1 tbd Sun Microsystems, Inc.
4 2.6.18-164.6.1.0.1.el5 (RHEL 5, OEL 5)
5 * Client support for unpatched kernels:
6 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
7 2.6.16 - 2.6.27 vanilla (kernel.org)
8 2.6.16.60-0.39.3 (SLES 10),
9 2.6.18-164.6.1.el5 (RHEL 5),
11 2.6.27 vanilla (kernel.org).
12 * Recommended e2fsprogs version: 1.41.6.sun1
13 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
14 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
15 removed cwd "./" (refer to Bugzilla 14399).
19 Description: Allow non-root access for "lfs check".
20 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
24 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
26 Severity : enhancement
28 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
29 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
33 Description: Sanity test_27p fails.
34 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
35 Sanity tests 27 were cleaned up and improved.
39 Description: Unify req format on client/servers
40 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
41 buffers and swabbers approach..
46 Description: drop join file stuff.
48 Severity : enhancement
50 Description: Add 2.6.27 and SLES11 (patchless client) support
54 Description: Fix in ptlrpc_expire_one_request() to print the signed time
57 Severity : enhancement
59 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
62 Severity : enhancement
64 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
66 Severity : enhancement
68 Description: correctly handle big reply message.
69 Details : send LNet event if reply is bigger then buffer and adjust this buffer
74 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
76 Severity : enhancement
78 Description: Add support for OEL5.
82 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
86 Description: client could not reconnect to OST because of an active request.
87 Details : The client now retries to reconnect to the same server, if a
88 connect request failed with EBUSY or -EAGAIN. Test case 35b
89 conf-sanity.sh has been created.
93 Description: Don't allow make backward step on assignin osc next id.
94 Details : race between allocation next id and ll_sync thread can be cause
95 of set wrong osc next id and can be kill valid ost objects.
99 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
103 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
107 Description: Error handling in osc_statfs_interpret() has been improved.
108 Details : Check in osc_statfs_interpret() for EBADR.
112 Description: Do not send statfs() requests to OSTs disabled by administrator.
113 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
115 Severity : enhancement
117 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
119 Severity : enhancement
121 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
123 Severity : enhancement
125 Description: Enhancement for lfs(1) command to use numeric uid/gid.
129 Description: Improved cleanup procedure for conf-sanity test_46a.
133 Description: Do not start bulk transfer, if client tried to reconnect.
134 Details : Do not start bulk transfer, if client tried to reconnect,
135 but failed due to active rpcs. Check for exp_abort_active_req
136 has been implemented.
140 Description: Prevent inconsistences between linux and lustre mount structures.
141 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
142 Make the sleep interruptible.
144 Severity : enhancement
146 Description: provide server to client comms path
147 Details : server to client communications path via new LDLM_SET_INFO rpc
151 Description: truncate starts GFP_FS allocation under transaction causing deadlock
152 Details : ldiskfs_truncate calls grab_cache_page which may start page
153 allocation under an open transaction. This may lead to
154 calling prune_icache with consequent lustre reentrance.
158 Description: send correctly lsm on open replay
159 Details : MDS is trust to LSM size on replay open, but client can set wrong size
164 Description: enable client interface failover
165 Details : When a child reconnects from another NID, properly update export
166 nid hash position and ldlm reverse import.
168 Severity : enhancement
170 Description: Communicate OST degraded/readonly state via statfs to MDS
171 Details : Flags in the statfs returned from OSTs indicate whether the
172 OST is in a degraded RAID state, or if the filesystem has
173 turned read-only after a filesystem error is detected.
178 Description: don't panic if EPROTO was hit when reading symlink
179 Details : correctly handling request reference in error cases.
181 Severity : enhancement
183 Description: Add LustreNetLink, a kernel-userspace communcation path.
185 Severity : enhancement
187 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
192 Description: access to llog context before init.
193 Details : move handling CATALOGS file at osc layer and forbid access to llog
197 Frequency : rare, only when client is writing to OST on local node
199 Description: Avoid deadlock for local client writes
200 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
201 memory freeing context. This allows OST threads to set the
202 PF_MEMALLOC flag on task structures in order to allocate memory
203 from reserved pools and complete IO.
204 Use GFP_HIGHUSER for OST allocations for non-local client writes,
205 so that the OST threads generate memory pressure and allow
206 inactive pages to be reclaimed.
208 Severity : enhancement
210 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
215 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
216 Details : move ldlm namespace creation in setup phase to avoid grab
217 _lprocfs_lock with cli_sem held.
221 Description: Temporarily disable grant shrink.
222 Details : Disable the feature for debugging.
225 Frequency : only when formatting test filesystems
227 Description: Unable to run several mkfs.lustre on loop devices at the same time
228 Details : mkfs.lustre returns error 256 on the concurrent loop devices
229 formatting. The solution is to proper handle the error.
231 Severity : enhancement
233 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
235 Severity : enhancement
237 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
242 Description: resolve race between obd_disconnect and class_disconnect_exports
243 Details : if obd_disconnect will be called to already disconnected export he
244 forget release one reference and osc module can't unloaded.
246 Severity : enhancement
248 Description: Allow tuning service thread via /proc
249 Details : For each service a new
250 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
251 created that can be used to set min/max thread counts, and get the
252 current number of running threads.
257 Descriptoin: don't return error if have partially created objects for file.
258 Details : lov_update_create_set uses set->set_success as index for created
259 objects, so if some requests will be failed, they will have hole
260 at end of array and qos_shrink_lsm for allocate correct lsm.
262 Severity : enhancement
264 Description: Update OFED support to 1.4.1
269 Description: Reduce small size read RPC
270 Details : Set read-ahead limit for every file and only do read-ahead when
271 available read-ahead pages are bigger than 1M to avoid small size
274 Severity : enhancement
276 Description: move AT tunable parameters for more consistent usage
277 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
279 Severity : enhancement
281 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
282 Details : allow skip disconnected ost for send statfs request and hide error
288 Description: fix racy locking of mballoc block bitmaps causing BUG
289 Details : The locking of the mballoc buddy bitmap and the in-memory
290 block bitmap was using two different spin locks in some
291 cases. This made it possible to incorrectly access the
292 mballoc bitmap while another process was modifying it,
293 causing a sanity assertion to fail. While no on-disk corruption
294 was reported, there was some risk of this happening.
297 Frequency : rare, on llog test 6
299 Descriptoin: don't allow connect to already connected import
300 Details : allowing connect to already connected import is hide connecting problem.
305 Descriptoin: allow kill process which wait statahead result
306 Details : for some reasons 'ls' can stick in waiting result from statahead,
307 in this case need way for kill this process.
309 Severity : enhancement
311 Description: Add state history info file, enhance import info file
312 Details : Track import connection state changes in a new osc/mdc proc file;
313 add overview-type data to the osc/mdc import proc file.
315 Severity : enhancement
317 Description: MDS create should not wait for statfs RPC while holding DLM lock.
319 Severity : enhancement
321 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
326 Descriptoin: OST grant too much space to client even there are not enough space.
327 Details : Client will shrink its grant cache to OST if there are no write
328 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
329 this grant cache if there are already not enough avaible space
330 (left_space < total_clients * 32M).
333 Frequency : start MDS on uncleanly shutdowned MDS device
335 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
336 Details : stay in waiting mds<>ost recovery finished produce random bugs
337 due race between two ll_sync thread for one lov target. send
338 ACTIVATE event only if connect realy finished and import have
342 Frequency : rare, connect and disconnect target at same time
344 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
345 Details : don't call obd_disconnect under lov_lock. this long time
346 operation and can block ptlrpcd which answer to connect request.
351 Descriptoin: don't lose wakeup for imp_recovery_waitq
352 Details : recover_import_no_retry or invalidate_import and import_close can
353 both sleep on imp_recovery_waitq, but we was send only one wakeup
357 Frequency : always with long access acl
359 Descriptoin: mds can't pack reply with long acl.
360 Details : mds don't control size of acl but they limited by reint/getattr
363 Severity : enhancement
365 Description: Update to SLES10 kernel-2.6.16.60-0.33.
367 Severity : enhancement
369 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
372 Frequency : start MDS on uncleanly shutdowned MDS device
374 Descriptoin: aborting recovery hang on MDS
375 Details : don't throttle destroy RPCs for the MDT.
378 Frequency : on remount
380 Description: external journal device not working after the remount
381 Details : clear dev_rdonly flag for external journal devices in
387 Description: shutdown vs evict race
388 Details : client_disconnect_export vs connect request race.
389 if client will evicted at this time - we start invalidate
390 thread without referece to import and import can be freed
394 Frequency : rare, need acl's on inode.
396 Description: client can't handle ost additional correctly
397 Details : if ost was added after client connected to mds client can have
398 hit lnet_try_match_md ... to big messages to wide striped files.
399 in this case need teach client to handle config events about add
400 lov target and update client max ea size at that event.
402 Severity : enhancement
404 Description: Changelogs
405 Details : Changelogs are a lightweight mechanism to track filesystem
406 metadata and namespace changes. The changelog is recorded
407 permanently on the MDTs, and is periodically "consumed" / purged
408 when records are no longer needed.
410 Severity : enhancement
412 Description: compact fld format with extents
413 Details : Store range of seq rather than every seq in FLD. Seq
414 controller update FLD rather than clients. In Case of CMD, mdt0
415 has FLD, all other metadata server act as non persistent proxy
416 for FLD queries and cache fld entries in fld cache.
421 Description: don't skip ost target if they assigned to file
422 Details : Drop slow OSCs if we can, but not for requested start idx.
423 This means "if OSC is slow and it is not the requested
424 start OST, then it can be skipped, otherwise skip it only
425 if it is inactive/recovering/out-of-space.
429 Description: more cleanup in mds_lov
430 Details : not send LOV EA under replay, we can't know about they size at this
431 time. Don't allow client connect to mds before any ost connected,
432 for avoid problems with LOV EA size and returning EIO to client.
434 Severity : enhancement
436 Description: Interoperability at server side (Disk interoperability)
438 Severity : enhancement
440 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
442 Severity : enhancement
444 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
446 Severity : enhancement
448 Description: New client IO stack (CLIO).
450 Severity : enhancement
452 Description: Commit on sharing. Eliminate inter-client dependencies between
453 uncommitted transactions by doing transaction commits.
454 Thereby clients may recovery independently.
457 Frequency : Create a symlink file with a very long name
459 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
460 Details : If there is no extra space in the request for early cancels,
461 ldlm_req_handles_avail() returns 0 instead of a negative value.
463 Severity : enhancement
465 Description: Add /proc entry for import status
466 Details : The mdc, osc, and mgc import directories now have
467 an import directory that contains useful import data for debugging
470 Severity : enhancement
472 Description: Re-disable certain /proc logging
473 Details : Enable and disable client's offset_stats, extents_stats and
474 extents_stats_per_process stats logging on the fly.
477 Frequency : Only on FC kernels 2.6.22+
479 Description: oops in statahead
480 Details : Do not drop reference count for the dentry from VFS when lookup,
481 VFS will do that by itself.
483 Severity : enhancement
485 Description: Generic /proc file permissions
486 Details : Set /Proc file permissions in a more generic way to enable non-
487 root users operate on some /proc files.
491 Description: Hitting mdc_commit_close() ASSERTION
492 Details : Properly handle request reference release in
493 ll_release_openhandle().
497 Description: quota recovery deadlock during mds failover
498 Details : This patch includes att18982, att18236, att18237 in bz14840.
500 1. fix osts hang when mds does failover with quotaon
501 2. prevent watchdog storm when osts threads wait for the
506 Frequency : only patchless client
507 Description: add workaround for race between add/remove dentry from hash
509 Severity : enhancement
511 Description: Allow OST glimpses to return PW locks
515 Description: LBUG when llog conf file is full
516 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
520 Description: Prevent import from entering FULL state when server in recovery
524 Description: service mount cannot take device name with ":"
525 Details : Only when device name contains ":/" will mount treat it as
531 Description: replace ptlrpcd with the statahead thread to interpret the async
532 statahead RPC callback
536 Frequency : on recovery
537 Description: I/O failures after umount during fail back
538 Details : if client reconnected to restarted server we need join to recovery
539 instead of find server handler is changed and process self
540 eviction with cancel all locks.
542 Severity : enhancement
544 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
546 Severity : enhancement
548 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
550 Severity : enhancement
552 Description: Upcall on Lustre log has been dumped
553 Details : Allow for a user mode script to be called once a Lustre log has
554 been dumped. It passes the filename of the dumped log to the
555 script, the location of the script can be specified via
556 /proc/sys/lnet/debug_log_upcall.
561 Description: avoid idr_remove called for id which is not allocated.
562 Details : Move assigment s_dev for clustered nfs to end of initialization,
563 to avoid problem with error handling.
568 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
569 Details : When connection is reused this not moved from CONN_UNUSED_HASH
570 into CONN_USED_HASH and this prodice warning when put connection
571 again in unused hash.
574 Severity : enhancement
576 Description: File striping can now be set to use an arbitrary pool of OSTs.
578 Severity : enhancement
580 Description: Export bytes_read/bytes_write count on OSC/OST.
584 Description: Early reply size mismatch, MGC loses connection
585 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
586 the connect flags are properly negotiated.
591 Description: quotas are not honored with O_DIRECT
592 Details : all writes with the flag O_DIRECT will use grants which leads to
593 this problem. Now using OBD_BRW_SYNC to guard this.
597 Description: add quota statistics
598 Details : 1. sort out quota proc entries and proc code.
599 2. add quota statistics
601 Severity : enhancement
603 Description: enable quota support for HEAD.
607 Description: Properly propagate oinfo flags from lov to osc for statfs
608 Details : restore missing copy oi_flags to lov requests.
610 Severity : enhancement
612 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
613 : l_getgroups(8), lst(8), routerstat(8)
614 Details : included man pages for llobdstat(8), llstat(8),
615 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
617 Severity : enhancement
619 Description: configure's --enable-quota should check the
620 : kernel .config for CONFIG_QUOTA
621 Details : configure is terminated if --enable-quota is passed but
622 : no quota support is in kernel
626 Description: Remove portals compatibility
627 Details : Remove portals compatibility, not interoperable with releases
632 Description: Resolve device initialization race
633 Details : Prevent proc handler from accessing devices added to the
634 obd_devs array but yet be intialized.
636 Severity : enhancement
638 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
640 Severity : enhancement
642 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
647 Description: Using wrong pointer in osc_brw_prep_request
648 Details : Access to array[-1] can produce panic if kernel compiled with
649 CONFIG_PAGE_ALLOC enabled
653 Description: Client runs out of low memory
654 Details : Consider only lowmem when counting initial number of llap pages
658 Description: *optional* service tags registration
659 Details : if the "service tags" package is installed on a Lustre node
660 When the filesystem is mounted, a local-node service tag will
661 be created. See http://inventory.sun.com/ for more information
662 about the Service Tags asset management system.
666 Description: Kernel BUG tries to release flock
667 Details : Lustre does not destroy flock lock before last reference goes
668 away. So always drop flock locks when client is evicted and
669 perform unlock regardless of successfulness of speaking to MDS.
673 Description: add refcount protection for osc callbacks, avoid panic on shutdown
677 Description: sanity test 65a fails if stripecount of -1 is set
678 Details : handle -1 striping on filesystem in ll_dirstripe_verify
683 Description: ASSERTION(CheckWriteback(page,cmd)) failed
684 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
687 Severity : enhancement
689 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
693 Description: do not process already freed flock
694 Details : flock can possibly be freed by another thread before it reaches
695 to ldlm_flock_completion_ast.
699 Description: LBUG during stress test
700 Details : Need properly lock accesses the flock deadlock detection list.
704 Description: oops in page fault handler
705 Details : kernel page fault handler can return two special 'pages' in error
706 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
710 Description: timeout with invalidate import.
711 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
712 should be handled by ptlrpcd. This produce long age waiting and
713 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
715 Severity : enhancement
717 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
721 Description: enable MGS and MDT services start separately
722 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
723 a co-located MGS without starting the MGS, which is a complement
724 to 'nosvc' mount option.
728 Frequency : after recovery
729 Description: precreate to many object's after del orphan.
730 Details : del orphan st in oscc last_id == next_id and this triger growing
731 count of precreated objects. Set flag LOW to skip increase count
732 of precreated objects.
736 Frequency : rare, on clear nid stats
737 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
738 Details : when clean nid stats sometimes try destroy live entry,
739 and this produce panic in free.
743 Description: Stack overflow during MDS log replay
744 ease stack pressure by using a thread dealing llog_process.
748 Description: wait until IO finished before start new when do lock cancel.
749 Details : VM protocol want old IO finished before start new, in this case
750 need wait until PG_writeback is cleared until check dirty flag and
751 call writepages in lock cancel callback.
753 Severity : enhancement
755 Description: using special macro for print time and cleanup in includes.
759 Description: mds_mfd_close() ASSERTION(rc == 0)
760 Details : In mds_mfd_close(), we need protect inode's writecount change
761 within its orphan write semaphore to prevent possible races.
765 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
769 Frequency : rare, on shutdown ost
770 Description: don't hit live lock with umount ost.
771 Details : shrink_dcache_parent can be in long loop with destroy dentries,
772 use shrink_dcache_sb instead.
776 Description: don't panic with use echo client
777 Details : echo client pass NULL as client nid pointer and this produce null
782 Description: fix build on ppc32
783 Details : compile code with -m64 flag produce wrong object file for ppc32.
787 Description: add message levels for liblustreapi
791 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
792 Details : Move the 'good_osts' check before the 'total_bavail' check. This
793 will result in an -EAGAIN and in the exit call path we call
794 alloc_rr() which will with increasing aggressiveness attempt to
795 aquire precreated objects on the minimum number of required OSCs.
799 Description: Use old size assignment to avoid deadlock
800 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
801 the scheduling while holding a spinlock. We do not need locking
802 for size in ll_update_inode() because size is only updated from
803 the MDS for directories or files without objects, so there is no
804 other place to do the update, and concurrent access to such inodes
805 are protected by the inode lock.
809 Description: resolve "_IOWR redefined" build error on SLES10
813 Description: dump the memory debugging after all modules are unloaded to
814 suppress false negative in conf_sanity test 39
816 Severity : enhancement
818 Description: build kernel-ib packages for OFED 1.3 in our release cycle
823 Description: fix SLES kernel versioning
824 Details : the kernel version for our SLES 10 kernel did not include a "-"
825 before the "smp" at the end. while this was not a problem in
826 general, it did mean that software trying to use the kernel
827 version to try to detect a vendor specific kernel would fail.
