency : rare (only users of deprecated and unsupported LDAP config)
3626 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3627 Details : LDAP apparently contains 'lustreName' attributes instead of
3628 'name'. A simple remapping of the name is sufficient.
3631 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3633 Description: waiting_locks_callback trips kernel BUG if client is evicted
3634 Details : Running with the dump_on_timeout debug flag turned on makes
3635 it possible that the waiting_locks_callback() can try to dump
3636 the Lustre kernel debug logs from an interrupt handler. Defer
3637 this log dumping to the expired_lock_main() thread.
3639 Severity : enhancement
3641 Description: Support NFS exporting on 2.6 kernels.
3642 Details : Implement non-rawops metadata methods for NFS server to use without
3643 changing NFS server code.
3646 Frequency : very rare (synthetic metadata workload only)
3648 Description: two racing renames might cause an MDS thread to deadlock
3649 Details : Running the "racer" program may cause one MDS thread to rename
3650 a file from being the source of a rename to being the target of
3651 a rename at exactly the same time that another thread is doing
3652 so, and the second thread has already enqueued these locks after
3653 doing a lookup of the target and is trying to relock them in
3654 order. Ensure that we don't try to re-lock the same resource.
3657 Frequency : only very large systems with liblustre clients
3659 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3660 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3661 all OSTs in parallel. This allows the eviction of stale liblustre
3662 clients to proceed much faster than if they were done in series,
3663 and also offers similar improvements for other set_info RPCs.
3668 Description: excessive CPU usage during initial read phase on client
3669 Details : During the initial read phase on a client, it would agressively
3670 retry readahead on the file, consuming too much CPU and impacting
3671 performance (since 1.4.5.8). Improve the readahead algorithm
3672 to avoid this, and also improve some other common cases (read
3673 of small files in particular, where "small" is files smaller than
3674 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3679 Description: MDS crash when receiving packet with unknown intent.
3680 Details : Do not LBUG in unknown intent case, just return -EFAULT
3682 Severity : enhancement
3683 Bugzilla : 9293, 9385
3684 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3685 Details : Do not serialize getattr (non-intent version) and statfs.
3688 Frequency : occasional, when OST network is overloaded/intermittent
3690 Description: client evicted by OST after bulk IO timeout
3691 Details : If a client sends a bulk IO request (read or write) the OST
3692 may evict the client if it is unresposive to its data GET/PUT
3693 request. This is incorrect if the network is overloaded (takes
3694 too long to transfer the RPC data) or dropped the OST GET/PUT
3695 request. There is no need to evict the client at all, since
3696 the pinger and/or lock callbacks will handle this, and the
3697 client can restart the bulk request.
3700 Frequency : Always when mmapping file with no objects
3702 Description: client crashes when mmapping file with no objects
3703 Details : Check that we actually have objects in a file before doing any
3704 operations on objects in ll_vm_open, ll_vm_close and
3710 Description: Request leak when working with deleted CWD
3711 Details : Introduce advanced request refcount tracking for requests
3712 referenced from lustre intent.
3714 Severity : Enhancement
3716 Description: Cache open file handles on client.
3717 Details : MDS now will return special lock along with openhandle, if
3718 requested and client is allowed to hold openhandle, even if unused,
3719 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3720 closing files for every read/write openration.
3722 Severity : Enhancement
3724 Description: Cache open negative dentries on client when possible.
3725 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3726 negative dentries on lock revocation.
3731 Description: Remounting a client read-only wasn't possible with a zconf mount
3732 Details : It wasn't possible to remount a client read-only with llmount.
3734 Severity : enhancement
3735 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3736 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3737 MPICH is included to add Lustre-specific ADIO interfaces.
3738 This is based closely on the UFS ADIO layer and only differs
3739 in file creation, in order to allow the OST striping to be set.
3740 This is user-contributed code and not supported by CFS.
3745 Description: extended inode attributes (immutable, append-only) work improperly
3746 when 2.4 and 2.6 kernels are used on client/server or vice versa
3747 Details : Introduce kernel-independent values for these flags.
3749 Severity : enhancement
3752 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3753 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3754 granularity. Now, Fractional MB granularity is be supported,
3755 this is very useful for low memory system.
3757 Severity : enhancement
3759 Description: Getattr by fid
3760 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3761 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3762 have locks and avoids getattr rpc requests for it on every stat.
3765 Frequency : Always, for filesystems larger than 2TB
3767 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3768 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3769 filesystems larger than 2TB to be mounted with mballoc enabled.
3772 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3774 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3775 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3776 filesystems larger than 2TB is mounted then the kernel will
3777 silently corrupt the start of the filesystem. CONFIG_LBD is
3778 enabled for all CFS-supported kernels, but the possibility of
3779 this happening with a modified kernel config exists.
3781 Severity : enhancement
3783 Description: add client O_DIRECT support for 2.6 kernels
3784 Details : It is now possible to do O_DIRECT reads and writes to files
3785 in the Lustre client mountpoint on 2.6 kernel clients.
3787 Severity : enhancement
3789 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3790 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3791 parallel, not waiting for response from every OST before sending
3792 a rpc to the next OST.
3797 Description: setattr vs write race when updating file timestamps
3798 Details : Client processes that update a file timestamp into the past
3799 right after writing to the file (e.g. tar) it is possible that
3800 the updated file modification time can be reset to the current
3801 time due to a race between processing the setattr and write RPC.
3803 Severity : enhancement
3805 Description: Bring 'lfs find' closer in line with regular Linux find.
3806 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3807 -print0 options and obtains all the needed info through the lustre
3810 Severity : enhancement
3812 Description: support up to 1024 configured devices on one node
3813 Details : change obd_dev array from statically allocated to dynamically
3814 allocated structs as they are first used to reduce memory usage
3819 Description: Flush dirty partially truncated pages during truncate
3820 Details : Immediatelly flush partially truncated pages in filter_setattr,
3821 this way we completely avoid having any pages in page cache on OST
3822 and can retire ugly workarounds during writes to flush such pages.
3827 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3828 Details : protect inode from truncation within vfs_unlink() context
3829 just take a reference before calling vfs_unlink() and release it
3830 when parent's i_sem is free.
3835 Description: last_id value checked outside lock on OST caused LASSERT failure
3836 Details : If there were multiple MDS->OST object precreate requests in
3837 flight, it was possible that the OST's last object id was checked
3838 outside a lock and incorrectly tripped an assertion. Move checks
3839 inside locks, and discard old precreate requests.
3842 Frequency : always, if extents are used on OSTs
3844 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3845 Details : OSTs using the extents format would not zero the high 16 bits of
3846 the index physical block number. This is not a problem for any
3847 OST filesystems smaller than 16TB, and no kernels support ext3
3848 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3849 new/modified files) and can be fixed for existing filesystems
3850 with e2fsprogs-1.39-cfs1.
3855 Description: import connection selection may be incorrect if timer wraps
3856 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3857 import connections to be ignored if the 32-bit jiffies counter
3858 wraps. Use a 64-bit jiffies counter.
3861 Frequency : very large clusters immediately after boot
3863 Description: LNET request buffers exhausted under heavy short-term load
3864 Details : If a large number of client requests are generated on a service
3865 that has previously never seen so many requests it is possible
3866 that the request buffer growth cannot keep up with the spike in
3867 demand. Instead of dropping incoming requests, they are held in
3868 the LND until the RPC service can accept more requests.
3871 Frequency : Sometimes during replay
3873 Description: Assertion failure in ll_local_open after replay.
3874 Details : If replay happened on an open request reply before we were able
3875 to set replay handler, reply will become not swabbed tripping the
3876 assertion in ll_local_open. Now we set the handler right after
3877 recognising of open request
3880 Frequency : very rare
3882 Description: kernel reports "badness in vsnprintf"
3883 Details : Reading from the "recovery_status" /proc file in small chunks
3884 may cause a negative length in lprocfs_obd_rd_recovery_status()
3885 call to vsnprintf() (which is otherwise harmless). Exit early
3886 if there is no more space in the output buffer.
3888 Severity : enhancement
3890 Description: clear OBD RPC statistics by writing to them
3891 Details : It is now possible to clear the OBD RPC statistics by writing
3892 to the "stats" file.
3897 Description: Client mtime is not the same on different clients after utimes
3898 Details : In some cases, the client was using the utimes() syscall on
3899 a file cached on another node. The clients now validate the
3900 ctime from the MDS + OSTs to determine which one is right.
3905 Description: Inability to activate failout mode
3906 Details : lconf script incorrectly assumed that in pythong string's numeric
3907 value is used in comparisons.
3910 Frequency : always with multiple stripes per file
3912 Description: Inefficient object allocation for mutli-stripe files
3913 Details : When selecting which OSTs to stripe files over, for files with
3914 a stripe count that divides evenly into the number of OSTs,
3915 the MDS is always picking the same starting OST for each file.
3916 Return the OST selection heuristic to the original design.
3921 Description: mount failures may take full timeout to return an error
3922 Details : Under some heavy load conditions it is possible that a
3923 failed mount can wait for the full obd_timeout interval,
3924 possibly several minutes, before reporting an error.
3925 Instead return an error as soon as the status is known.
3927 Frequency : quota enabled and large files being deleted
3929 Description: releasing more than 4GB of quota at once hangs OST
3930 Details : If a user deletes more than 4GB of files on a single OST it
3931 will cause the OST to spin in an infinite loop. Release
3932 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3937 Description: statfs data retrieved from /proc may be stale or zero
3938 Details : When reading per-device statfs data from /proc, in the
3939 {kbytes,files}_{total,free,avail} files, it may appear
3940 as zero or be out of date.
3943 Frequency : systems with MD RAID1 external journal devices
3945 Description: lconf's call to blkid is confused by RAID1 journal devices
3946 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3947 of returning all block devices that match the journal UUID.
3950 Frequency : always, for aggregate stripe size over 4GB
3952 Description: assertion fails when trying to use 4GB stripe size
3953 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3954 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3957 Frequency : always on ppc64
3959 Description: the first write on an ext3 filesystem with mballoc got stuck
3960 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3961 perform endianness conversion.
3963 ------------------------------------------------------------------------------
3965 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3967 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3968 INTEROPERATE with older versions automatically. Please read the
3969 user documentation before upgrading any part of a live system.
3970 * WARNING: Lustre networking configuration changes are required with
3971 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3974 * Support for kernels:
3975 2.6.9-22.0.2.EL (RHEL 4)
3976 2.6.5-7.244 (SLES 9)
3977 2.6.12.6 vanilla (kernel.org)
3980 Severity : enhancement
3981 Bugzilla : 7981/8208
3982 Description: Introduced Lustre Networking (LNET)
3983 Details : LNET is new networking infrastructure for Lustre, it includes
3984 a reorganized network configuration mode (see the user
3985 documentation for full details) as well as support for routing
3986 between different network fabrics. Lustre Networking Devices
3987 (LNDs) for the supported network fabrics have also been
3988 created for this new infrastructure.
3990 Severity : enhancement
3991 Description: Introduced Access control lists
3992 Details : clients can set ACLs on files and directories in order to have
3993 more fine-grained permissions than the standard Unix UGO+RWX.
3994 The MDS must be started with the "-o acl" mount option.
3996 Severity : enhancement
3997 Description: Introduced filesystem quotas
3998 Details : Administrators may now establish per-user quotas on the
4001 Severity : enhancement
4003 Description: Configuration change for the XT3
4004 The PTLLND is now used to run Lustre over Portals on the XT3
4005 The configure option(s) --with-cray-portals are no longer used.
4006 Rather --with-portals=<path-to-portals-includes> is used to
4007 enable building on the XT3. In addition to enable XT3 specific
4008 features the option --enable-cray-xt3 must be used.
4013 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4014 Details : A race between allocating a new llcd and re-getting the llcd_lock
4015 allowed another thread to grab newly-allocated llcd.
4017 Severity : enhancement
4019 Description: 2.6 OST async journal commit and locking fix to improve performance
4020 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4021 2.6 kernels are now async as they already were in 2.4 kernels so
4022 that they can commit concurrently with the network bulk transfer.
4023 For block-allocated files the filter allocation semaphore is held
4024 to avoid filesystem fragmentation during allocation. BKL lock
4025 removed for 2.6 xattr operations where it is no longer needed.
4030 Description: lconf incorrectly determined whether two IP networks could talk
4031 Details : In some more complicated routing and multiple-network
4032 configurations, lconf will avoid trying to make a network
4033 connection to a disjoint part of the IP space. It was doing the
4034 math incorrectly for one set of cases.
4039 Description: Fix for potential infinite loop processing records in an llog.
4040 Details : If an llog record is corrupted/zeroed, it is possible to loop
4041 forever in llog_process(). Validate the llog record length
4042 and skip the remainder of the block on error.
4045 Frequency : occasional (liblustre only)
4047 Description: liblustre could not open files whose last component is a symlink
4048 Details : sysio_path_walk() would incorrectly pass the open intent to
4049 intermediate path components.
4052 Frequency : rare (liblustre only with non-standard tuning)
4053 Bugzilla : 7201 (7350)
4054 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4055 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4056 after it has been released, only do this for referenced locks
4058 Severity : enhancement
4060 Description: specifying an (invalid) directory default stripe_size of -1
4061 would reset the directory default striping
4062 Details : stripe_size -1 was used internally to signal directory stripe
4063 removal, now use "all default" to signal dir stripe removal
4064 as a directory striping of "all default" is not useful
4067 Frequency : common for large clusters running liblustre clients
4069 Description: doing an ls when liblustre clients are running is slow
4070 Details : sending a glimpse AST to a liblustre client waits for every AST
4071 to time out, as liblustre clients will not respond. Since they
4072 cannot cache data we refresh the OST lock LVB from disk instead.
4074 Severity : enhancement
4076 Description: doing an ls at the same time as file IO can be slow
4077 Details : enqueue and other "small" requests can be blocked behind many
4078 large IO requests. Create a new OST IO portal for non-IO
4079 requests so they can be processed faster.
4082 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4084 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4085 created file with mode 000
4086 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4087 allow an ftruncate() as a non-root user to a file with mode 000.
4088 Lustre now respects this flag to disable mode checks when
4089 truncating a file owned by the user
4092 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4095 Description: Revoking locks from clients that went dead or catatonic might take
4097 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4098 cancellation of such locks instant on servers without waiting for
4099 any reply from clients. Clients drops these locks when cancel
4100 notification from server is received without replying.
4103 Frequency : liblustre-only, when liblustre client dies or becomes busy
4105 Description: Doing ls on Linux clients can take a long time with active
4107 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4108 granting such locks to it in the first place if possible. Locks
4109 are taken by proxy on the OST during the read or write and
4110 dropped immediately afterward. Add connect flags handling, do
4111 not grant locks to liblustre clients for glimpse ASTs.
4113 Severity : enhancement
4115 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4116 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4117 sequential reads and behaves badly with applications doing random
4118 reads. Improve it by reading ahead at least read region, and
4119 avoiding excessive large RPC for small reads.
4121 Severity : enhancement
4123 Description: Creating more than 1000 files for a single job may cause a load
4124 imbalance on the OSTs if there are also a large number of OSTs.
4125 Details : qos_prep_create() uses an OST index reseed value that is an
4126 even multiple of the number of available OSTs so that if the
4127 reseed happens in the middle of the object allocation it will
4128 still utilize the OSTs as uniformly as possible.
4133 Description: OST or MDS may oops in ping_evictor_main()
4134 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4135 but doesn't restart at beginning of obd_exports_timed list
4138 Severity : enhancement
4140 Description: improve by-nid export eviction on the MDS and OST
4141 Details : allow multiple exports with the same NID to be evicted at one
4142 time without re-searching the exports list.
4145 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4147 Description: MDS may oops in groups_free()
4148 Details : in rare race conditions a newly allocated group_info struct is
4149 freed again, and this can be NULL. The 2.4 compatibility code
4150 for groups_free() checked for a NULL pointer, but 2.6 did not.
4153 Frequency : common for liblustre clients doing little filesystem IO
4154 Bugzilla : 9352, 7313
4155 Description: server may evict liblustre clients accessing contended locks
4156 Details : if a client is granted a lock or receives a completion AST
4157 with a blocking AST already set it would not reply to the AST
4158 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4161 Frequency : lfs setstripe, only systems with more than 160 OSTs
4163 Description: unable to set striping with a starting offset beyond OST 160
4164 Details : llapi_create_file() incorrectly limited the starting stripe
4165 index to the maximum single-file stripe count.
4168 Frequency : LDAP users only
4170 Description: lconf did not handle in-kernel recovery with LDAP properly
4171 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4173 Severity : enhancement
4175 Description: bind OST threads to NUMA nodes to improve performance
4176 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4177 node and do their allocations there to localize memory access
4179 Severity : enhancement
4181 Description: llmount can determine client NID directly from Myrinet (GM)
4182 Details : the client NID code from gmnalnid was moved directly into
4183 llmount, removing the need to use this or specifying the
4184 client NID explicitly when mounting GM clients with zeroconf
4187 Frequency : if client is started with down MDS
4189 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4190 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4191 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4192 ptlrpc_import_delayed_req() because we didn't check if the
4193 request was interrupted, and we also didn't break out of the
4194 event loop if there was no timeout
4199 Description: data loss during non-page-aligned writes to a single file from
4200 both multiple nodes and multiple threads on one node at same time
4201 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4202 inconsistency led to false short-reads, that were cached and later
4203 used by ->prepare_write() to fill in partially written page,
4204 leading to data loss.
4207 Frequency : always, if lconf --abort_recovery used
4209 Description: lconf --abort_recovery fails with 'Operation not supported'
4210 Details : lconf was attempting to abort recovery on the MDT device and not
4213 Severity : enhancement
4215 Description: remove cleanup logs
4216 Details : replace lconf-generated cleanup logs with lustre internal
4217 cleanup routines. Eliminates the need for client-cleanup and
4220 Severity : enhancement
4222 Description: add support for EAs (user and system) on lustre filesystems
4223 Details : it is now possible to store extended attributes in the Lustre
4224 client filesystem, and with the user_xattr mount option it
4225 is possible to allow users to store EAs on their files also
4227 Severity : enhancement
4229 Description: Add possibility (config option) to show minimal available OST free
4231 Details : When compiled with --enable-mindf configure option, statfs(2)
4232 (and so, df) will return least minimal free space available from
4233 all OSTs as amount of free space on FS, instead of summary of
4234 free spaces of all OSTs.
4236 Severity : enhancement
4238 Description: do not expand extent locks acquired on OST-side
4239 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4240 by server: they are not cached anyway.
4243 Frequency : when mmap is used/binaries executed from Lustre
4245 Description: Unmmap pages before throwing them away from read cache.
4246 Details : llap_shrink cache now attempts to unmap pages before discarding
4247 them (if unmapping failed - do not discard). SLES9 kernel has
4248 extra checks that trigger if this unmapping is not done first.
4253 Description: lconf didn't resolve symlinks before checking to see whether a
4254 given mountpoint was already in use
4257 Frequency : when migrating failover services
4258 Bugzilla : 6395, 9514
4259 Description: When migrating a subset of services from a node (e.g. failback
4260 from a failover service node) the remaining services would
4261 time out and evict clients.
4262 Details : lconf --force (implied by --failover) sets the global obd_timeout
4263 to 5 seconds in order to quickly disconnect, but this caused
4264 other RPCs to time out too quickly. Do not change the global
4265 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4267 Severity : enhancement
4268 Frequency : if MDS is started with down OST
4269 Bugzilla : 9439,5706
4270 Description: Allow startup/shutdown of an MDS without depending on the
4271 availability of the OSTs.
4272 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4273 Add appropriate locking and lov-osc refcounts for safe
4274 cleaning. Add osc abort_inflight calls in case the
4275 synchronize never started.
4278 Frequency : occasional (Cray XT3 only)
4280 Description: root not authorized to access files in CRAY_PORTALS environment
4281 Details : The client process capabilities were not honoured on the MDS in
4282 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4283 been accessed by an authorized user then root was able to access
4284 the file on the local client also. The root user capabilities
4285 are now allowed on the MDS, as this environment has secure UID.
4288 Frequency : occasional
4290 Description: ldiskfs "too long searching" message happens too often
4291 Details : A debugging message (otherwise harmless) prints too often on
4292 the OST console. This has been reduced to only happen when
4293 there are fragmentation problems on the filesystem.
4298 Description: Division by zero in statfs when all OSCs are inactive
4299 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4300 lov_statfs divides by value returned by lov_get_stripecnt().
4304 Bugzilla : 9489, 3273
4305 Description: First write from each client to each OST was only 4kB in size,
4306 to initialize client writeback cache, which caused sub-optimal
4307 RPCs and poor layout on disk for the first writen file.
4308 Details : Clients now request an initial cache grant at (re)connect time
4309 and so that they can start streaming writes to the cache right
4310 away and always do full-sized RPCs if there is enough data.
4311 If the OST is rebooted the client also re-establishes its grant
4312 so that client cached writes will be honoured under the grant.