828 this was most evident by the OFED build scripts.
832 Description: Don't update lov_desc members until making sure they are valid
833 Details : When updating lov_desc members via proc fs, need fix their
834 validities before doing the real update.
838 Description: don't put request into delay list while invalidate in flight.
839 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
840 import in flight. this produce timeout for invalidate and sometimes
841 can cause stale data.
845 Frequency : on ppc only
846 Description: not convert ost objects for directory because it's not exist.
847 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
851 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
853 Severity : enhancement
855 Description: Update to support 2.6.22.14 vanilla kernel.
859 Frequency : rare, on recovery
860 Description: read procfs can produce deadlock in some situation
861 Details : Holding lprocfs lock which send rpc can produce block for destroy
862 obd objects and this also block reconnect with -EALREADY. This
863 isn't fix all lprocfs bugs - but make it rare.
865 Severity : enhancement
867 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
870 Frequency : frequent on X2 node
872 Description: mdc_set_open_replay_data LBUG
873 Details : Set replay data for requests that are eligible for replay.
877 Description: lustre_mgs: operation 101 on unconnected MGS
878 Details : When MGC is disconnected from MGS long enough, MGS will evict the
879 MGC, and late on MGC cannot successfully connect to MGS and a lot
880 of the error messages complaining that MGS is not connected.
884 Frequency : on network error
885 Description: panic with double free request if network error
886 Details : mdc_finish_enqueue is finish request if any network error ocuring,
887 but it's true only for synchronus enqueue, for async enqueue
888 (via ptlrpcd) this incorrect and ptlrpcd want finish request
891 Severity : enhancement
893 Description: client-side metadata stat-ahead during readdir(directory readahead)
894 Details : perform client-side metadata stat-ahead when the client detects
895 readdir and sequential stat of dir entries therein
898 Frequency : on start mds
900 Description: Implement get_info(last_id) in obdfilter.
903 Frequency : occasional
905 Description: Correctly check stale fid, not start epoch if ost not support SOM
906 Details : open with flag O_CREATE need set old fid in op_fid3 because
907 op_fid2 was overwritten with new generated fid, but mds can answer
908 with one of these two fids and both is not stale. Setattr
909 incorrectly started an epoch and assume will be called
910 done_writing, but without SOM done_writing ever being called.
913 Frequency : rare, depends on device drivers and load
915 Description: MDS or OSS nodes crash due to stack overflow
916 Details : Code changes in 1.8.0 increased the stack usage of some functions.
917 In some cases, in conjunction with device drivers that use a lot
918 of stack the MDS (or possibly OSS) service threads could overflow
919 the stack. One change which was identified to consume additional
920 stack has been reworked to avoid the extra stack usage.
923 Frequency : occasional
925 Description: Do not fail import if osc_interpret_create gets -EAGAIN
926 Details : If osc_interpret_create got -EAGAIN it immediately exits and
927 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
928 oscc_has_objects and see OSC has no objests and call
929 oscc_internal_create to resend create request.
931 Severity : enhancement
933 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
935 Severity : enhancement
937 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
940 Frequency : very rare
942 Description: Fix replay if there is an un-replied request and open
943 Details : In some cases, older replay request will revert the
944 mcd->mcd_last_xid on MDS which is used to record the client's
947 Severity : enhancement
949 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
951 Severity : enhancement
953 Description: Add rhel5 support to HEAD.
955 Serverity : enhancement
957 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
962 Description: Don't allow skipping OSTs if index has been specified.
963 Details : Don't allow skipping OSTs if index has been specified, make
964 locking in internal create lots better.
968 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
969 Details : access to bitfield in structure is always rounded to long
970 and this produce problem with not atomic change any bit.
974 Description: Lustre make rpms failed.
975 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
976 builds Lustre rpms from tarball.
979 Frequency : rare on shutdown ost
981 Description: If llog cancel was not send before clean_exports phase, this can
982 produce deadlock in llog code.
983 Details : If llog thread has last reference to obd and call class_import_put
984 this produce deadlock because llog_cleanup_commit_master wait when
985 last llog_commit_thread exited, but this never success because was
986 called from llog_commit_thread.
990 Description: allow userland application know is lost one of stripes.
991 Details : fill lvb_blocks with error code on ost and return it to
992 application if error flag found.
996 Description: NULL lov_tgts causing MDS oops
997 Details : more safe checks for NULL lov_tgts for avoid oops.
999 Severity : enhancement
1001 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1005 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1006 Details : If an OST is down the MDS will hang indefinitely in
1007 obd_statfs_rqset() waiting for the statfs data. While for
1008 MDS QOS usage of statfs, it should not stuck in waiting.
1010 Severity : enhancement
1012 Description: remote_acl support
1013 Details : Support ACL-based permission check for remote user.
1014 Support setfacl/getfacl for remote user with the utils
1015 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1016 the system "{s,g}etfacl" utils.
1018 Severity : enhancement
1020 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1022 Severity : enhancement
1024 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1028 Description: make mgs_setparam() handle fsname containing dash
1029 Details : fsname containing a dash does not work with lctl conf_param
1031 Severity : enhancement
1033 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1035 Severity : enhancement
1037 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1041 Frequency : rare, at shutdown
1042 Description: access already free / zero obd_namespace.
1043 Details : if client_disconnect_export was called without force flag set,
1044 and exist connect request in flight, this can produce access to
1045 NULL pointer (or already free pointer) when connect_interpret
1046 store ocd flags in obd_namespace.
1050 Frequency : only at startup
1051 Description: not alloc memory with spinlock held.
1052 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1053 if any spinlock held.
1055 Severity : enhancement
1057 Description: make lustre randomly fail allocating memory
1058 Details : Make lustre randomly failed allocating memory for testing purpose.
1060 Severity : enhancement
1062 Description: lost problems with lov objid file
1063 Details : Fixes some scability and access to not inited memory problems
1064 in work with lov objdid file.
1069 Description: lfs find does not continue on file error
1070 Details : Continue other files processing when a file/dir is absent.
1074 Description: Inconsistent usage of lustre_pack_reply()
1075 Details : Standardize the usage of lustre_pack_reply() such that it
1076 always generate a CERROR on failure.
1079 Frequency : occasional
1081 Description: MDS hang or stay in waiting lock
1082 Details : If client receive lock with CBPENDING flag ldlm need send lock
1083 cancel as separate rpc, to avoid situation when cancel request
1084 can't processed due all i/o threads stay in wait lock.
1088 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1092 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1096 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1100 Description: Update to RHEL4 latest kernel.
1102 Severity : enhancement
1104 Description: Build SLES10 patchless client fails
1105 Details : The configure was broken by run ./configure with
1106 --with-linux-obj=.... argument for patchless client. When the
1107 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1108 can't search header adequately. Use absolute path such as
1109 -I($LINUX)/include instead.
1113 Description: interrupt oig_wait produce painc on resend.
1114 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1115 private set, and this produce situation when rq_ptlrpcd_data not
1116 copyed to new allocated request and triggered LBUG on assert
1117 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1118 ptlrpcd set if request is changed and can be safety changed to use
1123 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1124 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1125 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1126 these macros and replace them with their difinitions instead.
1130 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1131 Details : Update which_patch & target file for SLES10 latest kernel.
1134 Description: Remove client patches from SLES 10 kernel.
1135 Details : This causes SLES 10 clients to behave as patchless clients
1136 even on a Lustre-patched (server) kernel.
1138 Severity : enhancement
1140 Description: self-adjustable client's lru lists
1141 Details : use adaptive algorithm for managing client cached locks lru
1142 lists according to current server load, other client's work
1143 pattern, memory activities, etc. Both, server and client
1144 side namespaces provide number of proc tunables for controlling
1147 Severity : enhancement
1149 Description: light-weight GSS support
1150 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1151 and reduce performance overhead.
1153 Severity : enhancement
1155 Description: Linux keyring support
1156 Details : Support using service of Linux keyring for Lustre GSS internal
1157 context refresh/cache mechanism.
1161 Description: Fix errors in lfs documentation
1162 Details : Fixes man pages
1166 Description: don't use GFP_* in generic Lustre code.
1167 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1171 Description: obdclass is limited by single OBD_ALLOC(idarray)
1172 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1176 Description: warnings with build patchless client with vanila 2.6.19 and up
1177 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1178 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1182 Frequency : Only for SLES
1183 Description: sanity_quota fail test_1
1184 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1185 file, which makes TSTID[2] inunique.
1189 Frequency : Only for SLES9
1190 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1191 Details : Update target/ChangeLog/which_patch .
1195 Frequency : Only for test_10 in sanity-quota.sh
1196 Description: a bug in quota test code
1197 Details : checking a test flag in a wrong place
1201 Frequency : Only for test_13 in sanity-quota.sh
1202 Description: a bug in quota test script
1203 Details : shouldn't check overfull allocated quota
1207 Description: jbd statistics
1208 Details : Port older jbd statistics patch for sles10
1210 Severity : enhancement
1212 Description: Add printing inode info into message about error in writepage.
1217 Description: reply_lock_interpret crash due to race with it and lock cancel.
1218 Details : Do not replay locks that are being cancelled. Do not reference
1219 locks by their address during replay, just by their handle.
1223 Frequency : When flocks are used.
1224 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1225 Details : Flock locks might destroy just granted lock if it could be merged
1226 with another existing flock, this is done in completion handler,
1227 so teach ldlm_cli_enquque_fini that this is a valid case for
1233 Description: Oops in read and write path when failing to allocate lock.
1234 Details : Check if lock allocation failed and return error back.
1238 Description: lstripe command fails for valid OST index
1239 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1240 instead of lov->desc.ld_active_tgt_count.
1244 Description: sanity.sh failed test 103
1245 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1250 Description: df doesn't work properly if diskfs blocksize != 4K
1251 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1255 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1256 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1258 Severity : enhancement
1260 Description: Adaptive timeouts
1261 Details : RPC timeouts adapt to changing server load and network
1262 conditions to reduce resend attempts and improve recovery time.
1266 Description: llapi_file_create() does not allow some changes
1267 Details : add llapi_file_open() that allows specifying the mode and
1268 open flags, and also returns an open file handle.
1272 Description: merge and cleanup kernel patches.
1273 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1277 Description: Add journal checksum support.(Kernel part)
1278 Details : The journal checksum feature adds two new flags i.e
1279 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1280 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1281 indicates that the commit block contains the checksum for
1282 the blocks described by the descriptor blocks. Now commit
1283 record can be sent to disk without waiting for descriptor
1284 blocks to be written to disk. This behavior is controlled
1285 using JBD2_FEATURE_ASYNC_COMMIT flag.
1289 Description: OSS needs mutliple precreate threads
1290 Details : Add ability to start more than one create thread per OSS.
1294 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1295 Details : execute lfs setstripe on client
1299 Description: mds_obd_create error creating tmp object
1300 Details : When the user sets quota on root, llog will be affected and can't
1301 create files and write files.
1304 Frequency : Always on ia64 patchless client, and possibly others.
1306 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1307 Details : This allows the patchless client to be loaded on architectures
1308 without this export.
1312 Description: RedHat Update kernel for RHEL5
1313 Details : Add the kernel config file for RHEL5.
1317 Description: RedHat Update kernel for RHEL5
1318 Details : Modify the kernel config file more closer RHEL5.
1322 Description: Build failure against Centos5 (RHEL5)
1323 Details : Define PAGE_SIZE when it isn't present.
1327 Description: test_7 on sanity-quota.sh can't be used on separate machines
1328 Details : add facet to handle it
1332 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1333 Details : with NFS, the anon dentry's parent was set to itself in
1334 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1335 corresponding dentry other than use rec->ur_name.
1337 Severity : enhancement
1339 Description: lfs setstripe enhancement
1340 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1344 Description: enable data checksumming by default
1345 Details : enable checksum by default, allow --disable-checksum
1346 configure option and "-o nochecksum" mount option. Checksums
1347 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1351 Description: lustre support for RHEL5
1352 Details : Add support for RHEL5.
1356 Description: Client eviction due to failover config
1357 Details : after a connection loss, the lustre client should attempt to
1358 reconnect to the last active server first before trying the
1359 other potential connections.
1363 Description: when mds and osts use different quota unit(32bit and 64bit),
1364 quota will be released repeatly.
1365 Details : void sending multiple quota reqs to mds, which will keep the
1366 status between the reqs.
1370 Description: osts not allocated evenly to files
1371 Details : change the condition to increase offset_idx
1374 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1376 Description: Data corruption for OSTs that are formatted larger than 2TB
1378 Details : When generating the bio request for lustre file writes the
1379 sector number would overflow a temporary variable before being
1380 used for the IO. The data reads correctly from Lustre (which
1381 will overflow in a similar manner) but other file data or
1382 filesystem metadata may be corrupted in some cases.
1386 Description: Tune the kernel for good SCSI performance.
1387 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1388 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1393 Description: rewrite ext2-derived code in obdclass/uuid.c
1394 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1395 from scratch preserving functionality.
1399 Description: "lfs find -obd UUID" prints directories
1400 Details : "lfs find -obd UUID" will return all directory names instead
1401 of just file names. It is incorrect because the directories
1402 do not reside on the OSTs.
1406 Description: use i_size_read and i_size_write in 2.6 port
1407 Details : replace inode->i_size access with i_size_read/write()
1410 Frequency : when removing large files
1412 Description: scheduling issue during removal of large Lustre files
1413 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1414 It causes scheduling issues when removing large files (17TB in the
1418 Frequency : only with liblustre clients on XT3
1420 Description: evictions taking too long
1421 Details : allow llrd to evict clients directly on OSTs
1424 Frequency : only on ppc
1426 Description: /proc/fs/lustre/devices broken on ppc
1427 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1428 We should make sure the type of 'index' is loff_t and then cast
1429 explicitly as needed below. Do not assign an explicitly cast
1433 Frequency : only for rhel5
1435 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1436 Details : Modify the target file & which_kernel.
1438 Severity : enhancement
1440 Description: omit set fsid for export NFS
1441 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1442 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1446 Frequency : Always, for kernels after 2.6.16
1447 Description: Fix warning idr_remove called for id=.. which is not allocated.
1448 Details : Last kernels save old s_dev before kill super and not allow
1449 to restore from callback - restore it before call kill_anon_super.
1453 Description: buffer overruns could theoretically occur
1454 Details : llapi_semantic_traverse() modifies the "path" argument by
1455 appending values to the end of the origin string, and a
1456 overrun may occur. Adding buffer overrun check in liblustreapi.
1460 Description: Fix error on 'ls .' at the top of the Lustre mount.
1461 Details : Don't revalidate dentry if it is a root dentry.
1465 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1466 Details : Modify vm-tunables-rhel4.patch.
1470 Description: Kernel config for 2.6.18-vanilla.
1471 Details : Modify targets/2.6-vanilla.target.in.
1472 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1473 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1474 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1475 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1479 Description: improve handling recoverable errors
1480 Details : if request processig with error which can be recoverable on server
1481 request should be resend, otherwise page released from cache and
1486 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1487 Details : Modify target file & which_patch.
1488 A flaw was found in the IA32 system call emulation provided
1489 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1490 value could be stored in the %RAX register, which could trigger an
1491 out-of-bounds system call table access. An untrusted local user
1492 could exploit this flaw to run code in the kernel
1493 (ie a root privilege escalation). (CVE-2007-4573).
1497 Description: O_DIRECT bypasses client statistics.
1498 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1499 read_ahead_stats were updated. copied the stats section from
1500 osc_send_oap_rpc() into async_internal().
1504 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1508 Description: change order of libsysio includes
1509 Details : '#include sysio.h' should always come before '#include xtio.h'
1513 Description: handle "serious error: objid * already exists" more gracefully
1514 Details : If LAST_ID value on disk is smaller than the objects existing in
1515 the O/0/d* directories, it indicates disk corruption and causes an
1516 LBUG(). If the object is 0-length, then we should use the existing
1517 object. This will help to avoid a full fsck in most cases.
1519 Severity : enhancement
1521 Description: adapt the lustre_config script to support the upgrade case
1522 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1523 server targets to 1.6 in parallel.
1527 Description: To avoid grant space > avaible space when the disk is almost
1528 full. Without this patch you might see the error "grant XXXX >
1529 available" or some LBUG about grant, when the disk is almost
1531 Details : In filter_check_grant, for non_grant cache write, we should
1532 check the left space by if (*left > ungranted + bytes), instead
1533 of (*left > ungranted), because only we are sure the left space
1534 is enough for another "bytes", then the ungrant space should be
1535 increase. In client, we should update cl_avail_grant only there
1536 is OBD_MD_FLGRANT in the reply.
1540 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1541 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1542 described in CVE-2007-4573. This problem could allow untrusted
1543 local users to gain root access.
1546 Frequency : when using O_DIRECT and quotas
1548 Description: Incorrect file ownership on O_DIRECT output files
1549 Details : block usage reported by 'lfs quota' does not take into account
1550 files that have been written with O_DIRECT.
1555 Description: touch file failed when fs is not full
1556 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1557 otherwise we can get ENOSP while fs is not full.
1561 Description: parallel lock callbacks
1562 Details : Instead of sending blocking and completion callbacks as separated
1563 requests, adding them to a set and sending in parallel.
1566 Frequency : only for Cray XT3
1567 Bugzilla : 12829/13455
1568 Description: Changing primary group doesn't change the group lustre assigns to
1570 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1571 overridden with the primary group provided by the group upcall,
1572 whereas the supplied fsgid can be trusted if it is in the list of
1573 supplementary groups returned by the group upcall.
1575 Severity : enhancement
1577 Description: Allow masking D_WARNING, D_ERROR messages from console
1578 Details : Console messages can now be disabled via lnet.printk.
1582 Description: User code with malformed file open parameter crashes client node
1583 Details : Before packing join_file req, all the related reference should be
1584 checked carefully in case some malformed flags cause fake
1585 join_file req on client.
1589 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1590 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1591 mode is equal to the granted mode. However, it should also check
1592 that LDLM_FL_CP_REQD is not set.
1596 Description: Heavy nfs access might result in deadlocks
1597 Details : After ELC code landed, it is now improper to enqueue any mds
1598 locks under och_sem, because enqueue might want to decide to
1599 cancel open locks for same inode we are holding och_sem for.
1603 Description: Client eviction while running blogbench
1604 Details : A lot of unlink operations with concurrent I/O can lead to a
1605 deadlock causing evictions. To address the problem, the number of
1606 oustanding OST_DESTROY requests is now throttled to
1607 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1608 callbacks are processed in priority.