4317 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4318 Details : Now I/O RPCs go to different portal number and (presumably) fast
4319 lock requests (and glimses) and other RPCs get their own service
4320 threads pool that should be able to service those RPCs
4323 Severity : enhancement
4325 Description: Ability to exchange lustre version between client and servers and
4326 issue warnings at client side if client is too old. Also for
4327 liblustre clients there is ability to refuse connection of too old
4329 Details : New 'version' field is added to connect data structure that is
4330 filled with version info. That info is later checked by server and
4334 Frequency : rare, liblustre only.
4335 Bugzilla : 9296, 9581
4336 Description: Two simultaneous writes from liblustre at offset within same page
4337 might proceed at the same time overwriting eachother with stale
4339 Details : I/O lock withing llu_file_prwv was released too early, before data
4340 actually was hitting the wire. Extended lock-holding time until
4341 server acknowledges receiving data.
4344 Frequency : extremely rare. Never observed in practice.
4346 Description: avoid generating lustre_handle cookie of 0.
4347 Details : class_handle_hash() generates handle cookies by incrementing
4348 global counter, and can hit 0 occasionaly (this is unlikely, but
4349 not impossible, because initial value of cookie counter is
4350 selected randonly). Value of 0 is used as a sentinel meaning
4351 "unassigned handle" --- avoid it. Also coalesce two critical
4352 sections in this function into one.
4354 Severity : enhancement
4356 Description: allow liblustre clients to delegate truncate locking to OST
4357 Details : To avoid overhead of locking, liblustre client instructs OST to
4358 take extent lock in ost_punch() on client's behalf. New connection
4359 flag is added to handle backward compatibility.
4361 Severity : enhancement
4362 Bugzilla : 4928, 7341, 9758
4363 Description: allow number of OST service threads to be specified
4364 Details : a module parameter allows the number of OST service threads
4365 to be specified via "options ost ost_num_threads={N}" in the
4366 OSS's /etc/modules.conf or /etc/modprobe.conf.
4370 Bugzilla : 6146, 9635, 9895
4371 Description: servers crash with bad pointer in target_handle_connect()
4372 Details : In rare cases when a client is reconnecting it was possible that
4373 the connection request was the last reference for that export.
4374 We would temporarily drop the export reference and get a new
4375 one, but this may have been the last reference and the export
4376 was just destroyed. Get new reference before dropping old one.
4378 Severity : enhancement
4379 Frequency : if client is started with failover MDS
4381 Description: Allow multiple MDS hostnames in the mount command
4382 Details : Try to read the configuration from all specified MDS
4383 hostnames during a client mount in case the "primary"
4386 Severity : enhancement
4388 Description: Stop sending data to evicted clients as soon as possible.
4389 Details : Check if the client we are about to send or are sending data to
4390 was evicted already. (Check is done every second of waiting,
4391 for which l_wait_event interface was extended to allow checking
4392 of exit condition at specified intervals).
4395 Frequency : rare, normally only when NFS exporting is done from client
4397 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4399 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4402 Frequency : always, if srand() is called before liblustre initialization
4404 Description: Liblustre uses system PRNG disturbing its usage by user application
4405 Details : Introduce internal to lustre fast and high-quality PRNG for
4406 lustre usage and make liblustre and some other places in generic
4407 lustre code to use it.
4409 Severity : enhancement
4410 Bugzilla : 9477, 9557, 9870
4411 Description: Verify that the MDS configuration logs are updated when xml is
4412 Details : Check if the .xml configuration logs are newer than the config
4413 logs stored on the MDS and report an error if this is the case.
4414 Request --write-conf, or allow starting with --old_conf.
4416 Severity : enhancement
4418 Description: Handle symlinks in the path when checking if Lustre is mounted.
4419 Details : Resolve intermediate symlinks when checking if a client has
4420 mounted a filesystem to avoid duplicate client mounts.
4425 Description: lconf can hit an error exception but still return success.
4426 Details : The lconf command catches the Command error exception at the top
4427 level script context and will exit with the associated exit
4428 status, but doesn't ensure that this exit status is non-zero.
4433 Description: failure of ptlrpc thread startup can cause oops
4434 Details : Starting a ptlrpc service thread can fail if there are a large
4435 number of threads or the server memory is very fragmented.
4436 Handle this without oopsing.
4439 Frequency : always, only if liblustre and non-default acceptor port was used
4441 Description: liblustre cannot connect to servers with non-default acceptor port
4442 Details : tcpnal_set_default_params() was not called and was therefore
4443 ignoring the environment varaible TCPNAL_PORT, as well as other
4444 TCPNAL_ environment variables
4449 Description: two objects could be created on the same OST for a single file
4450 Details : If an OST is down, in some cases it was possible to create two
4451 objects on a single OST for a single file. No problems other
4452 than potential performance impact and spurious error messages.
4456 Bugzilla : 5681, 9562
4457 Description: Client may oops in ll_unhash_aliases
4458 Details : Client dcache may become inconsistent in race condition.
4459 In some cases "getcwd" can fail if the current directory is
4465 Description: Inode refcounting problems in NFS export code
4466 Details : link_raw functions used to call d_instantiate without obtaining
4467 extra inode reference first.
4471 Bugzilla : 9942, 9903
4472 Description: Referencing freed requests leading to crash, memleaks with NFS.
4473 Details : We used to require that call to ll_revalidate_it was always
4474 followed by ll_lookup_it. Also with revalidate_special() it is
4475 possible to call ll_revalidate_it() twice for the same dentry
4476 even if first occurence returned success. This fix changes semantic
4477 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4478 reference on a request referred from the intent.
4479 ll_intent_release() then releases such a request.
4482 Frequency : rare, normally benchmark loads only
4484 Description: unlinked inodes were kept in memory on the client
4485 Details : If a client is repeatedly creating and unlinking files it
4486 can accumulate a lot of stale inodes in the inode slab cache.
4487 If there is no other client load running this can cause the
4488 client node to run out of memory. Instead flush old inodes
4489 from client cache that have the same inode number as a new inode.
4492 Frequency : SLES9 2.6.5 kernel and long filenames only
4493 Bugzilla : 9969, 10379
4494 Description: utime reports stale NFS file handle
4495 Details : SLES9 uses out-of-dentry names in some cases, which confused
4496 the lustre dentry revalidation. Change it to always use the
4500 Frequency : rare, unless heavy write-truncate concurrency is continuous
4501 Bugzilla : 4180, 6984, 7171, 9963, 9331
4502 Description: OST becomes very slow and/or deadlocked during object unlink
4503 Details : filter_destroy() was holding onto the parent directory lock
4504 while truncating+unlinking objects. For very large objects this
4505 may block other threads for a long time and slow overall OST
4506 responsiveness. It may also be possible to get a lock ordering
4507 deadlock in this case, or run out of journal credits because of
4508 the combined truncate+unlink. Solution is to do object truncate
4509 first in one transaction without parent lock, and then do the
4510 final unlink in a new transaction with the parent lock. This
4511 reduces the lock hold time dramatically.
4514 Frequency : rare, 2.4 kernels only
4516 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4517 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4518 before unmounting their filesystem in order to avoid deadlock.
4519 The kernel_locked() function in 2.4 kernels only checks whether
4520 the kernel lock is held, not whether it is this process that is
4521 holding it as 2.6 kernels do.
4526 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4527 Details : The client ptlrpc code may be trying to reconnect to a down
4528 server before a previous connection attempt has timed out.
4529 Increase the reconnect interval to be longer than the connection
4530 timeout interval to avoid sending duplicate connections to
4534 Frequency : echo_client brw_test command
4536 Description: fix echo_client to work with OST preallocated code
4537 Details : OST preallocation code (5137) didn't take echo_client IO path
4538 into account: echo_client calls filter methods outside of any
4539 OST thread and, hence, there is no per-thread preallocated
4540 pages and buffers to use. Solution: hijack pga pages for IO. As
4541 a byproduct, this avoids unnecessary data copying.
4545 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4546 Description: Client can oops in mdc_commit_close() after open replay
4547 Details : It was possible for the MDS to return an open request with no
4548 transaction number in mds_finish_transno() if the client was
4549 evicted, but without actually returning an error. Clients
4550 would later try to replay that open and may trip an assertion
4551 Simplify the client close codepath, and always return an error
4552 from the MDS in case the open is not successful.
4555 Frequency : rare, 2.6 OSTs only
4557 Description: OST may deadlock under high load on fragmented files
4558 Details : If there was a heavy load and highly-fragmented OST filesystems
4559 it was possible to have all the OST threads deadlock waiting on
4560 allocation of biovecs, because the biovecs were not released
4561 until the entire RPC IO was completed. Instead, release biovecs
4562 as soon as they are complete to ensure forward IO progress.
4564 Severity : enhancement
4566 Description: Support for specifying external journal device at mount
4567 Details : If an OST or MDS device is formatted with an external journal
4568 device, this device major/minor is stored in the ext3 superblock
4569 and may not be valid for failover. Allow detecting and
4570 specifying the external journal at mount time.
4575 Description: Mounting an MDS with pending unlinked files may cause oops
4576 Details : target_finish_recovery() calls mds_postrecov() which returned
4577 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4578 considers this an error and immediately begins cleaning up the
4579 lov, just after starting the mds_lov process
4581 Severity : enhancement
4583 Description: Implement 'lfs df' to report actual free space on per-OST basis
4584 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4585 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4586 usage of objects; '-h' to report in human readable format.
4588 ------------------------------------------------------------------------------
4590 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4597 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4598 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4599 at startup during orphan unlinking. Instead, skip buddy update
4600 before setup, as it will be regenerated after recovery is complete.
4605 Description: If an OST is inactive, its locks might reference stale inodes.
4606 Details : lov_change_cbdata() must iterate over all namespaces, even if
4607 they are inactive to clear inode references from the lock.
4609 Severity : enhancement
4610 Frequency : occasional, if non-standard max_dirty_mb used
4612 Description: Client will block write RPCs if not enough grant
4613 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4614 then the client will block writes while waiting for another RPC
4615 to complete instead of consuming its dirty limit. With change
4616 we get improved performance when max_dirty_mb is small.
4618 Severity : enhancement
4619 Bugzilla : 3389, 6253
4620 Description: Add support for supplementary groups on the MDS.
4621 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4622 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4623 lookups for user supplementary groups into a cache.
4627 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4628 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4629 for liblustre O_EXCL is enforced by MDS.
4633 Description: readdir never returns NULL in liblustre.
4634 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4635 dentry in struct dirent.