1612 Description: enable ACLs on MDS by default
1613 Details : ACLs must be enabled on MDS by default.
1616 Frequency : PPC/PPC64 only
1618 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1619 Details : fix duplicated definitions between asm-ppc64/types.h and
1620 lustre_types.h on PPC.
1623 Frequency : PPC/PPC64 only
1625 Description: asm-ppc/segment.h does not exist
1626 Details : fix compile issue on PPC.
1630 Description: better handle error messages in extents code
1633 Frequency : RHEL4 only
1635 Description: mkfs is very slow on IA64/RHEL4
1636 Details : A performance regression has been discovered in the MPT Fusion
1637 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1638 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1639 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1641 Severity : enhancement
1643 Description: SNMP support enhancement
1644 Details : Adding total number of sampled request for an MDS node in snmp
1647 Severity : enhancement
1649 Description: Optimize ldlm waiting list processing for PR extent locks
1650 Details : When processing waiting list for read extent lock and meeting read
1651 lock that is same or wider to it that is not contended, skip
1652 processing rest of the list and immediatelly return current
1653 status of conflictness, since we are guaranteed there are no
1654 conflicting locks in the rest of the list.
1658 Description: Time out and refuse to reconnect
1659 Details : When the failover node is the primary node, it is possible
1660 to have two identical connections in imp_conn_list. We must
1661 compare not conn's pointers but NIDs, otherwise we can defeat
1662 connection throttling.
1666 Description: port llog fixes from b1_6 into HEAD
1667 Details : Port llog reference couting and some llog cleanups from b1_6
1668 (bug 10800) into HEAD, for protect from panic and access to already
1669 free llog structures.
1673 Description: Detect stride IO mode in read-ahead
1674 Details : When a client does stride read, read-ahead should detect that and
1675 read-ahead pages according to the detected stride pattern.
1679 Description: data checksumming impacts single node performance
1680 Details : add support for several checksum algorithm. Currently, only CRC32
1681 and Adler-32 are supported. The checksum type can be changed on
1682 the fly via /proc/fs/lustre/osc/*/checksum_type.
1686 Description: use adler32 for page checksums
1687 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1692 Description: build for x2 fails
1693 Details : fix compile issue on Cray systems.
1697 Description: Properly match for duplicate locks
1698 Details : Due to different lock order from skiplists code, we need to
1699 traverse entire list for now
1702 Frequency : only on PPC/SLES10
1704 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1705 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1706 be defined. Add a hack in mkfs_lustre.c to work around this compile
1711 Description: LASSERT on MDS when client holding flock lock dies
1712 Details : ldlm pool logic depends on number of granted locks equal to
1713 number of released locks which is not true for flock locks, so
1714 just exclude such locks from consideration.
1718 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1719 Details : Use fsfilt_sync() for both the whole filesystem sync and
1720 individual file sync to eliminate dangerous inode locking
1721 with I_LOCK that can lead to a deadlock.
1725 Description: performance in 1.6.3
1726 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1727 sure that 1MB requests are not fragmented by the block layer.
1729 Severity : enhancement
1731 Description: organize the server-side client stats on per-nid basis
1732 Details : Change the structure of stats under obdfilter and mds to
1740 The "uuid"s file would list the uuids of _active_ exports.
1741 And the clear entry is to clear all stats and stale nids.
1743 Severity : enhancement
1745 Description: eliminate client locks in face of contention
1746 Details : file contention detection and lockless i/o implementation
1747 for contended files.
1751 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1754 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1756 Description: MDS may lose file striping (and hence file data) in some cases
1757 Details : If there are additional extended attributes stored on the MDS,
1758 in particular ACLs, SELinux, or user attributes (if user_xattr
1759 is specified for the client mount options) then there is a risk
1760 of attribute loss. Additionally, the Lustre file striping
1761 needs to be larger than default (e.g. striped over all OSTs),
1762 and an additional attribute must be stored initially in the
1763 inode and then increase in size enough to be moved to the
1764 external attribute block (e.g. ACL growing in size) for file
1769 Description: skiplist implementation simplification
1770 Details : skiplists are used to group compatible locks on granted list
1771 that was implemented as tracking first and last lock of each
1772 lock group the patch changes that to using doubly linked lists
1776 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1777 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1778 because LASSERT on both the data supplied by a client, and the
1779 data on disk is dangerous and incorrect.
1781 Severity : enhancement
1783 Description: Slow truncate/writes to huge files at high offsets.
1784 Details : Directly associate cached pages to lock that protect those pages,
1785 this allows us to quickly find what pages to write and remove
1786 once lock callback is received.
1790 Description: more ldlm soft lockups
1791 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1792 starve other threads from the resource lock for a long time in
1793 case of long waiting queue, so change the debug level from
1794 D_OTHER to the less frequently used D_INFO.
1796 Severity : enhancement
1798 Description: add -gid, -group, -uid, -user options to lfs find
1802 Description: Hung threads in invalidate_inode_pages2_range
1803 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1804 once one is completed. As a result, some RPCs are stuck in the
1805 queue and are never sent.
1809 Description: filter threads hungs on waiting journal commit
1810 Details : Cleanup filter group llog code, then only filter group llog will
1811 be only created in the MDS/OST syncing process.
1815 Description: Procfs and llog threads access destoryed import sometimes.
1816 Details : Sync the import destoryed process with procfs and llog threads by
1817 the import refcount and semaphore.
1819 Severity : enhancement
1821 Description: openlock cache of b1_6 port to HEAD
1826 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1827 Details : If insertion of an extent fails, then discard the inode
1828 preallocation and free data blocks else it can lead to duplicate
1833 Description: don't always update ctime in ext3_xattr_set_handle()
1834 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1835 In some cases the ctime should not be updated, for example for
1836 2.0->1.8 compatibility it is necessary to delete an xattr and it
1837 should not update the ctime.
1841 Bugzilla : 15713/16362
1842 Description: Assertion in iopen_connect_dentry in 1.6.3
1843 Details : looking up an inode via iopen with the wrong generation number can
1844 populate the dcache with a disconneced dentry while the inode
1845 number is in the process of being reallocated. This causes an
1846 assertion failure in iopen since the inode's dentry list contains
1847 both a connected and disconnected dentry.
1851 Description: assertion failure in ldlm_handle2lock()
1852 Details : fix a race between class_handle_unhash() and class_handle2object()
1853 introduced in lustre 1.6.5 by bug 13622.
1858 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1859 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1864 Description: oops in sd_iostats_seq_show()
1865 Details : unloading/reloading the scsi low level driver triggers a kernel
1866 bug when trying to access the sd iostat file.
1871 Description: Kernel panics during QLogic driver reload
1872 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1873 patch, causing memory corruption.
1878 Description: journal_dev option does not work in b1_6
1879 Details : pass mount option during pre-mount.
1881 Severity : enhancement
1883 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1884 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1885 extent information of a file. It can be used to map logical blocks
1886 in a file to physical blocks in the block device.
1890 Description: LDLM soft lockups - improvement
1891 Details : It is be possible to send the lock handle along with each read
1892 or write request because the client is already doing a lock match
1893 itself so there isn't any reason the OST should have to re-do that
1897 Frequency : only on Cray X2
1899 Description: X2 build failures
1900 Details : fix build failures on Cray X2.
1904 Description: xid & resent requests
1905 Details : Initialize RPC XID from clock at startup (randomly if clock is
1908 Severity : enhancement
1910 Description: Add lustre_start utility to start or stop multiple Lustre servers
1915 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1916 Details : In case of memory pressure, list_del() can be called twice on
1917 req->rq_history_list, causing a kernel oops.
1921 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1922 Details : fix stack overflow in the distributed lock manager by defering
1923 export eviction after a failed AST to the elt thread instead of
1924 handling it in the dlm interpret routine.
1928 Description: Convert some comments to new format.
1929 Details : Update documenting comments to match doxygen conventions.
1933 Description: Grammar fixes.
1934 Details : A couple of trivial spelling fixes.
1938 Description: OSD_COUNTERS-mandatory
1939 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1940 to simplify the code.
1944 Description: simplify cmm_device freeing logic.
1945 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1949 Description: Add lockdep support to dt_object_operations locking interface.
1950 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1951 parameter indicating lock ordering. Update mdd code to use new
1956 Description: Introduce failloc constants for lockless IO tests.
1957 Details : Add two new failloc constants to test lockless IO. Only one of
1958 them in implemented---another is checked in yet to be landed
1963 Description: Add lockdep support for inode mutex.
1964 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1965 in the situations where Lustre has to lock more than one inode
1970 Description: Add optional invariants checking support.
1971 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1972 configure switch. This macro is to be used for consistency and
1973 sanity checks that are too expensive to be left in `production'
1978 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1979 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1984 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1985 Details : Use spin_lock_nested() in (the only) situation where more than
1986 one ldlm_lock is locked simultaneously. Also, fix possible
1987 dead-lock in ldlm_lock_change_resource() by enforcing particular
1992 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1993 Details : Instead of specifying each ldlm_lock call-back through separate
1994 parameter, wrap them into struct ldlm_callback_suite.
1998 Description: Kill join_lru obd method and its callers.
1999 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2000 in memory---a requirement implemented through ->o_join_lru() obd
2001 method in HEAD. Remove this method and its users.
2005 Description: Add asynchronous ldlm ENQUEUE completion handler.
2006 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2007 case that stock ldlm_completion_ast() cannot handle as it waits
2008 until lock is granted. Introduce new ldlm_completion_ast_async()
2009 for this. Also comment ldlm_completion_ast().
2013 Description: ldlm_error <-> errno conversion.
2014 Details : Add functions to map (rather arbitrary) between LDLM error codes
2015 and standard errno values. CLIO needs this to prevent LDLM specific
2016 constants from escaping ldlm and osc.
2020 Description: Kill unused ldlm_handle2lock_ns() function.
2021 Details : Kill unused ldlm_handle2lock_ns() function.
2025 Description: Add lu_ref support to ldlm_lock
2026 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2027 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2028 and ldlm_resource. LDLM interface has to be changed, because code that
2029 releases a reference on a lock, has to "know" what reference this is.
2030 In the most frequent case
2032 lock = ldlm_handle2lock(handle);
2034 LDLM_LOCK_PUT(lock);
2036 no changes are required. When any other reference (received _not_ from
2037 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2038 instead of LDLM_LOCK_PUT().
2040 Arguably, changes are pervasive, and interface requires some discipline
2041 for proper use. On the other hand, it was very instrumental in finding
2042 a few leaked lock references.
2046 Description: Add ldlm_lock_addref_try().
2047 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2048 attempts to addref a lock that might be being canceled
2053 Description: Add ldlm_weigh_callback().
2054 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2055 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2056 hard-coded `number of pages' logic.
2060 Description: Add lockdep annotations to llog code.
2061 Details : Use appropriately tagged _nested() locking calls in the places
2062 where llog takes more than one ->lgh_lock lock.
2066 Description: Add loi_kms_set().
2067 Details : Wrap kms updates into a helper function.
2071 Description: Constify instances of struct lsm_operations.
2072 Details : Constify instances of struct lsm_operations.
2076 Description: lu_conf support.
2077 Details : On a server, a file system object is uniquely identified
2078 by a fid, which is sufficient to locate and load all object
2079 state (inode). On a client, on the other hand, more data are
2080 necessary instantiate an object. Change lu_object_find() and
2081 friends to take additional `lu_conf' argument describing object.
2082 Typically this includes layout information.
2086 Description: lu_context fixes.
2087 Details : Introduce new lu_context functions that are needed on the client
2088 side, where some system threads (ptlrpcd) are shared by multiple
2089 modules, and so cannot be stopped during module shutdown.
2093 Description: Add start and stop methods to lu_device_type_operations.
2094 Details : Introduce two new methods in lu_device_type_operations, that are
2095 invoked when first instance of a given type is created and last one
2096 is destroyed respectively. This is need by CLIO.
2100 Description: Add lu_ref support to struct lu_device.
2101 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2102 track leaked references.
2106 Description: Introduce lu_kmem_descr.
2107 Details : lu_kmem_descr and its companion interface allow to create
2108 and destroy a number of kmem caches at once.
2112 Description: Fix lu_object finalization race.
2113 Details : Fix a race between lu_object_find() finding an object and its
2114 concurrent finalization. This race is (most likely) not possible
2115 on the server, but might happen on the client.
2119 Description: Introduce lu_ref interface.
2120 Details : lu_ref is a debugging module allowing to track references to
2121 a given object. It is quite cpu expensive, and has to be
2122 explicitly enabled with --enable-lu_ref. See usage description
2127 Description: Factor lu_site procfs stats into a separate function.
2128 Details : Separate lu_site stats printing code into a separate function
2129 that can be reused on a client.
2133 Description: Constify instances of struct {lu,dt,md}_device_operations.
2134 Details : Constify instances of struct {lu,dt,md}_device_operations.
2138 Description: Introduce struct md_site and move meta-data specific parts of
2139 struct lu_site here.
2140 Details : Move md-specific fields out of struct lu_site into special struct
2141 md_site, so that lu_site can be used on a client.
2145 Description: Kill mdd_lov_destroy().
2146 Details : Remove unused mdd code.
2150 Description: Add st_block checking to multistat.c.
2151 Details : Add st_block checking to multistat.c.
2155 Description: Add lu_ref support to struct obd_device.
2156 Details : Add lu_ref tracking to obd_device.
2160 Description: Kill obd_set_fail_loc().
2161 Details : Remove unused code.
2165 Description: Add special type for ptlrpc_request interpret functions.
2166 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2167 there. Actual usage will be in CLIO.
2171 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2172 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2173 doesn't work with lockdep.
2177 Description: Add rwv.c test program.
2178 Details : New testing program exercising readv(2) and writev(2) (Qian).
2182 Description: Add sendfile.c test program.
2183 Details : New testing program exercising sendfile(2) (Jay).
2187 Description: Ratelimit a message that can be very frequent.
2188 Details : Ratelimit a memory allocation failure message that can
2193 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2194 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2198 Description: Liblustre build fixes.
2199 Details : Liblustre build fixes.
2203 Description: libcfs: add cfs_{need,cond}_resched() interface.
2204 Details : libcfs: add cfs_{need,cond}_resched() definition and
2205 implementations for Linux, NT, and liblustre.
2207 Severity : enhancement
2209 Description: More exported tunables for mballoc
2210 Details : Add support for tunable preallocation window and new tunables for
2211 large/small requests
2215 Description: Detect corruption of block bitmap and checking for preallocations
2216 Details : Checks validity of on-disk block bitmap. Also it does better
2217 checking of number of applied preallocations. When corruption is
2218 found, it turns filesystem readonly to prevent further corruptions.
2222 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2223 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2224 window to 0. If it does hit cache miss, and read-pattern is still
2225 stride-io mode, does not reset the stride window, but also does not
2226 increase the stride window length in this case.
2230 Frequency : only for big-endian servers
2231 Description: Check if system is big-endian while mounting fs with extents feature
2232 Details : Mounting a filesystem with extents feature will fail on big-endian
2233 systems since ext3-based ldiskfs is not supported on big-endian
2234 systems. This can be over-riden with "bigendian_extents" mount option.
2236 Severity : enhancement
2238 Description: The root squash functionality
2239 Details : A security feature, which is to prevent users from being able
2240 to mount lustre on their desktop, run as root, and delete
2241 all of the files in the filesystem. The goal is accomplished by
2242 remapping user id (UID) and group id (GID) of the root user to
2243 a UID and GID specified by the system administartor via Lustre
2244 configuration management server (MGS). The functionality also
2245 allows to specify sets of clients for which the remapping does
2250 Description: Excessive recovery window
2251 Details : With AT enabled, the recovery window can be excessively long (6000+
2252 seconds). To address this problem, we no longer use
2253 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2254 timeout no longer depends on the service time, it is set to
2255 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2256 time via pb_service_time.
2260 Description: Watchdog triggered on MDS failover
2261 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2262 the OSTs know that the MDS "UUID" can be reused for the same export
2263 from a different NID, so we do not need to wait for the export to be
2267 Frequency : rare, only if using MMP with Linux RAID
2269 Description: MMP doesn't work with Linux RAID
2270 Details : While using HA for Lustre servers with Linux RAID, it is possible
2271 that MMP will not detect multiple mounts. To make this work we
2272 need to unplug the device queue in RAID when the MMP block is being
2273 written. Also while reading the MMP block, we should read it from
2274 disk and not the cached one.
2276 Severity : enhancement
2278 Description: open file using fid
2279 Details : A file can be opened using just its fid, like
2280 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2283 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2286 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2287 Details : add proper locking for creation of jbd_4k slab cache
2291 Description: MMP check in ext3_remount() fails without displaying any error
2292 Details : When multiple mount protection fails during remount, proper error
2295 Severity : enhancement
2297 Description: Allow stripe size to be up to 4G-64k
2298 Details : Fix math logic to allow large stripe sizes.
2302 Description: add check for >8TB ldiskfs filesystems
2303 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2304 Don't allow >8TB ldiskfs filesystems to be mounted without
2305 force_over_8tb mount option
2310 Descriptoin: OSS complains "too many blocks" and crashes
2311 Details : In filter_commitrw_write():
2312 1. lquota_chkquota();
2313 2. obdo_from_inode();
2314 3. lquota_pending_commit();
2315 "2." has changed uid and gid in this case, then caused
2316 lquota_chkquota() and lquota_pending_commit() to use
2317 different uid and gid. That is the root of the bug.
2321 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2322 Details : Detect and print out the 'stripe count = -1' case properly.
2326 Description: Out or order replies might be lost on replay
2327 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2328 request with tid smaller then one currently being added, add it
2329 to the start, not end of the list.
2331 --------------------------------------------------------------------------------
2333 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2335 * Support for kernels:
2336 2.6.5-7.283 (SLES 9),
2337 2.6.9-55.EL (RHEL 4),
2338 2.6.16.46-0.14 (SLES 10),
2339 2.6.18.8 vanilla (kernel.org)
2340 * Client support for unpatched kernels:
2341 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2342 2.6.16 - 2.6.22 vanilla (kernel.org)
2343 * Due to recently discovered recovery problems, we do not recommend
2344 using patchless RHEL 4 clients with this or any earlier release.
2345 * Recommended e2fsprogs version: 1.39.cfs8
2346 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2347 * Starting with this release, the ldiskfs backing filesystem required
2348 by Lustre is now in its own package, lustre-ldiskfs. This package
2349 should be installed. It is versioned separately from Lustre and
2350 may be released separately in future.
2354 Description: block reactivating mgc import until all deactivates complete
2355 Details : Fix race when failing back MDT/MGS to itself (testing)
2358 Frequency : at statup only
2360 Description: mds_lov_synchronize race leads to various problems
2361 Details : simultaneous MDT->OST connections at startup can cause the
2362 sync to abort, leaving the OSC in a bad state.
2364 Severity : enhancement
2366 Description: add optional extra BUILD_VERSION info
2367 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2368 to override the lustre version.
2371 Frequency : 2.6.18 servers only
2373 Description: ll_kern_mount() doesn't release the module reference
2374 Details : The ldiskfs module reference count never drops down to 0
2375 because ll_kern_mount() doesn't release the module reference.
2380 Description: server LBUG when using old ost_num_threads parameter
2381 Details : Accept the old ost_num_threads parameter but warn that it
2382 is deprecated, and fix an off-by-one error that caused an LBUG.
2387 Description: Transient SCSI error results in persistent IO issue
2388 Details : iobuf->dr_error is not reinitialized to 0 between two
2392 Frequency : sometimes when underlying device returns I/O errors
2394 Description: OSTs not going read-only during write failures
2395 Details : OSTs are not remounted read-only when the journal commit threads
2396 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2397 instead of the ext3 wrappers.
2401 Description: poor connect scaling with increasing client count
2402 Details : Don't run filter_grant_sanity_check for more than 100 exports
2403 to improve scaling for large numbers of clients.
2406 Frequency : SLES10 only
2408 Description: sanity-quota.sh quotacheck failed: rc = -22
2409 Details : Quotas cannot be enabled on SLES10.
2412 Frequency : liblustre clients only
2414 Description: getdirentries does not give error when run on compute nodes
2415 Details : getdirentries does not fail when the size specified as an argument
2416 is too small to contain at least one entry
2418 Severity : enhancement
2420 Description: Add LNET router traceability for debug purposes
2421 Details : If a checksum failure occurs with a router as part of the
2422 IO path, the NID of the last router that forwarded the bulk data
2423 is printed so it can be identified.
2428 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2429 Details : Due to a race condition, liblustre clients were occasionally
2430 evicted incorrectly.
2432 Severity : enhancement
2434 Description: lfs setstripe use optional parameters instead of postional
2437 Severity : enhancement
2439 Description: Nanosecond timestamp support for ldiskfs
2440 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2441 resolution timestamps. There is not yet support for this at
2442 the Lustre filesystem level.
2445 Frequency : during server recovery
2447 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2448 Details : request with rq_no_resend flag not awake l_wait_event if they get
2452 Frequency : nfs export on patchless client
2454 Description: connectathon hang when test nfs export over patchless client
2455 Details : Disconnected dentry cannot be found with lookup, so we do not need
2456 to unhash it or make it invalid
2459 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2460 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2461 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2466 Description: open req refcounting wrong on reconnect
2467 Details : If reconnect happened between getting open reply from server and
2468 call to mdc_set_replay_data in ll_file_open, we will schedule
2469 replay for unreferenced request that we are about to free.
2470 Subsequent close will crash in variety of ways.
2471 Check that request is still eligible for replay in
2472 mdc_set_replay_data().
2477 Description: disable writes to filesystem when reading health_check file
2478 Details : the default for reading the health_check proc file has changed
2479 to NOT do a journal transaction and write to disk, because this
2480 can cause reads of the /proc file to hang and block HA state
2481 checking on a healthy but otherwise heavily loaded system. It
2482 is possible to return to the previous behaviour during configure
2483 with --enable-health-write.
2485 Severity : enhancement
2487 Description: 64-bit inode version
2488 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2489 to the inode. This will be required for version-based recovery.
2494 Description: MDS fails to start if a duplicate client export is detected
2495 Details : in some rare cases it was possible for a client to connect to
2496 an MDS multiple times. Upon recovery the MDS would detect this
2497 and fail during startup. Handle this more gracefully.
2499 Severity : enhancement
2501 Description: Add -o localflock option to simulate old noflock
2503 Details : This will achieve local-only flock/fcntl locks
2509 Description: log_commit_thread vs filter_destroy race leads to crash
2510 Details : Take import reference before releasing llog record semaphore
2515 Description: Wrong request locking in request set processing
2516 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2517 imp_delayed_list, in this place should be used imp_lock.
2520 Frequency : when reconnection
2522 Description: Grant Leak when osc reconnect to OST
2523 Details : When osc reconnect ost, OST(filter) should check whether it
2524 should grant more space to client by comparing fed_grant and
2525 cl_avail_grant, and return the granted space to client instead
2526 of "new granted" space, because client will call osc_init_grant
2527 to update the client grant space info.
2530 Frequency : when client reconnect to OST
2532 Description: Grant Leak when osc do resend and replay bulk write
2533 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2534 bulk write request, because the grant info while be sync between
2535 OSC and OST when reconnect, and we should ignore the grant info
2536 these of resend/replay write req.
2541 Description: Grant space more than avaiable left space sometimes.
2542 Details : When then OST is about to be full, if two bulk writing from
2543 different clients came to OST. Accord the avaliable space of the
2544 OST, the first req should be permitted, and the second one
2545 should be denied by ENOSPC. But if the seconde arrived before
2546 the first one is commited. The OST might wrongly permit second
2547 writing, which will cause grant space > avaiable space.
2550 Frequency : when client is evicted
2552 Description: Grant might be wrongly erased when osc is evicted by OST
2553 Details : when the import is evicted by server, it will fork another
2554 thread ptlrpc_invalidate_import_thread to invalidate the
2555 import, where the grant will be set to 0. While the original
2556 thread will update the grant it got when connecting. So if
2557 the former happened latter, the grant will be wrongly errased
2558 because of this race.
2563 Description: Checking Stale with correct fid
2564 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2565 to check whether it is stale, because sometimes, we want the
2566 enqueue happened anyway, and op_data.fid2 will not be initialized.
2568 Severity : enhancement
2570 Description: update patchless client
2571 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2574 Frequency : only with 2.4 kernel
2576 Description: random memory corruption
2577 Details : size of struct ll_inode_info is to big for union inode.u and this
2578 can be cause of random memory corruption.
2583 Description: Memory leak in recovery
2584 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2585 It should also check obd_fail before fsfilt_start, otherwise if
2586 fsfilt_start return -EROFS,(failover mds during mds recovery).
2587 then the req will return with repmsg->transno = 0 and rc = EROFS.
2588 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2589 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2590 be freed no matter whether fsfilt_commit success or not.
2593 Frequency : only with huge count clients
2595 Description: Prevents from taking the superblock lock in llap_from_page for
2597 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2598 for llap_from_page call in ll_removepage prevents from taking the
2599 superblock lock for a soon died page.
2604 Description: Not check open intent error before release open handle
2605 Details : in some rare cases, the open intent error is not checked before
2606 release open handle, which may cause
2607 ASSERTION(open_req->rq_transno != 0), because it tries to release
2608 the failed open handle.
2613 Description: Set cat log bitmap only after create log success.
2614 Details : in some rare cases, the cat log bitmap is set too early. and it
2615 should be set only after create log success.
2619 Description: Accessing a block bevice can re-enable I/O when Lustre is
2620 tearing down a device.
2621 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2626 Description: service threads may hog cpus when there are a lot of requests
2627 Details : Insert cond_resched to give other threads a chance to use some CPU
2632 Description: the cat log was not initialized in recovery
2633 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2634 unlink log on mds will not be initialized until mds post
2635 recovery. And also in mds post recovery, the unlink log will
2636 initialization will be done asynchronausly, so there will be race
2637 between add unlink log and unlink log initialization.
2641 Description: brw_stats were being printed incorrectly
2642 Details : brw_stats were being printed as log2 but all of them were not
2643 recorded as log2. Also remove some code duplication arising from
2644 filter_tally_{read,write}.
2648 Frequency : rare, only in recovery.
2649 Description: ASSERTION(req->rq_type != LI_POISON) failed
2650 Details : imp_lock should be held while iterating over imp_sending_list for
2651 prevent destroy request after get timeout in ptlrpc_queue_wait.
2655 Description: replay-single.sh test 52 fails
2656 Details : A lock's skiplist need to be cleanup when it being unlinked
2657 from its resource list.
2661 Description: Short directio read returns full requested size rather than
2663 Details : Direct I/O operations should return actual amount of bytes
2664 transferred rather than requested size.
2666 Severity : enhancement
2668 Description: metadata RPC reduction (e.g. for rm performance)
2669 Details : decrease the amount of synchronous RPC between clients and servers
2670 by canceling conflicing lock before the operation on the client
2671 and packing thier handles into the main operation RPC to server.
2673 Severity : enhancement
2675 Description: Async OSC create to avoid the blocking unnecessarily.
2676 Details : If a OST has no remain object, system will block on the creating
2677 when need to create a new object on this OST. Now, ways use
2678 pre-created objects when available, instead of blocking on an
2679 empty osc while others are not empty. If we must block, we block
2680 for the shortest possible period of time.
2684 Description: improve handling recoverable errors
2685 Details : if request processig with error which can be recoverable on server
2686 request should be resend, otherwise page released from cache and
2689 Severity : enhancement
2691 Description: refine locking for avoid write wrong info into lov_objid file
2692 Details : fix possible races with add new target and write/update data in
2695 --------------------------------------------------------------------------------
2697 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2702 Frequency : on some architectures
2704 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2705 Details : Uninitialized flags sometimes cause configuration commands to
2709 Frequency : patchless clients only
2711 Description: missing __iget() symbol export
2712 Details : The __iget() symbol export is missing. To avoid the need for
2713 this on patchless clients the deathrow inode reaper is turned
2714 off, and we depend on the VM to clean up old inodes. This
2715 dependency was during via the fix for bug 12181.
2717 --------------------------------------------------------------------------------
2719 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2721 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2722 INTEROPERATE with older versions automatically. In many cases a
2723 special upgrade step is needed. Please read the
2724 user documentation before upgrading any part of a live system.
2725 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2726 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2727 * WARNING: Lustre configuration and startup changes are required with
2728 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2732 Severity : enhancement
2734 Description: MountConf
2735 Details : Lustre configuration is now managed via mkfs and mount
2736 commands instead of lmc and lconf. New obd types (MGS, MGC)
2737 are added for dynamic configuration management. See
2738 https://mail.clusterfs.com/wikis/lustre/MountConf for
2741 Severity : enhancement
2743 Description: dynamic OST addition
2744 Details : OSTs can now be added to a live filesystem
2746 Severity : enhancement
2748 Description: startup order invariance
2749 Details : MDTs and OSTs can be started in any order. Clients only
2750 require the MDT to complete startup.
2752 Severity : enhancement
2754 Description: parallel, asynchronous orphan cleanup
2755 Details : orphan cleanup is now performed in separate threads for each
2756 OST, allowing parallel non-blocking operation.
2758 Severity : enhancement
2760 Description: optimized stripe assignment
2761 Details : stripe assignments are now made based on ost space available,
2762 ost previous usage, and OSS previous usage, in order to try
2763 to optimize storage space and networking resources.
2765 Severity : enhancement
2767 Description: Permanently set tunables
2768 Details : All writable /proc/fs/lustre tunables can now be permanently
2769 set on a per-server basis, at mkfs time or on a live system.
2771 Severity : enhancement
2773 Description: Lustre message v2
2774 Details : Add lustre message format v2.
2776 Severity : enhancement
2778 Description: client OST exclusion list
2779 Details : Clients can be started with a list of OSTs that should be
2780 declared "inactive" for known non-responsive OSTs.
2784 Description: SPEC SFS validation failure on NFS v2 over lustre.
2785 Details : Changes the blocksize for regular files to be 2x RPC size,
2786 and not depend on stripe size.
2788 Severity : enhancement
2790 Description: Multiple MD RPCs in flight.
2791 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2792 To avoid overly-overloading MDS, introduce a limit on number of
2793 MDS RPCs in flight for a single client and add /proc controls
2794 to adjust this limit.
2796 Severity : enhancement
2798 Description: client read/write statistics
2799 Details : Add client read/write call usage stats for performance
2800 analysis of user processes.
2801 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2802 file access. extents_stats shows chunk size distribution.
2803 extents_stats_per_process show chunk size distribution per
2806 Severity : enhancement
2808 Description: mds statistics
2809 Details : Add detailed mds operations statistics in
2810 /proc/fs/lustre/mds/*/stats.
2814 Description: Failure of copying files with lustre special EAs.
2815 Details : Client side always return success for setxattr call for lustre
2816 special xattr (currently only "trusted.lov").
2821 Description: Refcount LNET uuids
2822 Details : The global LNET uuid list grew linearly with every startup;
2823 refcount repeated list entries instead of always adding to
2826 Severity : enhancement
2828 Description: Dynamic service threads
2829 Details : Within a small range, start extra service threads
2830 automatically when the request queue builds up.
2833 Frequency : mixed-endian client/server environments
2835 Description: mixed-endian crashes
2836 Details : The new msg_v2 system had some failures in mixed-endian
2839 Severity : enhancement
2841 Description: Easy OST removal
2842 Details : OSTs can be permanently deactivated with e.g. 'lctl
2843 conf_param lustre-OST0001.osc.active=0'
2845 Severity : enhancement
2847 Description: MGS proc entries
2848 Details : Added basic proc entries for the MGS showing what filesystems
2851 Severity : enhancement
2853 Description: provide MGS failover
2854 Details : Added config lock reacquisition after MGS server failover.
2856 Severity : enhancement
2858 Description: add Linux 2.4 support
2859 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2863 Description: a large application tries to do I/O to the same resource and dies
2864 in the middle of it.
2865 Details : Check the req->rq_arrival time after the call to
2866 ost_brw_lock_get(), but before we do anything about
2867 processing it & sending the BULK transfer request. This
2868 should help move old stale pending locks off the queue as
2869 quickly as obd_timeout.
2872 Frequency : when an incorrect nid is specified during startup
2874 Description: ptlrpc connect to non-existant node causes kernel crash
2875 Details : LNET can't be re-entered from an event callback, which
2876 happened when we expire a message after the export has been
2877 cleaned up. Instead, hand the zombie cleanup off to another
2880 Severity : enhancement
2882 Description: plain/inodebits lock performance improvement
2883 Details : Grouping plain/inodebits in granted list by their request modes
2884 and bits policy, thus improving the performance of search through
2888 Frequency : only if OST filesystem is corrupted
2890 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2891 Details : for a short time RPCs with bulk IO are in the replay list,
2892 but replay of bulk IOs is unimplemented. If the OST filesystem
2893 is corrupted due to disk cache incoherency and then replay is
2894 started it is possible to trip an assertion. Avoid putting
2895 committed RPCs into the replay list at all to avoid this issue.
2898 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2900 Description: System hang on startup
2901 Details : This bug allowed the liblustre (e.g. catamount) client to
2902 return to the app before handling all startup RPCs. This
2903 could leave the node unresponsive to lustre network traffic
2904 and manifested as a server ptllnd timeout.
2906 Severity : enhancement
2908 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2909 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2910 boolean to control whether to print peer debug info when a
2911 client's RPC times out.
2914 Frequency : only for kernels with patches from Lustre below 1.4.3
2916 Description: Remove old rdonly API
2917 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2920 Frequency : only for devices with external journals
2922 Description: Set external device read-only also
2923 Details : During a commanded failover stop, we set the disk device
2924 read-only while the server shuts down. We now also set any
2925 external journal device read-only at the same time.
2928 Frequency : when upgrading from 1.4 while trying to change parameters
2930 Description: The wrong (new) MDC name was used when setting parameters for
2931 upgraded MDT's. Also allows changing of OSC (and MDC)
2932 parameters if --writeconf is specified at tunefs upgrade time.
2935 Frequency : when setting specific ost indicies
2937 Description: QOS code breaks on skipped indicies
2938 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2939 created with --index need not be sequential.
2941 Severity : enhancement
2943 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2944 Details : The uninit_groups feature works in conjunction with the kernel
2945 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2946 up the pass1 processing of e2fsck. This is a read-only feature
2947 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2948 mount filesystems that have had this feature enabled.
2950 Severity : enhancement
2952 Description: Improve multi-block allocation algorithm to avoid fragmentation
2953 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2954 allocation locality and avoid filesystem fragmentation.
2956 ------------------------------------------------------------------------------
2958 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2960 * Support for kernels:
2961 2.4.21-47.0.1.EL (RHEL 3)
2962 2.6.5-7.283 (SLES 9)
2963 2.6.9-42.0.10.EL (RHEL 4)
2964 2.6.12.6 vanilla (kernel.org)
2965 2.6.16.27-0.9 (SLES 10)
2966 * Recommended e2fsprogs version: 1.39.cfs5
2968 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2972 Frequency : occasional, depends on client load and configuration
2973 Bugzilla : 12181, 12203
2974 Description: data loss for recently-modified files
2976 Details : In some cases it is possible that recently written or created
2977 files may not be written to disk in a timely manner (this should
2978 normally be within 30s unless client IO load is very high).
2979 The problem appears as zero-length files or files that are a
2980 multiple of 1MB in size after a client crash or client eviction
2981 that are missing data at the end of the file.
2983 This problem is more likely to be hit on clients where files are
2984 repeatedly created and unlinked in the same directory, clients
2985 have a large amount of RAM, have many CPUs, the filesystem has
2986 many OSTs, the clients are rebooted frequently, and/or the files
2987 are not accessed by other nodes after being written.
2989 The presence of the problem can be detected by looking at
2990 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2991 smaller than the second (nr_unused) then dirty files will not
2992 be flushed automatically to disk. "sync; sleep 10" should be
2993 run several times on the node before unmounting it to update
2994 Lustre (this is also safe to run on nodes without this problem).
2996 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2997 that can cause this same problem, so customers using that kernel
2998 also need to update the kernel in addition to Lustre. In order
2999 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3001 It is normal that files written just before a client crash (less
3002 than 30s) may not yet have been flushed to disk, even for local
3006 Frequency : frequent on thin XT3 nodes
3008 Description: UUID collision on thin XT3 Linux nodes
3009 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3010 randomly, since we relied on an insufficiently-seeded PRNG.
3015 Description: OSS hangs after "All ost request buffers busy"
3016 Details : A deadlock between quota and journal operations caused OSS
3017 hangs after printing "All ost request buffers busy."