4639 Frequency : liblustre only, depends on application IO pattern
4640 Description: liblustre clients evicted if not contacting servers
4641 Details : Don't put liblustre clients into the ping_evictor list, so
4642 they will not be evicted by the pinger ever.
4644 Severity : enhancement
4646 Description: Add ability to evict clients by NID from MDS.
4647 Details : By echoing "nid:$NID" string into
4648 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4649 $NID would be instantly evicted from this MDS and from all active
4650 OSTs connected to it.
4654 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4655 Details : lookup_it_finish() used to query file size from OSTs that was not
4660 Description: service threads change working directory to that of init
4661 Details : Starting lustre service threads may pin the working directory
4662 of the parent thread, making that filesystem busy. Threads
4663 now change to the working directory of init to avoid this.
4667 Frequency : during shutdown only
4668 Description: shutdown with a failed MDS or OST can cause unmount to hang
4669 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4674 Frequency : 2.6 only
4675 Description: chmod/chown may include an extra supplementary group
4676 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4677 supplementary group and if none is specified this is used.
4680 Bugzilla : 5479 (6816)
4682 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4683 Details : If lookup is in progress on a file that is unlinked we might try
4684 to revalidate the inode and fail in revalidate after lookup is
4685 complete and ll_file_open() enqueues the open again but
4686 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4689 Frequency : always, if lconf --abort_recovery used
4691 Description: lconf --abort_recovery fails with 'Operation not supported'
4692 Details : lconf was attempting to abort recovery on the MDT device and not
4695 ------------------------------------------------------------------------------
4697 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4702 Frequency : rare (only unsupported configurations with a node running as an
4704 Bugzilla : 6514, 5137
4705 Description: Mounting a Lustre file system on a node running as an OST could
4707 Details : OSTs now preallocates memory needed to write out data at
4708 startup, instead of when needed, to avoid having to
4709 allocate memory in possibly low memory situations.
4710 Specifically, if the file system is mounted on on OST,
4711 memory pressure could force it to try to write out data,
4712 which it needed to allocate memory to do. Due to the low
4713 memory, it would be unable to do so and the node would
4714 become unresponsive.
4716 Severity : enhancement
4718 Description: Addition of lconf --service command line option
4719 Details : lconf now accepts a '--service <arg>' option, which is
4720 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4722 Severity : enhancement
4724 Description: Failover mode is now the default for OSTs.
4725 Details : By default, OSTs will now run in failover mode. To return to
4726 the old behaviour, add '--failout' to the lmc line for OSTs.
4728 Severity : enhancement
4730 Description: Health checks are now provided for MDS and OSTs
4731 Details : Additional detailed health check information on MSD and OSTs
4732 is now provided through the procfs health_check value.
4735 Frequency : occasional, depends on IO load
4737 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4738 after numerous create/delete cycles
4739 Details : The ext3 inode allocation policy would not allocate new inodes
4740 very well on the OSTs because there are no new directories
4741 being created. Instead we look for groups with free space if
4742 the parent directories are nearly full.
4747 Description: Network or server problems during mount may cause partially
4748 mounted clients instead of returning an error.
4749 Details : The config llog parsing code may overwrite the error return
4750 code during mount error handling, returning success instead
4756 Description: MDS can fail to allocate large reply buffers
4757 Details : After long uptimes the MDS can fail to allocate large reply
4758 buffers (e.g. zconf client mount config records) due to memory
4759 fragmentation or consumption by the buffer cache. Preallocate
4760 some large reply buffers so that these replies can be sent even
4761 under memory pressure.
4765 Frequency : rare (liblustre)
4766 Description: fsx running with liblustre complained that using truncate() to
4767 extend the file doesn't work. This patch corrects that issue.
4768 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4769 ATTR_SIZE and lsm use in llu_setattr_raw.
4773 Frequency : rare, only 2.6 kernels
4774 Description: Unusual file access patterns on the MDS may result in inode
4775 data being lost in very rare circumstances.
4776 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4777 code in the 2.6 kernel caused the inode and/or EA data not to
4778 be read from disk, causing single-file corruption.
4782 Frequency : rare, only 2.6 filesystems using extents
4783 Description: Heavy concurrent write and delete load may cause data corruption.
4784 Details : It was possible under high-load situations to have an extent
4785 metadata block in the block device cache from a just-unlinked
4786 file overwrite a newly-allocated data block. We now unmap any
4787 metadata buffers that alias just-allocated data blocks.
4791 Frequency : filesystems with default stripe_count larger than 77
4792 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4793 Details : lconf specifies an inode size of 4096 bytes when the default
4794 stripe_count is larger than 77. This conflicts with the default
4795 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4796 this case to avoid pinning too much memory for large EAs.
4798 ------------------------------------------------------------------------------
4800 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4805 Frequency : rare (extremely heavy IO load with hundreds of clients)
4807 Description: Client is evicted, gets IO error writing to file
4808 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4809 caused clients to be evicted for AST timeouts. The fixes in
4810 bug 5192 mean we no longer need to have such short AST timeouts
4811 so ldlm_timeout has been increased.
4814 Frequency : occasional during --force or --failover shutdown under load
4815 Bugzilla : 5949, 4834
4816 Description: Server oops/LBUG if stopped with --force or --failover under load
4817 Details : a collection of import/export refcount and cleanup ordering
4818 issues fixed for safer force cleanup
4821 Frequency : only filesystems larger than 120 OSTs
4822 Bugzilla : 5990, 6223
4823 Description: lfs getstripe would oops on a very large filesystem
4824 Details : lov_getconfig used kfree on vmalloc'd memory
4827 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4828 Bugzilla : 6242, 6243
4829 Description: reading from files that had been truncated to a non-zero size
4830 but never opened returned no data
4831 Details : ll_file_read() reads zeros from no-object files to EOF
4836 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4837 Details : There's little harm in aborting MDS/OSS recovery and letting it
4838 try again, so I removed the LASSERT and return an error instead.
4840 Severity : enhancement
4842 Description: New debugging infrastructure for tracking down data corruption
4843 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4844 (b) cover more of the client-side code path, and (c) try to narrow
4845 down where problems occurred
4849 Bugzilla : 3819, 4364, 4397, 6313
4850 Description: Racing close and eviction MDS could cause assertion in mds_close
4851 Details : It was possible to get multiple mfd references during close and
4852 client eviction, leading to one thread referencing a freed mfd.
4854 Severity: : enhancement
4855 Bugzilla : 3262, 6359
4856 Description: Attempts to reconnect to servers are now more aggressive.
4857 Details : This builds on the enhanced upcall-less recovery that was added
4858 in 1.4.2. When trying to reconnect to servers, clients will
4859 now try each server in the failover group every 10 seconds. By
4860 default, clients would previously try one server every 25 seconds.
4865 Description: After recovery, certain operations trigger a failed
4866 assertion on a client.
4867 Details : Failing over an mds, using lconf -d --failover, while a
4868 client was doing a readdir() call would cause the client to
4869 LBUG after recovery completed and the readdir() was resent.
4871 Severity : enhancement
4873 Description: Default groups are now added by lconf
4874 Details : You can now run lconf --group <servicename> without having to
4875 manually add groups with lmc.
4878 Frequency : occasional
4880 Description: Nodes with an elan id of 0 trigger a failed assertion
4883 Frequency : always when accessing e.g. tty/console device nodes
4885 Description: tty and some other devices nodes cannot be used on lustre
4886 Details : file's private_data field is used by device data and lustre
4887 values in there got lost. New field was added to struct file to
4888 store fs-specific private data.
4891 Frequency : when exporting Lustre via NFS
4893 Description: NFSD failed occasionally when looking up a path component
4894 Details : NFSD is looking up ".." which was broken in ext3 directories
4895 that had grown large enough to become hashed.
4898 Frequency : Clusters with multiple interfaces not on the same subnet
4900 Description: Nodes will repeatedly try to reconnect to an interface which it
4901 cannot reach and report an error to the log.
4902 Details : Extra peer list entries will be created by lconf with some peers
4903 unreachable. lconf now validates the peer before adding it.
4906 Frequency : Only if a default stripe is set on the filesystem root.
4908 Description: Setting a default stripe on the filesystem root prevented the
4909 filesystem from being remounted.
4910 Details : The client was sending extra request flags in the root getattr
4911 request and did not allocate a reply buffer for the dir EA.
4914 Frequency : occasional, higher if lots of files are accessed by one client
4915 Bugzilla : 6159, 6097
4916 Description: Client trips assertion regarding lsm mismatch/magic
4917 Details : While revalidating inodes the VFS looks up inodes with ifind()
4918 and in rare cases can find an inode that is being freed.
4919 The ll_test_inode() code will free the lsm during ifind()
4920 when it finds an existing inode and then the VFS later attaches
4921 this free lsm to a new inode.
4925 Bugzilla : 6422, 7030
4926 Description: MDS deadlock between mkdir and client eviction
4927 Details : Creating a new file via mkdir or mknod (starting a transaction
4928 and getting the ns lock) can deadlock with client eviction
4929 (gets ns lock and trying to finish a synchronous transaction).
4932 Frequency : occasional
4933 Description: While starting a server, the fsfilt_ext3 module could not be
4935 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4936 kernels. Previously, lconf would still use the ext3 name
4937 when trying to load modules. Now, it will correctly use
4938 ext3 on 2.4 and ldiskfs on 2.6.
4940 Severity : enhancement
4941 Description: The default stripe count has been changed to 1
4942 Details : The interpretation of the default stripe count (0, to lfs
4943 or lmc) has been changed to mean striping across a single
4944 OST, rather than all available. For general usage we have
4945 found a stripe count of 1 or 2 works best.
4947 Severity : enhancement
4948 Description: Add support for compiling against Cray portals.
4949 Details : Conditional compiling for some areas that are different
4953 Frequency : occasional
4954 Bugzilla : 6409, 6834
4955 Description: Creating files with an explicit stripe count may lead to
4956 a failed assertion on the MDS
4957 Details : If some OSTs are full or unavailable, creating files may
4958 trigger a failed assertion on the MDS. Now, Lustre will
4959 try to use other servers or return an error to the
4963 Frequency : occasional
4965 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4966 caused lock timeout during readahead (since 1.4.2).
4967 Details : Processes doing readahead might match a lock that hasn't been
4968 granted yet if there are overlapping and conflicting lock
4969 requests. The readahead process waits on ungranted lock
4970 (original lock is CBPENDING), while OST waits for that process
4971 to cancel CBPENDING read lock and eventually evicts client.