3020 Frequency : always on liblustre builds
3022 Description: Cleanup compiler warnings on liblustre
3025 Frequency : always on liblustre builds on XT3
3027 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3031 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3032 > 32000 subdirectories
3033 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3034 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3035 subdirectory count crosses 32000. This will aid e2fsck to
3036 correctly handle more than 32000 subdirectories.
3039 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3041 Description: System hang on startup
3042 Details : This bug allowed the liblustre (e.g. catamount) client to
3043 return to the app before handling all startup RPCs. This
3044 could leave the node unresponsive to lustre network traffic
3045 and manifested as a server ptllnd timeout.
3047 Severity : enhancement
3049 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3050 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3051 boolean to control whether to print peer debug info when a
3052 client's RPC times out.
3057 Description: make O_SYNC working on 2.6 kernels
3058 Details : 2.6 kernels use different method for mark pages for write,
3059 so need add a code to lustre for O_SYNC work.
3064 Description: Failure to close file and release space on NFS
3065 Details : Put inode details into lock acquired in ll_intent_file_open.
3066 Use mdc_intent_lock in ll_intent_open to properly
3067 detect all kind of errors unhandled by mdc_enqueue.
3072 Description: proc file read during shutdown sometimes raced obd removal,
3074 Details : Add lock to prevent obd access after proc file removal.
3077 Frequency : Only for files larger than 4GB on 32-bit clients.
3079 Description: improperly doing page alignment of locks
3080 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3081 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3086 Description: RPCs being resent when they shouldn't be
3087 Details : Some RPCs that should not be resent are being resent. This
3088 can cause inconsistencies in the RPC state machine. Do not
3089 resend such requests.
3092 Frequency : rare, only with NFS export
3094 Description: Crash on NFS re-export node
3095 Details : under very unusual load conditions an assertion is hit in
3096 ll_intent_file_open()
3099 Frequency : only if OST filesystem is corrupted
3101 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3102 Details : for a short time RPCs with bulk IO are in the replay list,
3103 but replay of bulk IOs is unimplemented. If the OST filesystem
3104 is corrupted due to disk cache incoherency and then replay is
3105 started it is possible to trip an assertion. Avoid putting
3106 committed RPCs into the replay list at all to avoid this issue.
3111 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3112 Details : Large single O_DIRECT read and write calls can fail to allocate
3113 a sufficiently large buffer to process the request. In case of
3114 allocation failure the allocation is retried with a smaller
3115 buffer and broken into smaller requests.
3117 Severity : enhancement
3119 Description: Add -o localflock option to simulate old noflock behaviour.
3120 Details : This will achieve local-only flock/fcntl locks coherentness.
3125 Description: versioning check is incomplete
3126 Details : Checking the version difference of client vs. server, report
3127 error if the gap is too big.
3132 Description: add support PG_writeback bit
3133 Details : add support for PG_writeback bit for Lustre, for more carefull
3134 work with page cache in 2.6 kernel. This also fix some deadlocks
3135 and remove hack for work O_SYNC with 2.6 kernel.
3137 Severity : enhancement
3139 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3140 Details : The uninit_groups feature works in conjunction with the kernel
3141 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3142 up the pass1 processing of e2fsck. This is a read-only feature
3143 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3144 mount filesystems that have had this feature enabled.
3146 Severity : enhancement
3148 Description: Improve multi-block allocation algorithm to avoid fragmentation
3149 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3150 allocation locality and avoid filesystem fragmentation.
3152 ------------------------------------------------------------------------------
3154 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3156 * Support for kernels:
3157 2.6.9-42.0.3.EL (RHEL 4)
3158 2.6.5-7.276 (SLES 9)
3159 2.4.21-47.0.1.EL (RHEL 3)
3160 2.6.12.6 vanilla (kernel.org)
3161 2.6.16.21-0.8 (SLES10)
3162 * Recommended e2fsprogs version: 1.39.cfs2-0
3164 * The backwards-compatible /proc/sys/portals symlink has been removed
3165 in this release. Before upgrading, please ensure that you change
3166 any configuration scripts or /etc/sysctl.conf files that access
3167 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3168 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3169 in advance of the upgrade on any system running Lustre 1.4.6 or
3170 newer, since /proc/sys/lnet was added in that version.
3171 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3175 Frequency : only when quota is used
3177 Description: avoid scanning export list for quota master
3178 Details : Change the algorithms to avoid scanning export list in order
3179 to improve the efficiency.
3182 Frequency : MDS failover only, very rarely
3184 Description: "went back in time" messages on mds failover
3185 Details : The greatest transno may be lost when the current operation
3186 finishes with an error (transno==0) and the client's last_rcvd
3187 record is over-written. Save the greatest transno in the
3188 mds_last_transno for this case.
3191 Frequency : always for specific kernels and striping counts
3193 Description: client may get "Matching packet too big" without ACL support
3194 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3195 when trying to access files in certain configurations. The
3196 clients should in fact be denied when mounting because they do
3197 not understand ACLs.
3200 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3202 Description: many clients connecting with IO in progress causes connect timeouts
3203 Details : Avoid synchronous journal commits to avoid delays caused by many
3204 clients connecting/disconnecting when bulk IO is in progress.
3205 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3206 OST_IO_PORTAL to avoid delays behind potentially many pending
3210 Frequency : occasionally with multiple writers to a single file
3212 Description: shared writes to file may result in wrong size reported by stat()
3213 Details : Allow growing of kms when extent lock is cancelled
3216 Frequency : always with random mmap IO to multi-striped file
3218 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3219 Details : Only if the hole is at the end of OST object so that kms is too
3220 small. Fix is to increase kms accordingly in ll_nopage.
3223 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3225 Description: writes to a missing object would leak memory on the OST
3226 Details : If there is an inconsistency between the MDS and OST filesystems,
3227 such that the MDS references an object that doesn't exist, writes
3228 to that object will leak memory due to incorrect cleanup in the
3229 error handling path, eventually running out of memory on the OST.
3234 Description: Creating too long symlink causes lustre errors
3235 Details : Check symlink and name lengths before sending requests to MDS.
3238 Frequency : only if flock is enabled (not on by default)
3240 Description: posix locks not released on fd closure on 2.6.9+
3241 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3242 kernels, this caused such locks not to be released on fd close and
3243 then assertions on fs unmount about still used locks.
3246 Frequency : MDS failover only, very rarely
3248 Description: clients may get ASSERTION(granted_lock != NULL)
3249 Details : When request was taking a long time, and a client was resending
3250 a getattr by name lock request. The were multiple lock requests
3251 with the same client lock handle and
3252 mds_getattr_name->fixup_handle_for_resent_request found one of the
3253 lock handles but later failed with ASSERTION(granted_lock != NULL).
3258 Description: handle->h_buffer_credits > 0, assertion failure
3259 Details : h_buffer_credits is zero after truncate, causing assertion
3260 failure. This patch extends the transaction or creates a new
3264 Frequency : NFS re-export or patchless client
3265 Bugzilla : 11179, 10796
3266 Description: Crash on NFS re-export node (__d_move)
3267 Details : We do not want to hash the dentry if we don't have a lock.
3268 But if this dentry is later used in d_move, we'd hit uninitialised
3269 list head d_hash, so we just do this to init d_hash field but
3270 leave dentry unhashed.
3273 Frequency : NFS re-export or patchless client
3275 Description: NFS exports has problem with symbolic link
3276 Details : lustre client didn't properly install dentry when re-exported
3277 to NFS or running patchless client.
3280 Frequency : NFS re-export or patchless client
3282 Description: Various nfs/patchless fixes.
3283 Details : fixes reuse disconected alias for lookup process - this fixes
3284 warning "find_exported_dentry: npd != pd",
3285 fix permission error with open files at nfs.
3286 fix apply umask when do revalidate.
3289 Frequency : occasional
3291 Description: Crash on NFS re-export node
3292 Details : calling clear_page() on the wrong pointer triggered oops in
3293 generic_mapping_read().
3296 Frequency : rarely, using O_DIRECT IO
3298 Description: unaligned directio crashes client with LASSERT
3299 Details : check for unaligned buffers before trying any requests.
3302 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3304 Description: stale data returned from RAID cache
3305 Details : If only a small amount of IO is done to the RAID device before
3306 reading it again it is possible to get stale data from the RAID
3307 cache instead of reading it from disk.
3310 Frequency : always for sles10 kernel
3312 Description: sles10 support
3313 Details : ll_follow_link: compile fixes and using of nd_set_link
3314 under newer kernels.
3317 Frequency : depends on arch, kernel and compiler version, always on sles10
3320 Description: recursive or deep enough symlinks cause stack overflow
3321 Details : getting rid of large stack-allocated variable in
3325 Frequency : depends on hardware
3327 Description: lustre write performance loss in the SLES10 kernel
3328 Details : the performance loss is caused by using of write barriers in the
3329 ext3 code. The SLES10 kernel turns barrier support on by
3330 default. The fix is to undo that change for ldiskfs.
3332 ------------------------------------------------------------------------------
3334 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3336 * Support for kernels:
3337 2.6.9-42.0.3EL (RHEL 4)
3338 2.6.5-7.276 (SLES 9)
3339 2.4.21-47.0.1.EL (RHEL 3)
3340 2.6.12.6 vanilla (kernel.org)
3344 Frequency : quota enabled and large files being deleted
3346 Description: releasing more than 4GB of quota at once hangs OST
3347 Details : If a user deletes more than 4GB of files on a single OST it
3348 will cause the OST to spin in an infinite loop. Release
3349 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3354 Description: statfs data retrieved from /proc may be stale or zero
3355 Details : When reading per-device statfs data from /proc, in the
3356 {kbytes,files}_{total,free,avail} files, it may appear
3357 as zero or be out of date.
3360 Frequency : systems with MD RAID1 external journal devices
3362 Description: lconf's call to blkid is confused by RAID1 journal devices
3363 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3364 of returning all block devices that match the journal UUID.
3367 Frequency : always, for aggregate stripe size over 4GB
3369 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3370 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3371 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3374 Frequency : always if "lfs find" used on a local file/directory
3376 Description: "lfs find" segfaults if used on a local file/directory
3377 Details : The case where a directory component was not specified wasn't
3378 handled correctly. Handle this properly.
3381 Frequency : always on ppc64
3383 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3384 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3385 perform endianness conversion.
3388 Frequency : rarely (truncate to non-zero file size after write under load)
3389 Bugzilla : 10730, 10687
3390 Description: Files padded with zeros to next 4K multiple
3391 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3392 it is possible that files that are truncated to a non-zero size
3393 immediately after being written are filled with zero bytes beyond
3394 the truncated size. No file data is lost.
3396 Severity : enhancement
3398 Description: Allow recovery/failover for liblustre clients.
3399 Details : liblustre clients were unaware of failover configurations until
3402 Severity : enhancement
3404 Description: user file locks should fail when not mounting with flock option
3405 Details : Set up an error-returning stub in ll_file_operations.lock field
3406 to prevent incorrect behaviour when client is mounted without
3407 flock option. Also, set up properly f_op->flock field for
3411 Frequency : always on ia64
3413 Description: "lfs df" loops on printing out MDS statfs information
3414 Details : The obd_ioctl_data was not initialized and in some systems
3415 this caused a failure during the ioctl that did not return
3416 an error. Initialize the struct and return an error on failure.
3419 Frequency : SLES 9 only
3421 Description: Error of copying files with lustre special EAs as root
3422 Details : Client side always return success for setxattr call for lustre
3423 special xattr (currently only "trusted.lov").
3426 Frequency : rarely on clusters with both ia64+i386 clients
3428 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3429 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3430 clients are doing concurrent writes to the same file it is
3431 possible that the ia64 clients may LASSERT because the OST
3432 extent locks are not PAGE_SIZE aligned. Ensure that grown
3433 locks are always aligned on the request boundary.
3436 Frequency : specific use, occasional
3438 Description: Overwriting in use executable truncates on-disk binary image
3439 Details : If one node attempts to overwrite an executable in use by
3440 another node, we now correctly return ETXTBSY instead of
3441 truncating the file.
3443 Severity : enhancement
3445 Description: Async OSC create to avoid the blocking unnecessarily.
3446 Details : If a OST has no remain object, system will block on the creating
3447 when need to create a new object on this OST. Now, ways use
3448 pre-created objects when available, instead of blocking on an
3449 empty osc while others are not empty. If we must block, we block
3450 for the shortest possible period of time.
3455 Description: chmod on Lustre root is propagated to other clients
3456 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3457 invalid by the follow_mount time.
3460 Frequency : liblustre clients only
3462 Description: Race in 'instant cancel' lock handling could lead to such locks
3463 never to be granted in case of SMP MDS
3464 Details : Do not destroy not yet granted but cbpending locks in
3468 Frequency : replay/resend of open
3470 Description: non null lock assetion failure in mds_intent_policy
3471 Details : Trying to replay/resend lockless open requests resulted in
3472 mds_open() returning 0 with no lock. Now it sets a flag if
3473 a lock is going to be returned.
3475 Severity : enhancement
3477 Description: Checksum enhancements
3478 Details : New checksum enhancements allow for resending RPCs that failed
3481 Severity : enhancement
3483 Description: Tunables on number of dirty pages in cacche
3484 Details : Allow to set limit on number of dirty pages cached.
3489 Description: client crash on unmount - lock still has references
3490 Details : In some error handling cases it was possible to leak a lock
3491 reference on a client while accessing a file. This was not
3492 harmful to the client during operation, but would cause the
3493 client to crash when the filesystem is unmounted.
3496 Frequency : specific case, rare
3498 Description: ETXTBSY on mds though file not in use by client
3499 Details : ETXTBSY is no longer incorrectly returned when attempting to
3500 chmod or chown a directory that the user previously tried to
3501 execute or a currently-executing binary.
3504 Frequency : extremely rare except on liblustre-based clients
3506 Description: Lustre space not freed when files are deleted
3507 Details : Clean up open-unlinked files after client eviction. Previously
3508 the unlink was skipped and the files remained as orphans.
3513 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3514 Details : In some cases it was possible to send a blocking callback to
3515 a client doing a glimpse, even though that client didn't get
3516 a lock granted. When the glimpse lock is cancelled on the OST
3517 the freed lock is left on the waiting list and corrupted the list.
3520 Frequency : all core dumps
3522 Description: Broke core dumps to lustre
3523 Details : Negative dentry may be unhashed if parent does not have UPDATE
3524 lock, but some callers, e.g. do_coredump, expect dentry to be
3525 hashed after successful create, hash it in ll_create_it.
3527 ------------------------------------------------------------------------------
3529 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3531 * Support for kernels:
3532 2.6.9-42.0.2.EL (RHEL 4)
3533 2.6.5-7.276 (SLES 9)
3534 2.4.21-40.EL (RHEL 3)
3535 2.6.12.6 vanilla (kernel.org)
3539 Frequency : always on RHEL 3
3541 Description: Number of open files grows over time
3542 Details : The number of open files grows over time, whether or not
3543 Lustre is started. This was due to a filp leak introduced
3544 by one of our kernel patches.
3546 ------------------------------------------------------------------------------
3548 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3550 * Support for kernels:
3551 2.6.9-42.EL (RHEL 4)
3552 2.6.5-7.276 (SLES 9)
3553 2.4.21-40.EL (RHEL 3)
3554 2.6.12.6 vanilla (kernel.org)
3559 Bugzilla : 5719, 9635, 9792, 9684
3560 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3561 Details : If a server is under heavy load and cannot reply to new
3562 connection requests before the client resends the (re)connect,
3563 the connection handling code can behave badly if two service
3564 threads are concurrently handing separate (re)connections from
3565 the same client. Add better locking to the connection handling
3566 code, and ensure that only a single connection will be processed
3567 for a given client UUID, even if the lock is dropped.
3569 Severity : enhancement
3571 Description: add TCP zero-copy support to kernel
3572 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3573 sends if the hardware supports scatter-gather and checksumming.
3574 This allows socklnd to do client-write and server-read more
3575 efficiently and reduce CPU utilization from skbuf copying.
3578 Frequency : only if NFS exporting from client
3580 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3581 Details : If setting attributes on a file created under NFS that had
3582 never been opened it would be possible to oops the client
3583 if the file had no objects.
3586 Frequency : always for liblustre
3588 Description: liblustre client does MDS+OSTs setattr RPC for each write
3589 Details : When doing a write from a liblustre client, the client
3590 incorrectly issued an RPC to the MDS and each OST the file was
3591 striped over in order to update the timestamps. When writing
3592 with small chunks and many clients this could overwhelm the MDS
3593 with RPCs. In all cases it would slow down the write because
3594 these RPCs are unnecessary.
3596 Severity : enhancement
3598 Description: allow number of MDS service threads to be changed at module load
3599 Details : It is now possible to change the number of MDS service threads
3600 running. Adding "options mds mds_num_threads={N}" to the MDS's
3601 /etc/modprobe.conf will set the number of threads for the next
3602 time Lustre is restarted (assuming the "mds" module is also
3603 reloaded at that time). The default number of threads will
3604 stay the same, 32 for most systems.
3609 Description: OST crash if filesystem is unformatted or corrupt
3610 Details : If an OST is started on a device that has never been formatted
3611 or if the filesystem is corrupt and cannot even mount then the
3612 error handling cleanup routines would dereference a NULL pointer.
3617 Description: NULL pointer deref in llap_from_page.
3618 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3619 incorrect) mapping if the page was truncated/reclaimed while it was
3620 searched for. Check for this condition and skip such pages when
3621 doing readahead. Introduce extra check to llap_from_page() to
3622 verify page->mapping->host is non-NULL (so page is not anonymous).
3625 Frequency : Sometimes when using sys_sendfile
3627 Description: "page not covered by a lock" warnings from ll_readpage
3628 Details : sendfile called ll_readpage without right page locks present.
3629 Now we introduced ll_file_sendfile that does necessary locking
3630 around call to generic_file_sendfile() much like we do in
3634 Frequency : with certain MDS communication failures at client mount time
3636 Description: NULL pointer deref after failed client mount
3637 Details : a client connection request may delayed by the network layer
3638 and not be sent until after the PTLRPC layer has timed out the
3639 request. If the client fails the mount immediately it will try
3640 to clean up before the network times out the request. Add a
3641 reference from the request import to the obd device and delay
3642 the cleanup until the network drops the request.