4973 Severity : enhancement
4975 Description: Initial enabling of flock support for clients
4976 Details : Implements fcntl advisory locking and file status functions.
4977 This feature is provided as an optional mount flag (default
4978 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4979 locking are implemented yet, and those that are are not guaranteed
4980 to be completely correct in production environments.
4981 mount -t lustre -o [flock|noflock] ...
4984 Frequency : occasional
4986 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4987 assertion in the ext3 JBD (journaling) layer.
4988 Details : The b_committed_data struct is protected by the big kernel lock
4989 in 2.4 kernels, serializing journal_commit_transaction() and
4990 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4991 access to this struct. In 2.6 kernels there is finer grained
4992 locking to improve SMP performance of the JBD layer.
4996 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5001 Description: merge and cleanup kernel patches.
5004 -----------------------------------------------------------------------------
5006 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5008 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5009 versions of Lustre. Please update all servers and clients to
5010 version 1.4.2 or later at the same time. You must also run
5011 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5012 to update the configuration logs.
5014 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5015 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5016 - swab llog records of type '0' so we get proper header size/idx (5861)
5017 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5018 - fix rename of one directory over another leaking an inode (5953)
5019 - avoid SetPageDirty on 2.6 (5981)
5020 - don't re-add just-being-destroyed locks to the waiting list (5653)
5021 - when creating new directories, inherit the parent's custom
5022 striping settings if present parent (3048)
5023 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5024 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5025 - don't hold client locks on temporary worklist from l_lru (5666)
5026 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5027 - automatically evict dead clients (5921)
5028 - Update file size properly in create+truncate+fstat case (6196)
5029 - Do not unhash mountpoint dentries, do not allow removal of
5031 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5032 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5033 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5034 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5035 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5036 - Correctly handle reads of files with no objects (6243)
5037 - lctl recover will also mark a device active if deactivate used (5933)
5039 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5040 - allow --write-conf on an MDS with different nettype than client (5619)
5041 - don't write config llogs to MDS for mounts not from that MDS (5617)
5042 - lconf should create multiple TCP connections from a client (5201)
5043 - init scripts are now turned off by default; run chkconfig --on
5044 lustre and chkconfig --on lustrefs to use them
5045 - upcalls are no longer needed for clients to recover to failover
5047 - add --abort-recovery option to lconf to abort recovery on device
5049 - add support for an arbitrary number of OSTs (3026)
5050 - Quota support protocol changes.
5051 - forward compatibility changes to wire structs (6007)
5052 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5053 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5054 - improved "lustre status" script
5055 - initialize blocksize for non-regular files (6062)
5056 - added --disable-server and --disable-client configure options (5782)
5057 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5058 - Vanilla 2.4.29 support
5059 - increase maximum number of obd devices to 520 (6242)
5060 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5061 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5062 SLES 9 2.6.5 kernels
5063 - limit stripes per file to 160 (the maximum EA size) (6093)
5065 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5068 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5069 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5070 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5071 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5072 - small ext3 extents cleanups and fixes (5733)
5073 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5074 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5075 - avoid needless client->OST connect, fix handle mismatch (5317)
5076 - fix DLM error path that led to out-of-sync client, long delays (5779)
5077 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5078 - fix several locking issues related to i_size (5492,5624,5654,5672)
5079 - don't move pending lock onto export if it is already evicted (5683)
5080 - fix kernel oops when creating .foo in unlinked directory (5548)
5081 - fix deadlock in obdfilter statistics vs. object create (5811)
5082 - use time_{before,after} to avoid timer jiffies wrap (5882)
5083 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5084 - Do not leak request if server was not able to process it (5154)
5085 - If mds_open unable to find parent dir, make that negative lookup(5154)
5086 - don't create new directories with extent-mapping (5909, 5936)
5088 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5089 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5090 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5091 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5092 - support for RHEL4 kernel on i686 (5773)
5093 - provide error messages when incompatible logs are encountered (5898)
5095 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5096 * version 1.4.0.10 (1.4.1 release candidate 1)
5098 - don't keep a lock reference when lock is not granted (4238)
5099 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5100 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5101 - drop import inflight refcount on signal_completed_replay error (5255)
5102 - unlock page after async write error during send (3677)
5103 - handle missing objects in filter_preprw_read properly (5265)
5104 - no transno return for symlink open, don't save no-trasno open (3440)
5105 - don't try to complete elan receive that already failed (4012)
5106 - free RPC server reply state on error (5406)
5107 - clean up thread from ptlrpc_start_thread() on error (5160)
5108 - readahead could read extra page into cache that wasn't ejected (5388)
5109 - prevent races in class_attach/setup/cleanup/detach (5260)
5110 - don't dereference de->d_inode after l_dput of de (5458)
5111 - use "int" for stripe value returned from lock_to_stripe (5544)
5112 - mballoc allocation and error-checking fixes in 2.6 (5504)
5113 - block device patches to fix I/O request sizes in 2.6 (5482)
5114 - look up hostnames for IB nals (5602)
5115 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5116 - don't start multiple acceptors for the same port (5277)
5117 - fix incorrect LASSERT in mds_getattr_name (5635)
5118 - export a proc file for general "ping" checking (5628)
5119 - fix "lfs check" to not block when the MDS is down (5628)
5121 - service request history (4965)
5122 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5123 - create an "evict_client" /proc entry on OSTs, like the MDS has
5124 - fix mount usage message, return errors per mount(8) (5168)
5125 - change grep [] to grep "[]" in tests so they work in more UMLs
5126 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5127 - remove ancient LOV_MAGIC_V0 EA support (5047)
5128 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5129 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5130 - allow readahead allocations to fail when low on memory (5383)
5131 - mmap locking landed again, after considerable improvement (2828)
5132 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5134 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5137 - send OST transaction number in read/write reply to free req (4966)
5138 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5139 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5140 - the watchdog thread now runs as interruptible (5246)
5141 - flock/lockf fixes (but it's still disabled, pending 5135)
5142 - don't use EXT3 constants in llite code (5094)
5143 - memory shortage at startup could cause assertion (5176)
5145 - reorganization of lov code
5146 - single portals codebase
5148 - add extents/mballoc support (5025)
5149 - direct I/O reads in the obdfilter (4048)
5150 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5152 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5155 - send OST transaction number in read/write reply to free req (4966)
5156 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5157 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5158 - the watchdog thread now runs as interruptible (5246)
5159 - handle missing objects in filter_preprw_read properly (5265)
5160 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5161 - drop import inflight refcount on signal_completed_replay error (5255)
5162 - unlock page after async write error during send (3677)
5163 - return original error code on reconstructed replies (3761)
5164 - no transno return for symlink open, don't save no-trasno open (3440)
5166 - add pid to ldlm debugging output (4922)
5167 - bump the watchdog timeouts -- we can't handle 30sec yet
5168 - extra debugging for orphan dentry/inode bug (5259)
5170 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5173 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5174 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5175 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5176 - fix NULL dereference in /proc/sys/portals/routes (4827)
5177 - allow failed mdc_close() operations to be interrupted (4561)
5178 - stop precreate on OST before MDS would time out on it (4778)
5179 - don't send partial-page writes before EOF from client (4410)
5180 - discard client grant for sub-page writes on large-page clients (4520)
5181 - don't free dentries not owned by NFS code, check generation (4806)
5182 - fix lsm leak if mds_create_objects() fails (4801)
5183 - limit debug_daemon file size, always print CERROR messages (4789)
5184 - use transno after validating reply (3892)
5185 - process timed out requests if import state changes (3754)
5186 - update mtime on OST during writes, return in glimpse (4829)
5187 - add mkfsoptions to LDAP (4679)
5188 - use ->max_readahead method instead of zapping global ra (5039)
5189 - don't interrupt __l_wait_event() during strace
5191 - add software watchdogs to catch hung threads quickly (4941)
5192 - make lustrefs init script start after nfs is mounted
5193 - fix CWARN/ERROR duplication (4930)
5194 - return async write errors to application if possible (2248)
5195 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5196 - print NAL number in %x format (4645)
5197 - update barely-supported suse-2.4.21-171 series (4842)
5198 - support for sles 9 %post scripts
5199 - support for building 2.6 kernel-source packages
5200 - support for sles km_* packages
5202 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5205 - ignore -ENOENT errors in osc_destroy (3639)
5206 - notify osc create thread that OSC is being cleaned up (4600)
5207 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5208 - reconstruct ost_handle() like mds_handle() (4657)
5209 - create a new thread to do import eviction to avoid deadlock (3969)
5210 - let lconf resolve symlinked-to devices (4629)
5211 - don't unlink "objects" from directory with default EA (4554)
5212 - hold socknal file ref over connect in case target is down (4394)
5213 - allow more than 32000 subdirectories in a single directory (3244)
5214 - fix blocks count for O_DIRECT writes (3751)
5215 - OST returns ENOSPC from object create when no space left (4539)
5216 - don't send truncate RPC if file size isn't changing (4410)
5217 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5218 - bind to privileged port in socknal and tcpnal (3689)
5220 - rate limit CERROR/CWARN console message to avoid overload (4519)
5221 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5222 - basic mmap support (3918)
5223 - kernel patch series update from b1_4 (4711)
5225 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5228 - avoid crash during MDS cleanup with OST shut down (2775)
5229 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5230 - don't use bad inodes on the MDS (3744)
5231 - dynamic object preallocation to improve recovery speed (4236)
5232 - don't hold spinlock over lock dumping or change debug flags (4401)
5233 - don't zero obd_dev when it is force cleaned (3651)
5234 - print grants to console if they go negative (4431)
5235 - "lctl deactivate" will stop automatic recovery attempts (3406)
5236 - look for existing locks in ldlm_handle_enqueue() (3764)
5237 - don't resolve lock handle twice in recovery avoiding race (4401)
5238 - revalidate should check working dir is a directory (4134)
5240 - don't always mark "slow" obdfilter messages as errors (4418)
5242 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5245 - don't close LustreDB during write_conf until it is done (3860)
5246 - fix typo in lconf for_each_profile (3821)
5247 - allow dumping logs from multiple threads at one time (3820)
5248 - don't allow multiple threads in OSC recovery (3812)
5249 - fix debug_size parameters (3864)
5250 - fix mds_postrecov to initialize import for llog ctxt (3121)
5251 - replace config semaphore with spinlock (3306)
5252 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5253 - don't allow enqueue to complete on a destroyed export (3822)
5254 - down write_lock before checking llog header bitmap (3825)
5255 - recover from lock replay timeout (3764)
5256 - up llog sem before sending rpc (3652)
5257 - reduce ns lock hold times when setting kms (3267)
5258 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5259 - fix NULL deref and obd_dev leak on setup error (3312)
5260 - replace some LBUG about llog ops with error handling (3841)
5261 - don't match INVALID dentries from d_lookup and spin (3784)
5262 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5263 - fix invalid assertion in ptlrpc_set_wait (3880)
5265 - add libwrap support for the TCP acceptor (3996)
5266 - add /proc/sys/portals/routes for non-root route listing (3994)
5267 - allow setting MDS UUID in .xml (2580)
5268 - print the stack of a process that LBUGs (4228)
5270 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5273 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5274 - make sure to unset replay flag from failed open requests (3440)
5275 - if default stripe count is 0, use OST count for inode size (3636)
5276 - update parent mtime/ctime on client for create/unlink (2611)
5277 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5278 - free recovery state on server during a forced cleanup (3571)
5279 - unregister_reply for resent reqs (3063)
5280 - loop back devices mounting and status check on 2.6 (3563)
5281 - fix resource-creation race that can provoke i_size == 0 (3513)
5282 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5283 - more debugging for page-accounting assertion (3746)
5284 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5285 - avoid deadlock after precreation failure (3758)
5286 - fix race and lock order deadlock in orphan handling (3450, 3750)
5287 - add validity checks when grabbing inodes from l_ast_data (3599)
5289 - add /proc/.../recovery_status to obdfilter (3428)
5290 - lightweight CDEBUG infrastructure, debug daemon (3668)
5291 - change default OSC RPC parameters to be better on small clusters
5292 - turn off OST read cache for files smaller than 32MB
5293 - install man pages and include them in rpms (3100)
5294 - add new init script for (un)mounting lustre filesystems (2593)
5295 - run chkconfig in %post for init scripts (3701)
5296 - drop scimac NAL (unmaintained)
5298 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5301 - clean kiobufs before and after use (3485)
5302 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5303 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5304 - change init script to fail more clearly if not run as root (1528)
5305 - allow clients to reconnect during replay (1742)
5306 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5307 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5308 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5309 - fix cond_resched (3554)
5310 - only evict unfinished clients after recovery (3515)
5311 - allow bulk resend, prevent data loss (3570)
5312 - dynamic ptlrpc request buffer allocation (2102)
5313 - don't allow unlinking open directory if it isn't empty (2904)
5314 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5315 - remove extraneous obd dereference causing LASSERT failure (3334)
5316 - don't use get_cycles() when creating temp. files on the mds (3156)
5317 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5318 - handle EEXIST for set-stripe, set proper directory name (3336)
5320 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5321 - fix ksocknal_fmb_callback() error messages (2918)
5323 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5326 - don't copy lvb into (possibly NULL) reply on error (2983)
5327 - don't deref dentry after dput, don't free lvb on error (2922)
5328 - use the kms to determine writeback rpc length (2947)
5329 - increment oti_logcookies when osc is inactive (2948)
5330 - update client's i_blocks count via lvb messages (2543)
5331 - handle intent open/close of special files properly (1557)
5332 - mount MDS with errors=remount-ro, like obdfilter (2009)
5333 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5334 - don't use cancelling-locks' kms values (2947)
5335 - use highest lock extent for kms, not last one (2925)
5336 - don't dereference ERR_PTR() dentry in error handling path (3107)
5337 - fix thread race in portals_debug_dumplog() (3122)
5338 - create lprocfs device entries at setup instead of at attach (1519)
5339 - common AST error handler, don't evict client on completion race (3145)
5340 - zero nameidata in detach_mnt in 2.6 (3118)
5341 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5342 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5343 - initialize RPC timeout timer earlier for 2.6 (3219)
5344 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5345 - print nal/nid for unknown nid (3258)
5346 - additional checks for oscc recovery before doing precreate (3284)
5347 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5348 - don't crash in mdc_close for bad permissions on open (3285)
5349 - zero i_rdev for non-device files (3147)
5350 - clear page->private before handing to FS, better assertion (3119)
5351 - tune the read pipeline (3236)
5352 - fix incorrect decref of invalidated dentry (2350)
5353 - provide read-ahead stats and refine rpc in flight stats (3328)
5354 - don't hold journal transaction open across create RPC (3313)
5355 - update atime on MDS at close time (3265)
5356 - close LDAP connection when recovering to avoid server load (3315)
5357 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5358 - don't leak open file on MDS after open resend (3325)
5359 - serialize filter_precreate and filter_destroy_precreated (3329)
5360 - loop device shouldn't call sync_dev() for nul device (3092)
5361 - clear page cache after eviction (2766)
5362 - resynchronize MDS->OST in background (2824)
5363 - refuse to mount the same filesystem twice on same mountpoint (3394)
5364 - allow llmount to create routes for mounting behind routers (3320)
5365 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5366 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5367 - fix rare problem with rename on htree directories (3417)
5369 - allow default OST striping configuration per directory (1414)
5370 - fix compilation for qswnal for 2.6 kernels (3125)
5371 - increase maximum number of MDS request buffers for large systems
5372 - change liblustreapi to be useful for external progs like lfsck (3098)
5373 - increase local configuration timeout for slow disks (3353)
5374 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5376 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5379 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5380 - don't overwrite extent policy data in reply if lock was blocked (2901)
5381 - drop filter export grants atomically with removal from device (2663)
5382 - del obd_self_export from work_list in class_disconnect_exports (2908)
5383 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5384 - swab reply message in mdc_close, other PPC fixes (2464)
5385 - fix destroying of named logs (2325)
5386 - overwrite old logs when running lconf --write_conf (2264)
5387 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5388 - fix race in target_handle_connect (2898)
5389 - mds_reint_create() should take same inode create lock (2926)
5390 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5391 - don't close files for self_export to avoid uninitialized obd (2936)
5392 - allow MDS with the same name as client node (2939)
5393 - hold dentry reference for closed log files for unlink (2325)
5394 - reserve space for all logs during transactions (2059)
5395 - don't evict page beyond end of stripe extent (2925)
5396 - don't oops on a deleted current working directory (2399)
5397 - handle hard links to targets without a parent properly (2517)
5398 - don't dereference NULL lock when racing during eviction (2867)
5399 - don't grow lock extents when lots of conflicting locks (2919)
5401 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5404 - account for cache space usage on clients to avoid data loss (974)
5405 - lfsck support in lustre kernel code (2349)
5406 - reduce journal credits needed for BRW writes (2370)
5407 - orphan handling to avoid losing space on client/server crashes
5408 - ptlrpcd can be blocked, stopping ALL progress (2477)
5409 - use lock value blocks to assist in proper KMS, faster stat (1021)
5410 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5411 - recovery for initial connections (2355)
5412 - fixes for mds_cleanup_orphans (1934)
5413 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5414 - block all file creations until orphan recovery completes (1901)
5415 - client remove rq_connection from request struct (2423)
5416 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5417 - recovery timer race (2670)
5418 - mdc_close recovey bug (2532)
5419 - ptlrpc cleanup bug (2710)
5420 - mds timeout on local locks (2588)
5421 - namespace lock held during RPCs (2431)
5422 - handle interrupted sync write properly (2503)
5423 - don't try to handle a message that hasn't been replied to (2699)
5424 - client assert failure during cleanup after abort recovery (2701)
5425 - leak mdc device after failed mount (2712)
5426 - ptlrpc_check_set allows timedout requests to complete (2714)
5427 - wait for inflight reqs when ptlrpcd finishes (2710)
5428 - make sure unregistered services are removed from the srv_list
5429 - reset bulk XID's when resending them (caught by 1138 test)
5430 - unregister_bulk after timeout
5431 - fix lconf error (2694)
5432 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5433 - readahead locks pages, leaves pending causing memory pressure (2673)
5434 - increase OST request buffers to 4096 on large machines (2729)
5435 - fix up permission of existing directories in simple_mkdir (2661)
5436 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5437 - don't assign transno to errored transactions (2742)
5438 - don't delete objects on OST if given a bogus objid from MDS (2751)
5439 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5440 - if rq_no_resend, then timeout request after recovery (2432)
5441 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5442 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5443 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5444 - MDS recovery shouldn't delete objects that it has given out (2730)
5445 - if enqueue arrives after completion, don't clobber LVB (2819)
5446 - don't unlock pages twice when trigger_group_io returns error (2814)
5447 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5448 - don't write pages to disk if there was an error (1450)
5449 - don't ping imports that have recovery disabled (2676)
5450 - take buffered bytes into account when balancing socknal conn (2817)
5451 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5452 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5453 - remove little-endian swabbing of llog records (1987)
5454 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5455 - retry reposting request buffers if they fail (1191)
5456 - grow extent at grant time to avoid granting a revoked lock (2809)
5457 - lock revoke doesn't evict page if covered by a second lock (2765)
5458 - disable VM readahead to avoid reading outside lock extents (2805)
5460 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5461 - updated kernel patches for hp-2.4.20 kernel (2681)
5463 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5466 - fix truncated write corruption (2366)
5467 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5469 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5470 - ensure len(uuid) < 37 in lmc (1171)
5471 - fix ia64 OOPS in llog_test (2255)
5472 - zero end of page at obdfilter for partial page writes (2648)
5473 - don't leave stale dentries around after renames (bug 2428)
5474 - fix timeouts when evicting a client with a single lock held (2642)
5475 - set deadline for the initial HELLO message to drain (2634)
5476 - print out dotted-quad IP addresses in the socknal (2302)
5478 - additional debugging for MDS client eviction problem (2443)
5479 - fix mkfsoptions support for osts (2603, 2604)