3645 Frequency : occasionally during client (re)connect
3647 Description: assertion failure during client (re)connect
3648 Details : processing a client connection request may be delayed by the
3649 client or server longer than the client connect timeout. This
3650 causes the client to resend the connection request. If the
3651 original connection request is replied in this interval, the
3652 client may trip an assertion failure in ptlrpc_connect_interpret()
3653 which thought it would be the only running connect process.
3656 Frequency : only with obd_echo servers and clients that are rebooted
3658 Description: kernel BUG accessing uninitialized data structure
3659 Details : When running an obd_echo server it did not start the ping_evictor
3660 thread, and when a client was evicted an uninitialized data
3661 structure was accessed. Start the ping_evictor in the RPC
3662 service startup instead of the OBD startup.
3664 Severity : enhancement
3665 Bugzilla : 10193 (patchless)
3666 Description: Remove dependency on various unexported kernel interfaces.
3667 Details : No longer need reparent_to_init, exit_mm, exit_files,
3668 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3671 Frequency : rare (only users of deprecated and unsupported LDAP config)
3673 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3674 Details : LDAP apparently contains 'lustreName' attributes instead of
3675 'name'. A simple remapping of the name is sufficient.
3678 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3680 Description: waiting_locks_callback trips kernel BUG if client is evicted
3681 Details : Running with the dump_on_timeout debug flag turned on makes
3682 it possible that the waiting_locks_callback() can try to dump
3683 the Lustre kernel debug logs from an interrupt handler. Defer
3684 this log dumping to the expired_lock_main() thread.
3686 Severity : enhancement
3688 Description: Support NFS exporting on 2.6 kernels.
3689 Details : Implement non-rawops metadata methods for NFS server to use without
3690 changing NFS server code.
3693 Frequency : very rare (synthetic metadata workload only)
3695 Description: two racing renames might cause an MDS thread to deadlock
3696 Details : Running the "racer" program may cause one MDS thread to rename
3697 a file from being the source of a rename to being the target of
3698 a rename at exactly the same time that another thread is doing
3699 so, and the second thread has already enqueued these locks after
3700 doing a lookup of the target and is trying to relock them in
3701 order. Ensure that we don't try to re-lock the same resource.
3704 Frequency : only very large systems with liblustre clients
3706 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3707 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3708 all OSTs in parallel. This allows the eviction of stale liblustre
3709 clients to proceed much faster than if they were done in series,
3710 and also offers similar improvements for other set_info RPCs.
3715 Description: excessive CPU usage during initial read phase on client
3716 Details : During the initial read phase on a client, it would agressively
3717 retry readahead on the file, consuming too much CPU and impacting
3718 performance (since 1.4.5.8). Improve the readahead algorithm
3719 to avoid this, and also improve some other common cases (read
3720 of small files in particular, where "small" is files smaller than
3721 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3726 Description: MDS crash when receiving packet with unknown intent.
3727 Details : Do not LBUG in unknown intent case, just return -EFAULT
3729 Severity : enhancement
3730 Bugzilla : 9293, 9385
3731 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3732 Details : Do not serialize getattr (non-intent version) and statfs.
3735 Frequency : occasional, when OST network is overloaded/intermittent
3737 Description: client evicted by OST after bulk IO timeout
3738 Details : If a client sends a bulk IO request (read or write) the OST
3739 may evict the client if it is unresposive to its data GET/PUT
3740 request. This is incorrect if the network is overloaded (takes
3741 too long to transfer the RPC data) or dropped the OST GET/PUT
3742 request. There is no need to evict the client at all, since
3743 the pinger and/or lock callbacks will handle this, and the
3744 client can restart the bulk request.
3747 Frequency : Always when mmapping file with no objects
3749 Description: client crashes when mmapping file with no objects
3750 Details : Check that we actually have objects in a file before doing any
3751 operations on objects in ll_vm_open, ll_vm_close and
3757 Description: Request leak when working with deleted CWD
3758 Details : Introduce advanced request refcount tracking for requests
3759 referenced from lustre intent.
3761 Severity : Enhancement
3763 Description: Cache open file handles on client.
3764 Details : MDS now will return special lock along with openhandle, if
3765 requested and client is allowed to hold openhandle, even if unused,
3766 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3767 closing files for every read/write openration.
3769 Severity : Enhancement
3771 Description: Cache open negative dentries on client when possible.
3772 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3773 negative dentries on lock revocation.
3778 Description: Remounting a client read-only wasn't possible with a zconf mount
3779 Details : It wasn't possible to remount a client read-only with llmount.
3781 Severity : enhancement
3782 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3783 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3784 MPICH is included to add Lustre-specific ADIO interfaces.
3785 This is based closely on the UFS ADIO layer and only differs
3786 in file creation, in order to allow the OST striping to be set.
3787 This is user-contributed code and not supported by CFS.
3792 Description: extended inode attributes (immutable, append-only) work improperly
3793 when 2.4 and 2.6 kernels are used on client/server or vice versa
3794 Details : Introduce kernel-independent values for these flags.
3796 Severity : enhancement
3799 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3800 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3801 granularity. Now, Fractional MB granularity is be supported,
3802 this is very useful for low memory system.
3804 Severity : enhancement
3806 Description: Getattr by fid
3807 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3808 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3809 have locks and avoids getattr rpc requests for it on every stat.
3812 Frequency : Always, for filesystems larger than 2TB
3814 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3815 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3816 filesystems larger than 2TB to be mounted with mballoc enabled.
3819 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3821 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3822 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3823 filesystems larger than 2TB is mounted then the kernel will
3824 silently corrupt the start of the filesystem. CONFIG_LBD is
3825 enabled for all CFS-supported kernels, but the possibility of
3826 this happening with a modified kernel config exists.
3828 Severity : enhancement
3830 Description: add client O_DIRECT support for 2.6 kernels
3831 Details : It is now possible to do O_DIRECT reads and writes to files
3832 in the Lustre client mountpoint on 2.6 kernel clients.
3834 Severity : enhancement
3836 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3837 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3838 parallel, not waiting for response from every OST before sending
3839 a rpc to the next OST.
3844 Description: setattr vs write race when updating file timestamps
3845 Details : Client processes that update a file timestamp into the past
3846 right after writing to the file (e.g. tar) it is possible that
3847 the updated file modification time can be reset to the current
3848 time due to a race between processing the setattr and write RPC.
3850 Severity : enhancement
3852 Description: Bring 'lfs find' closer in line with regular Linux find.
3853 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3854 -print0 options and obtains all the needed info through the lustre
3857 Severity : enhancement
3859 Description: support up to 1024 configured devices on one node
3860 Details : change obd_dev array from statically allocated to dynamically
3861 allocated structs as they are first used to reduce memory usage
3866 Description: Flush dirty partially truncated pages during truncate
3867 Details : Immediatelly flush partially truncated pages in filter_setattr,
3868 this way we completely avoid having any pages in page cache on OST
3869 and can retire ugly workarounds during writes to flush such pages.
3874 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3875 Details : protect inode from truncation within vfs_unlink() context
3876 just take a reference before calling vfs_unlink() and release it
3877 when parent's i_sem is free.
3882 Description: last_id value checked outside lock on OST caused LASSERT failure
3883 Details : If there were multiple MDS->OST object precreate requests in
3884 flight, it was possible that the OST's last object id was checked
3885 outside a lock and incorrectly tripped an assertion. Move checks
3886 inside locks, and discard old precreate requests.
3889 Frequency : always, if extents are used on OSTs
3891 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3892 Details : OSTs using the extents format would not zero the high 16 bits of
3893 the index physical block number. This is not a problem for any
3894 OST filesystems smaller than 16TB, and no kernels support ext3
3895 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3896 new/modified files) and can be fixed for existing filesystems
3897 with e2fsprogs-1.39-cfs1.
3902 Description: import connection selection may be incorrect if timer wraps
3903 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3904 import connections to be ignored if the 32-bit jiffies counter
3905 wraps. Use a 64-bit jiffies counter.
3908 Frequency : very large clusters immediately after boot
3910 Description: LNET request buffers exhausted under heavy short-term load
3911 Details : If a large number of client requests are generated on a service
3912 that has previously never seen so many requests it is possible
3913 that the request buffer growth cannot keep up with the spike in
3914 demand. Instead of dropping incoming requests, they are held in
3915 the LND until the RPC service can accept more requests.
3918 Frequency : Sometimes during replay
3920 Description: Assertion failure in ll_local_open after replay.
3921 Details : If replay happened on an open request reply before we were able
3922 to set replay handler, reply will become not swabbed tripping the
3923 assertion in ll_local_open. Now we set the handler right after
3924 recognising of open request
3927 Frequency : very rare
3929 Description: kernel reports "badness in vsnprintf"
3930 Details : Reading from the "recovery_status" /proc file in small chunks
3931 may cause a negative length in lprocfs_obd_rd_recovery_status()
3932 call to vsnprintf() (which is otherwise harmless). Exit early
3933 if there is no more space in the output buffer.
3935 Severity : enhancement
3937 Description: clear OBD RPC statistics by writing to them
3938 Details : It is now possible to clear the OBD RPC statistics by writing
3939 to the "stats" file.
3944 Description: Client mtime is not the same on different clients after utimes
3945 Details : In some cases, the client was using the utimes() syscall on
3946 a file cached on another node. The clients now validate the
3947 ctime from the MDS + OSTs to determine which one is right.
3952 Description: Inability to activate failout mode
3953 Details : lconf script incorrectly assumed that in pythong string's numeric
3954 value is used in comparisons.
3957 Frequency : always with multiple stripes per file
3959 Description: Inefficient object allocation for mutli-stripe files
3960 Details : When selecting which OSTs to stripe files over, for files with
3961 a stripe count that divides evenly into the number of OSTs,
3962 the MDS is always picking the same starting OST for each file.
3963 Return the OST selection heuristic to the original design.
3968 Description: mount failures may take full timeout to return an error
3969 Details : Under some heavy load conditions it is possible that a
3970 failed mount can wait for the full obd_timeout interval,
3971 possibly several minutes, before reporting an error.
3972 Instead return an error as soon as the status is known.
3974 Frequency : quota enabled and large files being deleted
3976 Description: releasing more than 4GB of quota at once hangs OST
3977 Details : If a user deletes more than 4GB of files on a single OST it
3978 will cause the OST to spin in an infinite loop. Release
3979 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3984 Description: statfs data retrieved from /proc may be stale or zero
3985 Details : When reading per-device statfs data from /proc, in the
3986 {kbytes,files}_{total,free,avail} files, it may appear
3987 as zero or be out of date.
3990 Frequency : systems with MD RAID1 external journal devices
3992 Description: lconf's call to blkid is confused by RAID1 journal devices
3993 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3994 of returning all block devices that match the journal UUID.
3997 Frequency : always, for aggregate stripe size over 4GB
3999 Description: assertion fails when trying to use 4GB stripe size
4000 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4001 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4004 Frequency : always on ppc64
4006 Description: the first write on an ext3 filesystem with mballoc got stuck
4007 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4008 perform endianness conversion.
4010 ------------------------------------------------------------------------------
4012 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4014 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4015 INTEROPERATE with older versions automatically. Please read the
4016 user documentation before upgrading any part of a live system.
4017 * WARNING: Lustre networking configuration changes are required with
4018 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4021 * Support for kernels:
4022 2.6.9-22.0.2.EL (RHEL 4)
4023 2.6.5-7.244 (SLES 9)
4024 2.6.12.6 vanilla (kernel.org)
4027 Severity : enhancement
4028 Bugzilla : 7981/8208
4029 Description: Introduced Lustre Networking (LNET)
4030 Details : LNET is new networking infrastructure for Lustre, it includes
4031 a reorganized network configuration mode (see the user
4032 documentation for full details) as well as support for routing
4033 between different network fabrics. Lustre Networking Devices
4034 (LNDs) for the supported network fabrics have also been
4035 created for this new infrastructure.
4037 Severity : enhancement
4038 Description: Introduced Access control lists
4039 Details : clients can set ACLs on files and directories in order to have
4040 more fine-grained permissions than the standard Unix UGO+RWX.
4041 The MDS must be started with the "-o acl" mount option.
4043 Severity : enhancement
4044 Description: Introduced filesystem quotas
4045 Details : Administrators may now establish per-user quotas on the
4048 Severity : enhancement
4050 Description: Configuration change for the XT3
4051 The PTLLND is now used to run Lustre over Portals on the XT3
4052 The configure option(s) --with-cray-portals are no longer used.
4053 Rather --with-portals=<path-to-portals-includes> is used to
4054 enable building on the XT3. In addition to enable XT3 specific
4055 features the option --enable-cray-xt3 must be used.
4060 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4061 Details : A race between allocating a new llcd and re-getting the llcd_lock
4062 allowed another thread to grab newly-allocated llcd.
4064 Severity : enhancement
4066 Description: 2.6 OST async journal commit and locking fix to improve performance
4067 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4068 2.6 kernels are now async as they already were in 2.4 kernels so
4069 that they can commit concurrently with the network bulk transfer.
4070 For block-allocated files the filter allocation semaphore is held
4071 to avoid filesystem fragmentation during allocation. BKL lock
4072 removed for 2.6 xattr operations where it is no longer needed.
4077 Description: lconf incorrectly determined whether two IP networks could talk
4078 Details : In some more complicated routing and multiple-network
4079 configurations, lconf will avoid trying to make a network
4080 connection to a disjoint part of the IP space. It was doing the
4081 math incorrectly for one set of cases.
4086 Description: Fix for potential infinite loop processing records in an llog.
4087 Details : If an llog record is corrupted/zeroed, it is possible to loop
4088 forever in llog_process(). Validate the llog record length
4089 and skip the remainder of the block on error.
4092 Frequency : occasional (liblustre only)
4094 Description: liblustre could not open files whose last component is a symlink
4095 Details : sysio_path_walk() would incorrectly pass the open intent to
4096 intermediate path components.
4099 Frequency : rare (liblustre only with non-standard tuning)
4100 Bugzilla : 7201 (7350)
4101 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4102 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4103 after it has been released, only do this for referenced locks
4105 Severity : enhancement
4107 Description: specifying an (invalid) directory default stripe_size of -1
4108 would reset the directory default striping
4109 Details : stripe_size -1 was used internally to signal directory stripe
4110 removal, now use "all default" to signal dir stripe removal
4111 as a directory striping of "all default" is not useful
4114 Frequency : common for large clusters running liblustre clients
4116 Description: doing an ls when liblustre clients are running is slow
4117 Details : sending a glimpse AST to a liblustre client waits for every AST
4118 to time out, as liblustre clients will not respond. Since they
4119 cannot cache data we refresh the OST lock LVB from disk instead.
4121 Severity : enhancement
4123 Description: doing an ls at the same time as file IO can be slow
4124 Details : enqueue and other "small" requests can be blocked behind many
4125 large IO requests. Create a new OST IO portal for non-IO
4126 requests so they can be processed faster.
4129 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4131 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4132 created file with mode 000
4133 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4134 allow an ftruncate() as a non-root user to a file with mode 000.
4135 Lustre now respects this flag to disable mode checks when
4136 truncating a file owned by the user
4139 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4142 Description: Revoking locks from clients that went dead or catatonic might take
4144 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4145 cancellation of such locks instant on servers without waiting for
4146 any reply from clients. Clients drops these locks when cancel
4147 notification from server is received without replying.
4150 Frequency : liblustre-only, when liblustre client dies or becomes busy
4152 Description: Doing ls on Linux clients can take a long time with active
4154 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4155 granting such locks to it in the first place if possible. Locks
4156 are taken by proxy on the OST during the read or write and
4157 dropped immediately afterward. Add connect flags handling, do
4158 not grant locks to liblustre clients for glimpse ASTs.
4160 Severity : enhancement
4162 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4163 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4164 sequential reads and behaves badly with applications doing random
4165 reads. Improve it by reading ahead at least read region, and
4166 avoiding excessive large RPC for small reads.
4168 Severity : enhancement
4170 Description: Creating more than 1000 files for a single job may cause a load
4171 imbalance on the OSTs if there are also a large number of OSTs.
4172 Details : qos_prep_create() uses an OST index reseed value that is an
4173 even multiple of the number of available OSTs so that if the
4174 reseed happens in the middle of the object allocation it will
4175 still utilize the OSTs as uniformly as possible.
4180 Description: OST or MDS may oops in ping_evictor_main()
4181 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4182 but doesn't restart at beginning of obd_exports_timed list
4185 Severity : enhancement
4187 Description: improve by-nid export eviction on the MDS and OST
4188 Details : allow multiple exports with the same NID to be evicted at one
4189 time without re-searching the exports list.
4192 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4194 Description: MDS may oops in groups_free()
4195 Details : in rare race conditions a newly allocated group_info struct is
4196 freed again, and this can be NULL. The 2.4 compatibility code
4197 for groups_free() checked for a NULL pointer, but 2.6 did not.
4200 Frequency : common for liblustre clients doing little filesystem IO
4201 Bugzilla : 9352, 7313
4202 Description: server may evict liblustre clients accessing contended locks
4203 Details : if a client is granted a lock or receives a completion AST
4204 with a blocking AST already set it would not reply to the AST
4205 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4208 Frequency : lfs setstripe, only systems with more than 160 OSTs
4210 Description: unable to set striping with a starting offset beyond OST 160
4211 Details : llapi_create_file() incorrectly limited the starting stripe
4212 index to the maximum single-file stripe count.
4215 Frequency : LDAP users only
4217 Description: lconf did not handle in-kernel recovery with LDAP properly
4218 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4220 Severity : enhancement
4222 Description: bind OST threads to NUMA nodes to improve performance
4223 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4224 node and do their allocations there to localize memory access
4226 Severity : enhancement
4228 Description: llmount can determine client NID directly from Myrinet (GM)
4229 Details : the client NID code from gmnalnid was moved directly into
4230 llmount, removing the need to use this or specifying the
4231 client NID explicitly when mounting GM clients with zeroconf
4234 Frequency : if client is started with down MDS
4236 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4237 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4238 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4239 ptlrpc_import_delayed_req() because we didn't check if the
4240 request was interrupted, and we also didn't break out of the
4241 event loop if there was no timeout
4246 Description: data loss during non-page-aligned writes to a single file from
4247 both multiple nodes and multiple threads on one node at same time
4248 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4249 inconsistency led to false short-reads, that were cached and later
4250 used by ->prepare_write() to fill in partially written page,
4251 leading to data loss.
4254 Frequency : always, if lconf --abort_recovery used
4256 Description: lconf --abort_recovery fails with 'Operation not supported'
4257 Details : lconf was attempting to abort recovery on the MDT device and not
4260 Severity : enhancement
4262 Description: remove cleanup logs
4263 Details : replace lconf-generated cleanup logs with lustre internal
4264 cleanup routines. Eliminates the need for client-cleanup and
4267 Severity : enhancement
4269 Description: add support for EAs (user and system) on lustre filesystems
4270 Details : it is now possible to store extended attributes in the Lustre
4271 client filesystem, and with the user_xattr mount option it
4272 is possible to allow users to store EAs on their files also
4274 Severity : enhancement
4276 Description: Add possibility (config option) to show minimal available OST free
4278 Details : When compiled with --enable-mindf configure option, statfs(2)
4279 (and so, df) will return least minimal free space available from
4280 all OSTs as amount of free space on FS, instead of summary of
4281 free spaces of all OSTs.
4283 Severity : enhancement
4285 Description: do not expand extent locks acquired on OST-side
4286 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4287 by server: they are not cached anyway.
4290 Frequency : when mmap is used/binaries executed from Lustre
4292 Description: Unmmap pages before throwing them away from read cache.
4293 Details : llap_shrink cache now attempts to unmap pages before discarding
4294 them (if unmapping failed - do not discard). SLES9 kernel has
4295 extra checks that trigger if this unmapping is not done first.
4300 Description: lconf didn't resolve symlinks before checking to see whether a
4301 given mountpoint was already in use
4304 Frequency : when migrating failover services
4305 Bugzilla : 6395, 9514
4306 Description: When migrating a subset of services from a node (e.g. failback
4307 from a failover service node) the remaining services would
4308 time out and evict clients.
4309 Details : lconf --force (implied by --failover) sets the global obd_timeout
4310 to 5 seconds in order to quickly disconnect, but this caused
4311 other RPCs to time out too quickly. Do not change the global
4312 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4314 Severity : enhancement
4315 Frequency : if MDS is started with down OST
4316 Bugzilla : 9439,5706
4317 Description: Allow startup/shutdown of an MDS without depending on the
4318 availability of the OSTs.
4319 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4320 Add appropriate locking and lov-osc refcounts for safe
4321 cleaning. Add osc abort_inflight calls in case the
4322 synchronize never started.
4325 Frequency : occasional (Cray XT3 only)
4327 Description: root not authorized to access files in CRAY_PORTALS environment
4328 Details : The client process capabilities were not honoured on the MDS in
4329 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4330 been accessed by an authorized user then root was able to access
4331 the file on the local client also. The root user capabilities
4332 are now allowed on the MDS, as this environment has secure UID.
4335 Frequency : occasional
4337 Description: ldiskfs "too long searching" message happens too often
4338 Details : A debugging message (otherwise harmless) prints too often on
4339 the OST console. This has been reduced to only happen when
4340 there are fragmentation problems on the filesystem.
4345 Description: Division by zero in statfs when all OSCs are inactive
4346 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4347 lov_statfs divides by value returned by lov_get_stripecnt().
4351 Bugzilla : 9489, 3273
4352 Description: First write from each client to each OST was only 4kB in size,
4353 to initialize client writeback cache, which caused sub-optimal
4354 RPCs and poor layout on disk for the first writen file.
4355 Details : Clients now request an initial cache grant at (re)connect time
4356 and so that they can start streaming writes to the cache right
4357 away and always do full-sized RPCs if there is enough data.
4358 If the OST is rebooted the client also re-establishes its grant
4359 so that client cached writes will be honoured under the grant.
4364 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4365 Details : Now I/O RPCs go to different portal number and (presumably) fast
4366 lock requests (and glimses) and other RPCs get their own service
4367 threads pool that should be able to service those RPCs
4370 Severity : enhancement
4372 Description: Ability to exchange lustre version between client and servers and
4373 issue warnings at client side if client is too old. Also for
4374 liblustre clients there is ability to refuse connection of too old
4376 Details : New 'version' field is added to connect data structure that is
4377 filled with version info. That info is later checked by server and
4381 Frequency : rare, liblustre only.
4382 Bugzilla : 9296, 9581
4383 Description: Two simultaneous writes from liblustre at offset within same page
4384 might proceed at the same time overwriting eachother with stale
4386 Details : I/O lock withing llu_file_prwv was released too early, before data
4387 actually was hitting the wire. Extended lock-holding time until
4388 server acknowledges receiving data.
4391 Frequency : extremely rare. Never observed in practice.
4393 Description: avoid generating lustre_handle cookie of 0.
4394 Details : class_handle_hash() generates handle cookies by incrementing
4395 global counter, and can hit 0 occasionaly (this is unlikely, but
4396 not impossible, because initial value of cookie counter is
4397 selected randonly). Value of 0 is used as a sentinel meaning
4398 "unassigned handle" --- avoid it. Also coalesce two critical
4399 sections in this function into one.
4401 Severity : enhancement
4403 Description: allow liblustre clients to delegate truncate locking to OST
4404 Details : To avoid overhead of locking, liblustre client instructs OST to
4405 take extent lock in ost_punch() on client's behalf. New connection
4406 flag is added to handle backward compatibility.
4408 Severity : enhancement
4409 Bugzilla : 4928, 7341, 9758
4410 Description: allow number of OST service threads to be specified
4411 Details : a module parameter allows the number of OST service threads
4412 to be specified via "options ost ost_num_threads={N}" in the
4413 OSS's /etc/modules.conf or /etc/modprobe.conf.
4417 Bugzilla : 6146, 9635, 9895
4418 Description: servers crash with bad pointer in target_handle_connect()
4419 Details : In rare cases when a client is reconnecting it was possible that
4420 the connection request was the last reference for that export.
4421 We would temporarily drop the export reference and get a new
4422 one, but this may have been the last reference and the export
4423 was just destroyed. Get new reference before dropping old one.
4425 Severity : enhancement
4426 Frequency : if client is started with failover MDS
4428 Description: Allow multiple MDS hostnames in the mount command
4429 Details : Try to read the configuration from all specified MDS
4430 hostnames during a client mount in case the "primary"
4433 Severity : enhancement
4435 Description: Stop sending data to evicted clients as soon as possible.
4436 Details : Check if the client we are about to send or are sending data to
4437 was evicted already. (Check is done every second of waiting,
4438 for which l_wait_event interface was extended to allow checking
4439 of exit condition at specified intervals).
4442 Frequency : rare, normally only when NFS exporting is done from client
4444 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4446 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4449 Frequency : always, if srand() is called before liblustre initialization
4451 Description: Liblustre uses system PRNG disturbing its usage by user application
4452 Details : Introduce internal to lustre fast and high-quality PRNG for
4453 lustre usage and make liblustre and some other places in generic
4454 lustre code to use it.
4456 Severity : enhancement
4457 Bugzilla : 9477, 9557, 9870
4458 Description: Verify that the MDS configuration logs are updated when xml is
4459 Details : Check if the .xml configuration logs are newer than the config
4460 logs stored on the MDS and report an error if this is the case.
4461 Request --write-conf, or allow starting with --old_conf.
4463 Severity : enhancement
4465 Description: Handle symlinks in the path when checking if Lustre is mounted.
4466 Details : Resolve intermediate symlinks when checking if a client has
4467 mounted a filesystem to avoid duplicate client mounts.
4472 Description: lconf can hit an error exception but still return success.
4473 Details : The lconf command catches the Command error exception at the top
4474 level script context and will exit with the associated exit
4475 status, but doesn't ensure that this exit status is non-zero.
4480 Description: failure of ptlrpc thread startup can cause oops
4481 Details : Starting a ptlrpc service thread can fail if there are a large
4482 number of threads or the server memory is very fragmented.
4483 Handle this without oopsing.
4486 Frequency : always, only if liblustre and non-default acceptor port was used
4488 Description: liblustre cannot connect to servers with non-default acceptor port
4489 Details : tcpnal_set_default_params() was not called and was therefore
4490 ignoring the environment varaible TCPNAL_PORT, as well as other
4491 TCPNAL_ environment variables
4496 Description: two objects could be created on the same OST for a single file
4497 Details : If an OST is down, in some cases it was possible to create two
4498 objects on a single OST for a single file. No problems other
4499 than potential performance impact and spurious error messages.
4503 Bugzilla : 5681, 9562
4504 Description: Client may oops in ll_unhash_aliases
4505 Details : Client dcache may become inconsistent in race condition.
4506 In some cases "getcwd" can fail if the current directory is
4512 Description: Inode refcounting problems in NFS export code
4513 Details : link_raw functions used to call d_instantiate without obtaining
4514 extra inode reference first.
4518 Bugzilla : 9942, 9903
4519 Description: Referencing freed requests leading to crash, memleaks with NFS.
4520 Details : We used to require that call to ll_revalidate_it was always
4521 followed by ll_lookup_it. Also with revalidate_special() it is
4522 possible to call ll_revalidate_it() twice for the same dentry
4523 even if first occurence returned success. This fix changes semantic
4524 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4525 reference on a request referred from the intent.
4526 ll_intent_release() then releases such a request.
4529 Frequency : rare, normally benchmark loads only
4531 Description: unlinked inodes were kept in memory on the client
4532 Details : If a client is repeatedly creating and unlinking files it
4533 can accumulate a lot of stale inodes in the inode slab cache.
4534 If there is no other client load running this can cause the
4535 client node to run out of memory. Instead flush old inodes
4536 from client cache that have the same inode number as a new inode.
4539 Frequency : SLES9 2.6.5 kernel and long filenames only
4540 Bugzilla : 9969, 10379
4541 Description: utime reports stale NFS file handle
4542 Details : SLES9 uses out-of-dentry names in some cases, which confused
4543 the lustre dentry revalidation. Change it to always use the
4547 Frequency : rare, unless heavy write-truncate concurrency is continuous
4548 Bugzilla : 4180, 6984, 7171, 9963, 9331
4549 Description: OST becomes very slow and/or deadlocked during object unlink
4550 Details : filter_destroy() was holding onto the parent directory lock
4551 while truncating+unlinking objects. For very large objects this
4552 may block other threads for a long time and slow overall OST
4553 responsiveness. It may also be possible to get a lock ordering
4554 deadlock in this case, or run out of journal credits because of
4555 the combined truncate+unlink. Solution is to do object truncate
4556 first in one transaction without parent lock, and then do the
4557 final unlink in a new transaction with the parent lock. This
4558 reduces the lock hold time dramatically.
4561 Frequency : rare, 2.4 kernels only
4563 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4564 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4565 before unmounting their filesystem in order to avoid deadlock.
4566 The kernel_locked() function in 2.4 kernels only checks whether
4567 the kernel lock is held, not whether it is this process that is
4568 holding it as 2.6 kernels do.
4573 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4574 Details : The client ptlrpc code may be trying to reconnect to a down
4575 server before a previous connection attempt has timed out.
4576 Increase the reconnect interval to be longer than the connection
4577 timeout interval to avoid sending duplicate connections to
4581 Frequency : echo_client brw_test command
4583 Description: fix echo_client to work with OST preallocated code
4584 Details : OST preallocation code (5137) didn't take echo_client IO path
4585 into account: echo_client calls filter methods outside of any
4586 OST thread and, hence, there is no per-thread preallocated
4587 pages and buffers to use. Solution: hijack pga pages for IO. As
4588 a byproduct, this avoids unnecessary data copying.
4592 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4593 Description: Client can oops in mdc_commit_close() after open replay
4594 Details : It was possible for the MDS to return an open request with no
4595 transaction number in mds_finish_transno() if the client was
4596 evicted, but without actually returning an error. Clients
4597 would later try to replay that open and may trip an assertion
4598 Simplify the client close codepath, and always return an error
4599 from the MDS in case the open is not successful.
4602 Frequency : rare, 2.6 OSTs only
4604 Description: OST may deadlock under high load on fragmented files
4605 Details : If there was a heavy load and highly-fragmented OST filesystems
4606 it was possible to have all the OST threads deadlock waiting on
4607 allocation of biovecs, because the biovecs were not released
4608 until the entire RPC IO was completed. Instead, release biovecs
4609 as soon as they are complete to ensure forward IO progress.
4611 Severity : enhancement
4613 Description: Support for specifying external journal device at mount
4614 Details : If an OST or MDS device is formatted with an external journal
4615 device, this device major/minor is stored in the ext3 superblock
4616 and may not be valid for failover. Allow detecting and
4617 specifying the external journal at mount time.
4622 Description: Mounting an MDS with pending unlinked files may cause oops
4623 Details : target_finish_recovery() calls mds_postrecov() which returned
4624 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4625 considers this an error and immediately begins cleaning up the
4626 lov, just after starting the mds_lov process
4628 Severity : enhancement
4630 Description: Implement 'lfs df' to report actual free space on per-OST basis
4631 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4632 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4633 usage of objects; '-h' to report in human readable format.
4635 ------------------------------------------------------------------------------
4637 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4644 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4645 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4646 at startup during orphan unlinking. Instead, skip buddy update
4647 before setup, as it will be regenerated after recovery is complete.
4652 Description: If an OST is inactive, its locks might reference stale inodes.
4653 Details : lov_change_cbdata() must iterate over all namespaces, even if
4654 they are inactive to clear inode references from the lock.
4656 Severity : enhancement
4657 Frequency : occasional, if non-standard max_dirty_mb used
4659 Description: Client will block write RPCs if not enough grant
4660 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4661 then the client will block writes while waiting for another RPC
4662 to complete instead of consuming its dirty limit. With change
4663 we get improved performance when max_dirty_mb is small.
4665 Severity : enhancement
4666 Bugzilla : 3389, 6253
4667 Description: Add support for supplementary groups on the MDS.
4668 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4669 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4670 lookups for user supplementary groups into a cache.
4674 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4675 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4676 for liblustre O_EXCL is enforced by MDS.
4680 Description: readdir never returns NULL in liblustre.
4681 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4682 dentry in struct dirent.
4686 Frequency : liblustre only, depends on application IO pattern
4687 Description: liblustre clients evicted if not contacting servers
4688 Details : Don't put liblustre clients into the ping_evictor list, so
4689 they will not be evicted by the pinger ever.
4691 Severity : enhancement
4693 Description: Add ability to evict clients by NID from MDS.
4694 Details : By echoing "nid:$NID" string into
4695 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4696 $NID would be instantly evicted from this MDS and from all active
4697 OSTs connected to it.
4701 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4702 Details : lookup_it_finish() used to query file size from OSTs that was not
4707 Description: service threads change working directory to that of init
4708 Details : Starting lustre service threads may pin the working directory
4709 of the parent thread, making that filesystem busy. Threads
4710 now change to the working directory of init to avoid this.
4714 Frequency : during shutdown only
4715 Description: shutdown with a failed MDS or OST can cause unmount to hang
4716 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4721 Frequency : 2.6 only
4722 Description: chmod/chown may include an extra supplementary group
4723 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4724 supplementary group and if none is specified this is used.
4727 Bugzilla : 5479 (6816)
4729 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4730 Details : If lookup is in progress on a file that is unlinked we might try
4731 to revalidate the inode and fail in revalidate after lookup is
4732 complete and ll_file_open() enqueues the open again but
4733 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4736 Frequency : always, if lconf --abort_recovery used
4738 Description: lconf --abort_recovery fails with 'Operation not supported'
4739 Details : lconf was attempting to abort recovery on the MDT device and not
4742 ------------------------------------------------------------------------------
4744 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4749 Frequency : rare (only unsupported configurations with a node running as an
4751 Bugzilla : 6514, 5137
4752 Description: Mounting a Lustre file system on a node running as an OST could
4754 Details : OSTs now preallocates memory needed to write out data at
4755 startup, instead of when needed, to avoid having to
4756 allocate memory in possibly low memory situations.
4757 Specifically, if the file system is mounted on on OST,
4758 memory pressure could force it to try to write out data,
4759 which it needed to allocate memory to do. Due to the low
4760 memory, it would be unable to do so and the node would
4761 become unresponsive.
4763 Severity : enhancement
4765 Description: Addition of lconf --service command line option
4766 Details : lconf now accepts a '--service <arg>' option, which is
4767 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4769 Severity : enhancement
4771 Description: Failover mode is now the default for OSTs.
4772 Details : By default, OSTs will now run in failover mode. To return to
4773 the old behaviour, add '--failout' to the lmc line for OSTs.
4775 Severity : enhancement
4777 Description: Health checks are now provided for MDS and OSTs
4778 Details : Additional detailed health check information on MSD and OSTs
4779 is now provided through the procfs health_check value.
4782 Frequency : occasional, depends on IO load
4784 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4785 after numerous create/delete cycles
4786 Details : The ext3 inode allocation policy would not allocate new inodes
4787 very well on the OSTs because there are no new directories
4788 being created. Instead we look for groups with free space if
4789 the parent directories are nearly full.
4794 Description: Network or server problems during mount may cause partially
4795 mounted clients instead of returning an error.
4796 Details : The config llog parsing code may overwrite the error return
4797 code during mount error handling, returning success instead
4803 Description: MDS can fail to allocate large reply buffers
4804 Details : After long uptimes the MDS can fail to allocate large reply
4805 buffers (e.g. zconf client mount config records) due to memory
4806 fragmentation or consumption by the buffer cache. Preallocate
4807 some large reply buffers so that these replies can be sent even
4808 under memory pressure.
4812 Frequency : rare (liblustre)
4813 Description: fsx running with liblustre complained that using truncate() to
4814 extend the file doesn't work. This patch corrects that issue.
4815 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4816 ATTR_SIZE and lsm use in llu_setattr_raw.
4820 Frequency : rare, only 2.6 kernels
4821 Description: Unusual file access patterns on the MDS may result in inode
4822 data being lost in very rare circumstances.
4823 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4824 code in the 2.6 kernel caused the inode and/or EA data not to
4825 be read from disk, causing single-file corruption.
4829 Frequency : rare, only 2.6 filesystems using extents
4830 Description: Heavy concurrent write and delete load may cause data corruption.
4831 Details : It was possible under high-load situations to have an extent
4832 metadata block in the block device cache from a just-unlinked
4833 file overwrite a newly-allocated data block. We now unmap any
4834 metadata buffers that alias just-allocated data blocks.
4838 Frequency : filesystems with default stripe_count larger than 77
4839 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4840 Details : lconf specifies an inode size of 4096 bytes when the default
4841 stripe_count is larger than 77. This conflicts with the default
4842 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4843 this case to avoid pinning too much memory for large EAs.