5481 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5484 - add series for the vanilla 2.6.0 kernel
5485 - add series for the vanilla 2.4.24 kernel
5486 - add series for a cray x86/64 UL kernel drop
5487 - fix xattr patches for the vanilla 2.4.19 series
5489 - generate true UUIDs in lmc (1171)
5490 - have portals stack dumping break in UML (2466)
5491 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5492 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5493 - wake callback waiting threads on client eviction (2460)
5494 - Add --ptldebug and --subsystem to lmc (1719)
5495 - update assertion to allow safe interrupt allocation
5496 - set rq_no_resend for cancel requests (2432)
5497 - recalculate ptlrpcd timeout after resend (2494)
5498 - call vfs_rmdir when removing pending directories (2368)
5499 - fix renaming a file to itself (2429)
5500 - lmc creates a default one-stripe lov (2454)
5501 - expand procfs space to handle large clusters (2326)
5502 - increase UML stack to avoid overflow
5503 - update lconf's list of debug and subsystem masks
5504 - fix lfs find --obd (2510)
5505 - /proc tunable for disabling filter read caching (2591)
5506 - stop rpm packages from altering slapd.conf (2301)
5507 - disable nagle in the socknal under 0conf (2578)
5508 - choose mds inode size based on stripe count (2572)
5509 - fix kernel-source rpm problems (2516)
5511 - add --disable-doc to avoid pdf generation (2421)
5512 - update documentation, tests, type-os, comments
5513 - avoid format warnings on ia64
5514 - remove the TOE NAL
5515 - tiny code cleanups by removing unused fields
5517 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5520 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5521 - remove the most blatant lies from BUILDING file (2371)
5522 - change default debug level to reasonable production setting
5523 - reduce client side cache size to reduce cache flush time
5524 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5525 - make TCP zerocopy and pinger support enabled by default (2476)
5526 - sync writes completed after process exits caused crashes (2319)
5527 - maintain correct mount count on the MDS (2356)
5528 - backout 1557, because 2316 wasn't really fixed
5529 - better file I/O statistics gathering in /proc
5530 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5531 - another kernel patch to fix zero-copy TCP function export
5532 - don't take duplicate lock when processing re-sent getattr (2420)
5533 - lctl uses obd_self_export instead of creating new conn (2353)
5534 - MDS/OST recovery case which requires object creation asserted (2425)
5535 - move lfs from /usr/sbin to /usr/bin in packages
5536 - fix race between mds_client_add and mds_client_free (2417)
5537 - use kmalloc instead of slabs in portals (2430)
5538 - don't create duplicate records when a failover MDS is present (2442)
5539 - remove unnecessary mount age check (2332)
5540 - don't remove directory inodes from locks prematurely (2451)
5541 - don't break if MDS service name is the same as hostname (2103)
5542 - fix races in client write RPC generation when cache full (2482)
5544 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5547 - remove now-unused request->rq_obd (278)
5548 - if an allocation fails, print out how much memory we've used (1933)
5549 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5550 - add the "configurable stack size" patch to most series files (1256)
5551 - ability to write large log records, for 100+ OST configs (2306)
5552 - fix NULL deref when filter_prep fails (2314)
5553 - fix operator precedence error in filter_sync
5554 - dynamic allocation of socknal TX descriptors (2315)
5555 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5556 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5557 - ensure that CWARN messages in Portals always get to the syslog
5558 - __init/__exit are not for prototype decls (ldlm_init/exit)
5559 - x86-64 compile warning fixes
5560 - fix gateway LMC keyword conflict (2318)
5561 - fix MDS lock inversions in getattr/reint paths (1844)
5562 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5563 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5564 - DLM race condition prevented some lock evictions (2328)
5565 - ENOMEM detection and retry on socknal sends (2230)
5566 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5567 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5568 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5569 - better error messages when a client is rejected during recovery (1505)
5570 - avoid cancelling locks which were never granted, after failure (2330)
5571 - fix i_sem/journal inversion in mds_client_add (2333)
5572 - fix truncate/getattr lock cycle deadlock (2334)
5573 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5574 - fix two rare exit paths which could leak an l_lock() ref (2321)
5575 - fencepost error in MDS/OST orphan recovery (2226)
5576 - make log record alignment 8 bytes (1988)
5577 - lstripe now fails when requested offset > ost_count (2237)
5578 - ensure that all kernel series have a complete list.h (1607)
5579 - fix crashes in special-file operations (2316)
5580 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5582 - allow configurable automake binary, for testing new versions
5583 - small update to the lfs documentation
5585 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5587 * fix negative export reference count in fsfilt_sync (2312)
5589 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5590 * release candidate 0.9.1
5592 - orphans are moved into the PENDING directory for possible recovery
5593 - replayed opens now open by fid for orphan/rename safety (1042)
5594 - last close of an orphan inode generates a transno (683)
5595 - chdir() and mount() now pin the directory entry (1020)
5596 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5597 - discard very old requests without processing them (1502)
5598 - remove some common, well-understood CERRORs (1505)
5599 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5600 - clear "grant" flags in OST replies until OST grant code lands (1644)
5601 - fix read performance by not clobbering i_blksize on client (1598)
5602 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5603 - make LRU size a /proc tunable, clears locks when reduced (707)
5604 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5605 - server locks take references on exports now (1558)
5606 - build fixes for 2.4.20-rh trees (1663)
5607 - return an error from lov_create if all OSCs are inactive (1751)
5608 - fix import levels when a reconnect happens without a timeout (1597)
5609 - exit early from mds_open if we get a lookup error (1749)
5610 - partial page read at EOF wouldn't wait for disk before sending (1642)
5611 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5612 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5613 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5614 - write/truncate lock inversion (1639)
5615 - Don't auto-load obdclass, portals modules during cleanup (1495)
5616 - fix timestamps from jumping to "now" (1763)
5617 - extra journal assertions (1648)
5618 - add an extra multiunlink test (1771)
5619 - fix read_record/write_record API (1776)
5620 - fix leak of offset_extent, possible incorrect i_size later (1772)
5621 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5622 - Debugging for the kqswnal_get_idle_tx problems (1820)
5623 - Allow recovery to be attempted multiple times (1536)
5624 - Write out MDS last_rcvd file after it is first created (1600)
5625 - Fix tx_descriptor leak in failed transmit situations (1827)
5626 - ext3 journaling fixes for assertion failure after IO error (1871)
5627 - class_export_put() on freed export after completion AST error (1896)
5628 - Fix revalidate looping in VFS (1322)
5629 - Don't access a freed export during MDS_REINT timeout (1521)
5630 - Add open-unlink recovery support on the MDS (1673,1764)
5631 - Return an error if no MDS data was read from last_rcvd (1946)
5632 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5633 - Don't setup a disk device that doesn't match exported UUID (317)
5634 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5635 - avoid committing NULL handle in force close
5636 - local.sh is now a one-stripe LOV configuration
5637 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5638 - fix ext3 htree duplicate directory entry corruption (1516)
5639 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5640 - update to new LOV EA format (2097)
5641 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5643 2003-06-15 Phil Schwan <phil@clusterfs.com>
5646 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5647 - per-import/export recovery handling (958, 931, 959)
5648 - multiple last-rcvd slots, for serving multiple FSes (949)
5649 - connections are again shared between multiple imp/exports (963, 964)
5650 - "umount -f" would hang if any requests needed to be sent (393, 978)
5651 - avoid pinning large req buffer by copying for queued messages (989)
5652 - add "uuid" to "lctl device" command to help upcalls (991)
5653 - "open" RPCs with transnos would confuse recovery counters (1037)
5654 - do proper endian conversion of all wire messages (288, 340, 891)
5655 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5656 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5657 - fix ptlrpc_connection leak in target_handle_connect (1174)
5658 - fix import refcounting bug in OST and MDS cleanup (1134)
5659 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5660 - fix very unlikely obd_types race condition (501)
5661 - remove osc_open hack for echo_client (1187)
5662 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5663 - a failure in read_inode2 leads to deadlock (1139)
5664 - cancel ack-locks as soon as transaction is committed (1072)
5665 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5666 - make sure to commitrw after any preprw to avoid deadlock (1162)
5667 - failing to execute a file in a lustre FS would lock inode (1203)
5668 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5669 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5670 - obd_uuid2tgt would walk off the end of the list (1255)
5671 - on IA64 the osc would give portals incorrect bulk size (1258)
5672 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5673 - fix lock inversion caused by new llite matching code (1282)
5674 - limit the number of dirty pages on a client to 10MB (1286)
5675 - timed out locks were not being corrected cancelled (1289)
5676 - fix O_DIRECT above 4GB on IA-32 (1292)
5677 * major user-visible changes
5678 - fail out/fail over policy now controlled by the upcall (993)
5680 - add OBD_PING to check server availability and failure (954)
5681 - lustre messages are now sent in sending host order (288, 340, 891)
5682 - add eadatalen to MDS getattr reply (340)
5683 - OST read replies may contain second buffer, with per-page status (593)
5685 2003-03-11 Phil Schwan <phil@clusterfs.com>
5688 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5689 - failed open()s could cause deadlock; fixed (867, 869)
5690 - stop cancelling OST locks when files are closed (481)
5691 - overlapping XID spaces caused network corruption (851, 853)
5692 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5693 - setattr_raw added, to do single-RPC, server-side setattrs
5694 - lmc/lconf syntax change for OST UUIDs
5695 - fix crashy race condition between ptlrpc_free_req and osc_close
5696 - don't use request in mdc_enqueue if we hit a timeout (889)
5697 - don't set the inode i_size for regular files from the MDS (896)
5698 - handle out of order completion AST (842)
5699 - don't LBUG if a lock request times out after receiving AST (913)
5700 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5701 - if a bad lock AST arrives, send an error instead of dropping entirely
5702 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5703 - fix leak in bulk IO when only partially completed (899, 900, 926)
5704 - fix O_DIRECT for ia64 (55)
5705 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5706 - C-z after timeout could hang a process forever; fixed (977)
5708 - client-side I/O cache (678, 924, 929, 941, 970)
5710 - READPAGE and SETATTRs which don't take server-side locks get
5713 2003-02-11 Phil Schwan <phil@clusterfs.com>
5716 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5717 - Fix fsfilt_extN_readpage() to read a full page of directory
5718 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5719 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5720 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5721 - Most intent operations were removed, in favour of a new RPC mode
5722 that does a single RPC to the server and bypasses most of the VFS
5723 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5724 - Aggressively cancel local locks on DLM servers
5725 - mds_reint_unlink sends EA to the client if it's the last nlink.
5726 client uses that EA to unlink OST objects.
5727 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5728 - recursive symlinks were fixed (439)
5729 - fixed NULL deref in DEBUG_REQ
5730 - filter_update_lastobjid no longer calls sync, which annoyed extN
5731 - fixed multi-client small-writes to a single file problem (445)
5732 - fixed mtime updates during file writes (607)
5733 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5734 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5735 - fixed Linux OST size reporting problem (444, 656)
5736 - OST now updates object mtime with writes or setattr (607, 619)
5737 - client verifies file size before zeroing page past EOF (445)
5738 - OST now writes last allocated objid to disk with allocation (108)
5739 - LOV on echo now works (409)
5741 - mds_reint_unlink sends a new buffer, with the EA included. this
5742 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5743 set if a regular file was being unlinked, and it was the last link
5744 - use PtlGet from the target for bulk writes (315)
5745 - OST now updates object mtime with writes or setattr (607, 619)
5746 - LDLM now has a grant-time callback to revalidate locked items, if
5748 - Many MDS operations were reorganized to combat race conditions
5750 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5751 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5753 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5756 - Fully reactivate OST imports after reconnection (512, others)
5757 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5758 - More graceful error handling for truncating on dead OST (515)
5759 - Don't error out unless we're actually accessing dead stripes (474)
5760 - Fix garbage sizes when stripes are missing (410)
5761 - LRU counters were broken, causing constant lock purge (433, 432)
5762 - garbage on read from stripes with failed OSTs (441)
5763 - mark OSCs as active before reconnecting during recovery (438)
5764 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5765 - lfind did not preserve OST order in output (443)
5766 - symlinks cause hung clients, incorrect data (439)
5767 - stop dereferencing request after dropping refcount (457)
5768 - don't LASSERT(spin_is_locked) on non-SMP (455)
5769 - fixes for many rename() bugs
5770 - fstat didn't correctly synchronize attributes (399)
5771 - server must handle lock cancellation during blocking AST prep (487)
5772 - bulk descriptors were free()d too soon (511)
5773 - fix paths in lconf, which would load incorrect modules (451, 507)
5774 - fix confusing lconf 'host not found' error message (386)
5775 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5776 - fix race condition in mdc_blocking_ast() for inode access (526)
5777 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5778 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5779 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5780 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5781 - fix lstripe in several ways (kernel side) (527)
5782 - fix request leak in ldlm_cli_enqueue (262)
5783 - incorrect OSC was marked inactive after OST failure
5784 - call mds_fs_cleanup before unmounting filesystem (524)
5785 - fix races between taking ns_lock and ldlm_lock_change_resource
5786 - fix races updating LOV export open file list
5787 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5788 - fix recovery NULL deref in ldlm_cli_cancel_unused
5789 - fix some DLM races by using new hash table for lock handles (419)
5790 - permit the client to specify desired inodes, at replay
5791 - duplicate requests when we queue them for replay reintegration
5792 - fix last_rcvd offset calculation
5793 - sync after each recovered transaction, so we always make progress
5794 - never, not always, ERESTART requests without transnos
5795 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5797 - skip replay if the MDS says that the client is already connected
5798 - don't check for a recovery-enabled export to match lctl's UUID
5799 - don't INC_USE_COUNT for phantom exports
5800 - don't crash when cleaning up phantom exports (567)
5801 - don't double-finish or set replay data for errored mdc_open requests
5802 - abort requests when they time out, so we don't get old replies
5803 - send/receive replies for AST messages again
5804 - if the client says that it doesn't have the lock, cancel it on the
5806 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5807 mark it as destroyed, it will all work out when decref is called
5808 - fix module use counts (22, 581)
5810 - ASTs now expect a reply (server cancels lock on error reply)
5812 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5815 - fix many simultaneous client startup (392)
5816 - fix dentry->d_it clobbering
5817 - credentials weren't being shipped for readdir/getattr operations
5818 - remove invalid assertions triggered during some concurrent MD
5820 - proper Lustre versions added (336, 389)
5821 - fix memory leak for create error case (398)
5822 - fix LOV locking bug that would get cli/srv out of sync
5823 - fix echo client over LOV (409)
5824 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5825 - fix double-O_EXCL intent crash (424)
5826 - avoid sending multiple lock CANCELs (352)
5828 - MDS can do multi-client recovery (modulo bugs in new code)
5830 - many updates, edits, cleanups
5832 2002-11-18 Phil Schwan <phil@clusterfs.com>
5835 - fix null d_it dereference (346)
5836 - fix full OST/dbench hang (333)
5837 - fix permission problem with file removal (286)
5838 - fix removal of OSCs from LOV when they fail
5839 - fix NULL deref during bulk timeout (214)
5840 - fix problems related to multiple filesystems on one MDS (241)
5841 - fixed serious subtle metadata locking bugs
5842 - free locks on clients when inodes are removed due to memory
5844 - fix inode pointer in lock data (285)
5845 - partial support for multiple MDS on a single host (241)
5846 - data locks weren't cancelled at clear_inode time (290, 311)
5847 - intent locks could lead to unbounded lock growth (205)
5848 - added a maximum lock count, an LRU list, and a flusher
5849 - fix multiple rename (365)
5850 - properly abstracted the echo client
5851 - OSC locked 1 byte too many; fixed
5852 - rewrote brw callback code:
5853 - fixed recovery bugs related to LOVs (306)
5854 - fixed too-many-pages-in-one-write crash (191)
5855 - fixed (again) crash in sync_io_timeout (214)
5856 - probably fixed callback-related race (385)
5858 - Add capability to MDS protocol
5859 - LDLM cancellations and callbacks on different portals
5861 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5864 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5865 - defer open object destruction to close time (601981, 138)
5866 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5867 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5868 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5869 - don't call obd_setattr in ll_file_release for destroyed objects
5870 * protocol change to lustre_msg: move |version| and add |flags|
5871 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5872 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5873 * added replay of create, unlink, link and rename operations during
5874 MDS failover; recovery should be much more robust now
5875 * remove failed OSCs from LOVs (only lov_create uses this so far)
5876 * the lustre-HOWTO was brought (more) up to date (582544)
5878 2002-10-23 Phil Schwan <phil@clusterfs.com>
5881 - in-use dentries weren't being reused properly (617851)
5882 - prevent multiple LDLM setup (599178)
5883 - fix LOV size calculations for truncate (617853)
5884 - fix client handling of MDS intent errors (POSIX)
5885 - fix permission bug in lovstripe.c test (624321)
5886 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5887 - truncate past end of file could corrupt data
5888 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5889 - a race in recovery could return ETIMEDOUT to apps (623947)
5890 - building outside the source directory was fixed
5891 * the lustre-HOWTO was brought (more) up to date (582544)
5892 * major progress was made on recovery functionality
5894 2002-10-10 Phil Schwan <phil@clusterfs.com>
5897 - recovery deadlock fix
5898 - rm -rf causes LBUG fix (617817)
5899 - file open by multiple tasks fix (618962)
5900 - directory permissions bugs (602707 and 620007)
5901 - journal_stop fixed with locking (611313)
5902 - O_APPEND failures resolved (618273, perhaps 614459)
5903 - lconf PATH fix (619770)
5904 - IA64 build fix (621450)
5905 - RPC buffer sizes scale with amount of memory
5907 2002-10-01 Phil Schwan <phil@clusterfs.com>
5910 - locks would be cancelled without throwing away data pages,
5911 resulting in inconsistent data (605627)
5912 - inode attributes were not always being refreshed (605627, 612449)
5913 - lconf now continues to cleanup after lctl reports an error
5914 - MDS now enforces user permissions (602707)
5915 - lprocfs cleanup fixed, but not yet enabled (614157)
5916 - fixed infinite server hang, should a client not respond to an AST
5917 - avoid going into recovery if user calls readlink() with a buffer
5918 that's too small (613941)
5919 - AST RPCs no longer require replies (614867) -- this may be changed
5920 - don't crash server if client sends an IOV that's too big (611336)
5921 - fixed lock conversion deadlock (611892)
5922 - fixed the following of symlinks (614622)
5923 * recovery: the server can remove locks from a client that dies, other
5924 clients can make progress
5925 * more extN patch fixes
5926 * compile-time configurable ptlrpc buffer allocations
5928 - collaborative read cache document
5929 - Lustre Lite Performance CDR document-in-progress
5931 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5934 - fix typo in patch-2.4.18
5936 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5939 - clear ptlrpc request each time in handle_incoming_request()
5940 - unlink of files now destroys the object on the OST
5942 2002-09-19 Peter Braam <braam@clusterfs.com>
5944 * add hard link support
5945 * change obdfile creation method
5946 * kernel patch changed
5948 2002-09-19 Peter Braam <braam@clusterfs.com>
5951 - stack overflow bug in extN fixed
5953 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5955 * documentation updates
5956 - add man pages for config tools
5957 - update tests/README to describe testing with new config tools
5958 - finish metadata API descriptions
5959 * bug fixes and cleanups
5960 - statfs workaround for 16TB limit
5961 - LOV stripe allocation improved, can stripe on subset of OSTs
5962 - LOV file size/IO offset was wrong for files > 4GB in size
5963 - object EA data was being dropped, caused files to be unreadable
5964 - memory overflow with non-LOV OST caused memory corruption
5965 - fixed regression tests to work with new config tools, obdfilter
5966 - fixed bug when directory size became larger than 1 block
5967 - fixed bug (for single client case) when PWD was deleted
5968 - invalidate local directory pages when doing intent-based ops
5969 - avoid LDLM oops when lock callback contained bad data
5971 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5973 * documentation updates
5974 * bug fixes and cleanups
5975 - configuration tools
5978 - 64-bit compile warnings
5979 - 64-bit internal statfs data
5981 * test_brw on persistent OST devices
5983 * lprocfs (disabled)
5985 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5987 * documentation updates
5988 * bug fixes and cleanups
5989 * configuration tools
5991 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5994 * many small fixes to 0_5_4
5995 * io/network handling
5996 * thinkos in MDS operations
5998 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6001 * crucial basic fixes to 0.5.3
6002 * IOR, Iozone work over Elan
6005 2002-08-07 Phil Schwan <phil@clusterfs.com>
6006 * version 0_5_3, our first alpha
6007 * we use the new Portals iovs
6008 * documentation updates
6009 * bug fixes and cleanups
6010 * small changes in the DLM wire protocol
6012 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6013 * version 0_5_1 with some initial stability,
6014 * locking on MD and file I/O.
6015 * documentation updates
6016 * several bug fixes since 0.5.0
6017 * small changes in wire protocol
6019 2002-07-18 Phil Schwan <phil@clusterfs.com>
6021 * delivered as Lustre Light Alpha
6022 * fixed a crash after handling invalid MDS requests
6023 * fixed directory pages for architectures with non-4k pages sizes
6025 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6026 * release version v0_4_4
6027 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6029 * Fixes a number of interruption problems with OST operations.
6030 * Update documentation for portals header changes
6031 * Move all wire protocol structs/defines to lustre_idl.h
6032 * Fixes symlink length bug.
6033 * Add tcpdump to repository.
6035 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6036 * release version v0_4_3
6037 * Fixes statfs for inodes on extN.
6038 * Fixes bug in runtests which would delete /etc/hosts.
6039 * Use 64-bit object IDs wherever possible (not into VFS though)
6040 Remove ost_get_info, which is unused by lustre, and out of date.
6042 2002-07-03 Peter Braam <braam@clusterfs.com>
6043 * release version v0_4_2 Fixes a lookup error (type not passed)
6044 * move forward to head of Portals
6045 * move forward to latest Lustre kernel
6047 2002-06-25 Peter Braam <braam@clusterfs.com>
6048 * release version v0_4_1. Hopefully stable on single node use.