4845 ------------------------------------------------------------------------------
4847 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4852 Frequency : rare (extremely heavy IO load with hundreds of clients)
4854 Description: Client is evicted, gets IO error writing to file
4855 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4856 caused clients to be evicted for AST timeouts. The fixes in
4857 bug 5192 mean we no longer need to have such short AST timeouts
4858 so ldlm_timeout has been increased.
4861 Frequency : occasional during --force or --failover shutdown under load
4862 Bugzilla : 5949, 4834
4863 Description: Server oops/LBUG if stopped with --force or --failover under load
4864 Details : a collection of import/export refcount and cleanup ordering
4865 issues fixed for safer force cleanup
4868 Frequency : only filesystems larger than 120 OSTs
4869 Bugzilla : 5990, 6223
4870 Description: lfs getstripe would oops on a very large filesystem
4871 Details : lov_getconfig used kfree on vmalloc'd memory
4874 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4875 Bugzilla : 6242, 6243
4876 Description: reading from files that had been truncated to a non-zero size
4877 but never opened returned no data
4878 Details : ll_file_read() reads zeros from no-object files to EOF
4883 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4884 Details : There's little harm in aborting MDS/OSS recovery and letting it
4885 try again, so I removed the LASSERT and return an error instead.
4887 Severity : enhancement
4889 Description: New debugging infrastructure for tracking down data corruption
4890 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4891 (b) cover more of the client-side code path, and (c) try to narrow
4892 down where problems occurred
4896 Bugzilla : 3819, 4364, 4397, 6313
4897 Description: Racing close and eviction MDS could cause assertion in mds_close
4898 Details : It was possible to get multiple mfd references during close and
4899 client eviction, leading to one thread referencing a freed mfd.
4901 Severity: : enhancement
4902 Bugzilla : 3262, 6359
4903 Description: Attempts to reconnect to servers are now more aggressive.
4904 Details : This builds on the enhanced upcall-less recovery that was added
4905 in 1.4.2. When trying to reconnect to servers, clients will
4906 now try each server in the failover group every 10 seconds. By
4907 default, clients would previously try one server every 25 seconds.
4912 Description: After recovery, certain operations trigger a failed
4913 assertion on a client.
4914 Details : Failing over an mds, using lconf -d --failover, while a
4915 client was doing a readdir() call would cause the client to
4916 LBUG after recovery completed and the readdir() was resent.
4918 Severity : enhancement
4920 Description: Default groups are now added by lconf
4921 Details : You can now run lconf --group <servicename> without having to
4922 manually add groups with lmc.
4925 Frequency : occasional
4927 Description: Nodes with an elan id of 0 trigger a failed assertion
4930 Frequency : always when accessing e.g. tty/console device nodes
4932 Description: tty and some other devices nodes cannot be used on lustre
4933 Details : file's private_data field is used by device data and lustre
4934 values in there got lost. New field was added to struct file to
4935 store fs-specific private data.
4938 Frequency : when exporting Lustre via NFS
4940 Description: NFSD failed occasionally when looking up a path component
4941 Details : NFSD is looking up ".." which was broken in ext3 directories
4942 that had grown large enough to become hashed.
4945 Frequency : Clusters with multiple interfaces not on the same subnet
4947 Description: Nodes will repeatedly try to reconnect to an interface which it
4948 cannot reach and report an error to the log.
4949 Details : Extra peer list entries will be created by lconf with some peers
4950 unreachable. lconf now validates the peer before adding it.
4953 Frequency : Only if a default stripe is set on the filesystem root.
4955 Description: Setting a default stripe on the filesystem root prevented the
4956 filesystem from being remounted.
4957 Details : The client was sending extra request flags in the root getattr
4958 request and did not allocate a reply buffer for the dir EA.
4961 Frequency : occasional, higher if lots of files are accessed by one client
4962 Bugzilla : 6159, 6097
4963 Description: Client trips assertion regarding lsm mismatch/magic
4964 Details : While revalidating inodes the VFS looks up inodes with ifind()
4965 and in rare cases can find an inode that is being freed.
4966 The ll_test_inode() code will free the lsm during ifind()
4967 when it finds an existing inode and then the VFS later attaches
4968 this free lsm to a new inode.
4972 Bugzilla : 6422, 7030
4973 Description: MDS deadlock between mkdir and client eviction
4974 Details : Creating a new file via mkdir or mknod (starting a transaction
4975 and getting the ns lock) can deadlock with client eviction
4976 (gets ns lock and trying to finish a synchronous transaction).
4979 Frequency : occasional
4980 Description: While starting a server, the fsfilt_ext3 module could not be
4982 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4983 kernels. Previously, lconf would still use the ext3 name
4984 when trying to load modules. Now, it will correctly use
4985 ext3 on 2.4 and ldiskfs on 2.6.
4987 Severity : enhancement
4988 Description: The default stripe count has been changed to 1
4989 Details : The interpretation of the default stripe count (0, to lfs
4990 or lmc) has been changed to mean striping across a single
4991 OST, rather than all available. For general usage we have
4992 found a stripe count of 1 or 2 works best.
4994 Severity : enhancement
4995 Description: Add support for compiling against Cray portals.
4996 Details : Conditional compiling for some areas that are different
5000 Frequency : occasional
5001 Bugzilla : 6409, 6834
5002 Description: Creating files with an explicit stripe count may lead to
5003 a failed assertion on the MDS
5004 Details : If some OSTs are full or unavailable, creating files may
5005 trigger a failed assertion on the MDS. Now, Lustre will
5006 try to use other servers or return an error to the
5010 Frequency : occasional
5012 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5013 caused lock timeout during readahead (since 1.4.2).
5014 Details : Processes doing readahead might match a lock that hasn't been
5015 granted yet if there are overlapping and conflicting lock
5016 requests. The readahead process waits on ungranted lock
5017 (original lock is CBPENDING), while OST waits for that process
5018 to cancel CBPENDING read lock and eventually evicts client.
5020 Severity : enhancement
5022 Description: Initial enabling of flock support for clients
5023 Details : Implements fcntl advisory locking and file status functions.
5024 This feature is provided as an optional mount flag (default
5025 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5026 locking are implemented yet, and those that are are not guaranteed
5027 to be completely correct in production environments.
5028 mount -t lustre -o [flock|noflock] ...
5031 Frequency : occasional
5033 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5034 assertion in the ext3 JBD (journaling) layer.
5035 Details : The b_committed_data struct is protected by the big kernel lock
5036 in 2.4 kernels, serializing journal_commit_transaction() and
5037 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5038 access to this struct. In 2.6 kernels there is finer grained
5039 locking to improve SMP performance of the JBD layer.
5043 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5048 Description: merge and cleanup kernel patches.
5051 -----------------------------------------------------------------------------
5053 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5055 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5056 versions of Lustre. Please update all servers and clients to
5057 version 1.4.2 or later at the same time. You must also run
5058 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5059 to update the configuration logs.
5061 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5062 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5063 - swab llog records of type '0' so we get proper header size/idx (5861)
5064 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5065 - fix rename of one directory over another leaking an inode (5953)
5066 - avoid SetPageDirty on 2.6 (5981)
5067 - don't re-add just-being-destroyed locks to the waiting list (5653)
5068 - when creating new directories, inherit the parent's custom
5069 striping settings if present parent (3048)
5070 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5071 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5072 - don't hold client locks on temporary worklist from l_lru (5666)
5073 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5074 - automatically evict dead clients (5921)
5075 - Update file size properly in create+truncate+fstat case (6196)
5076 - Do not unhash mountpoint dentries, do not allow removal of
5078 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5079 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5080 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5081 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5082 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5083 - Correctly handle reads of files with no objects (6243)
5084 - lctl recover will also mark a device active if deactivate used (5933)
5086 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5087 - allow --write-conf on an MDS with different nettype than client (5619)
5088 - don't write config llogs to MDS for mounts not from that MDS (5617)
5089 - lconf should create multiple TCP connections from a client (5201)
5090 - init scripts are now turned off by default; run chkconfig --on
5091 lustre and chkconfig --on lustrefs to use them
5092 - upcalls are no longer needed for clients to recover to failover
5094 - add --abort-recovery option to lconf to abort recovery on device
5096 - add support for an arbitrary number of OSTs (3026)
5097 - Quota support protocol changes.
5098 - forward compatibility changes to wire structs (6007)
5099 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5100 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5101 - improved "lustre status" script
5102 - initialize blocksize for non-regular files (6062)
5103 - added --disable-server and --disable-client configure options (5782)
5104 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5105 - Vanilla 2.4.29 support
5106 - increase maximum number of obd devices to 520 (6242)
5107 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5108 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5109 SLES 9 2.6.5 kernels
5110 - limit stripes per file to 160 (the maximum EA size) (6093)
5112 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5115 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5116 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5117 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5118 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5119 - small ext3 extents cleanups and fixes (5733)
5120 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5121 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5122 - avoid needless client->OST connect, fix handle mismatch (5317)
5123 - fix DLM error path that led to out-of-sync client, long delays (5779)
5124 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5125 - fix several locking issues related to i_size (5492,5624,5654,5672)
5126 - don't move pending lock onto export if it is already evicted (5683)
5127 - fix kernel oops when creating .foo in unlinked directory (5548)
5128 - fix deadlock in obdfilter statistics vs. object create (5811)
5129 - use time_{before,after} to avoid timer jiffies wrap (5882)
5130 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5131 - Do not leak request if server was not able to process it (5154)
5132 - If mds_open unable to find parent dir, make that negative lookup(5154)
5133 - don't create new directories with extent-mapping (5909, 5936)
5135 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5136 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5137 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5138 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5139 - support for RHEL4 kernel on i686 (5773)
5140 - provide error messages when incompatible logs are encountered (5898)
5142 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5143 * version 1.4.0.10 (1.4.1 release candidate 1)
5145 - don't keep a lock reference when lock is not granted (4238)
5146 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5147 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5148 - drop import inflight refcount on signal_completed_replay error (5255)
5149 - unlock page after async write error during send (3677)
5150 - handle missing objects in filter_preprw_read properly (5265)
5151 - no transno return for symlink open, don't save no-trasno open (3440)
5152 - don't try to complete elan receive that already failed (4012)
5153 - free RPC server reply state on error (5406)
5154 - clean up thread from ptlrpc_start_thread() on error (5160)
5155 - readahead could read extra page into cache that wasn't ejected (5388)
5156 - prevent races in class_attach/setup/cleanup/detach (5260)
5157 - don't dereference de->d_inode after l_dput of de (5458)
5158 - use "int" for stripe value returned from lock_to_stripe (5544)
5159 - mballoc allocation and error-checking fixes in 2.6 (5504)
5160 - block device patches to fix I/O request sizes in 2.6 (5482)
5161 - look up hostnames for IB nals (5602)
5162 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5163 - don't start multiple acceptors for the same port (5277)
5164 - fix incorrect LASSERT in mds_getattr_name (5635)
5165 - export a proc file for general "ping" checking (5628)
5166 - fix "lfs check" to not block when the MDS is down (5628)
5168 - service request history (4965)
5169 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5170 - create an "evict_client" /proc entry on OSTs, like the MDS has
5171 - fix mount usage message, return errors per mount(8) (5168)
5172 - change grep [] to grep "[]" in tests so they work in more UMLs
5173 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5174 - remove ancient LOV_MAGIC_V0 EA support (5047)
5175 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5176 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5177 - allow readahead allocations to fail when low on memory (5383)
5178 - mmap locking landed again, after considerable improvement (2828)
5179 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5181 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5184 - send OST transaction number in read/write reply to free req (4966)
5185 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5186 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5187 - the watchdog thread now runs as interruptible (5246)
5188 - flock/lockf fixes (but it's still disabled, pending 5135)
5189 - don't use EXT3 constants in llite code (5094)
5190 - memory shortage at startup could cause assertion (5176)
5192 - reorganization of lov code
5193 - single portals codebase
5195 - add extents/mballoc support (5025)
5196 - direct I/O reads in the obdfilter (4048)
5197 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5199 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5202 - send OST transaction number in read/write reply to free req (4966)
5203 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5204 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5205 - the watchdog thread now runs as interruptible (5246)
5206 - handle missing objects in filter_preprw_read properly (5265)
5207 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5208 - drop import inflight refcount on signal_completed_replay error (5255)
5209 - unlock page after async write error during send (3677)
5210 - return original error code on reconstructed replies (3761)
5211 - no transno return for symlink open, don't save no-trasno open (3440)
5213 - add pid to ldlm debugging output (4922)
5214 - bump the watchdog timeouts -- we can't handle 30sec yet
5215 - extra debugging for orphan dentry/inode bug (5259)
5217 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5220 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5221 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5222 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5223 - fix NULL dereference in /proc/sys/portals/routes (4827)
5224 - allow failed mdc_close() operations to be interrupted (4561)
5225 - stop precreate on OST before MDS would time out on it (4778)
5226 - don't send partial-page writes before EOF from client (4410)
5227 - discard client grant for sub-page writes on large-page clients (4520)
5228 - don't free dentries not owned by NFS code, check generation (4806)
5229 - fix lsm leak if mds_create_objects() fails (4801)
5230 - limit debug_daemon file size, always print CERROR messages (4789)
5231 - use transno after validating reply (3892)
5232 - process timed out requests if import state changes (3754)
5233 - update mtime on OST during writes, return in glimpse (4829)
5234 - add mkfsoptions to LDAP (4679)
5235 - use ->max_readahead method instead of zapping global ra (5039)
5236 - don't interrupt __l_wait_event() during strace
5238 - add software watchdogs to catch hung threads quickly (4941)
5239 - make lustrefs init script start after nfs is mounted
5240 - fix CWARN/ERROR duplication (4930)
5241 - return async write errors to application if possible (2248)
5242 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5243 - print NAL number in %x format (4645)
5244 - update barely-supported suse-2.4.21-171 series (4842)
5245 - support for sles 9 %post scripts
5246 - support for building 2.6 kernel-source packages
5247 - support for sles km_* packages
5249 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5252 - ignore -ENOENT errors in osc_destroy (3639)
5253 - notify osc create thread that OSC is being cleaned up (4600)
5254 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5255 - reconstruct ost_handle() like mds_handle() (4657)
5256 - create a new thread to do import eviction to avoid deadlock (3969)
5257 - let lconf resolve symlinked-to devices (4629)
5258 - don't unlink "objects" from directory with default EA (4554)
5259 - hold socknal file ref over connect in case target is down (4394)
5260 - allow more than 32000 subdirectories in a single directory (3244)
5261 - fix blocks count for O_DIRECT writes (3751)
5262 - OST returns ENOSPC from object create when no space left (4539)
5263 - don't send truncate RPC if file size isn't changing (4410)
5264 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5265 - bind to privileged port in socknal and tcpnal (3689)
5267 - rate limit CERROR/CWARN console message to avoid overload (4519)
5268 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5269 - basic mmap support (3918)
5270 - kernel patch series update from b1_4 (4711)
5272 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5275 - avoid crash during MDS cleanup with OST shut down (2775)
5276 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5277 - don't use bad inodes on the MDS (3744)
5278 - dynamic object preallocation to improve recovery speed (4236)
5279 - don't hold spinlock over lock dumping or change debug flags (4401)
5280 - don't zero obd_dev when it is force cleaned (3651)
5281 - print grants to console if they go negative (4431)
5282 - "lctl deactivate" will stop automatic recovery attempts (3406)
5283 - look for existing locks in ldlm_handle_enqueue() (3764)
5284 - don't resolve lock handle twice in recovery avoiding race (4401)
5285 - revalidate should check working dir is a directory (4134)
5287 - don't always mark "slow" obdfilter messages as errors (4418)
5289 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5292 - don't close LustreDB during write_conf until it is done (3860)
5293 - fix typo in lconf for_each_profile (3821)
5294 - allow dumping logs from multiple threads at one time (3820)
5295 - don't allow multiple threads in OSC recovery (3812)
5296 - fix debug_size parameters (3864)
5297 - fix mds_postrecov to initialize import for llog ctxt (3121)
5298 - replace config semaphore with spinlock (3306)
5299 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5300 - don't allow enqueue to complete on a destroyed export (3822)
5301 - down write_lock before checking llog header bitmap (3825)
5302 - recover from lock replay timeout (3764)
5303 - up llog sem before sending rpc (3652)
5304 - reduce ns lock hold times when setting kms (3267)
5305 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5306 - fix NULL deref and obd_dev leak on setup error (3312)
5307 - replace some LBUG about llog ops with error handling (3841)
5308 - don't match INVALID dentries from d_lookup and spin (3784)
5309 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5310 - fix invalid assertion in ptlrpc_set_wait (3880)
5312 - add libwrap support for the TCP acceptor (3996)
5313 - add /proc/sys/portals/routes for non-root route listing (3994)
5314 - allow setting MDS UUID in .xml (2580)
5315 - print the stack of a process that LBUGs (4228)
5317 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5320 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5321 - make sure to unset replay flag from failed open requests (3440)
5322 - if default stripe count is 0, use OST count for inode size (3636)
5323 - update parent mtime/ctime on client for create/unlink (2611)
5324 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5325 - free recovery state on server during a forced cleanup (3571)
5326 - unregister_reply for resent reqs (3063)
5327 - loop back devices mounting and status check on 2.6 (3563)
5328 - fix resource-creation race that can provoke i_size == 0 (3513)
5329 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5330 - more debugging for page-accounting assertion (3746)
5331 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5332 - avoid deadlock after precreation failure (3758)
5333 - fix race and lock order deadlock in orphan handling (3450, 3750)
5334 - add validity checks when grabbing inodes from l_ast_data (3599)
5336 - add /proc/.../recovery_status to obdfilter (3428)
5337 - lightweight CDEBUG infrastructure, debug daemon (3668)
5338 - change default OSC RPC parameters to be better on small clusters
5339 - turn off OST read cache for files smaller than 32MB
5340 - install man pages and include them in rpms (3100)
5341 - add new init script for (un)mounting lustre filesystems (2593)
5342 - run chkconfig in %post for init scripts (3701)
5343 - drop scimac NAL (unmaintained)
5345 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5348 - clean kiobufs before and after use (3485)
5349 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5350 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5351 - change init script to fail more clearly if not run as root (1528)
5352 - allow clients to reconnect during replay (1742)
5353 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5354 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5355 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5356 - fix cond_resched (3554)
5357 - only evict unfinished clients after recovery (3515)
5358 - allow bulk resend, prevent data loss (3570)
5359 - dynamic ptlrpc request buffer allocation (2102)
5360 - don't allow unlinking open directory if it isn't empty (2904)
5361 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5362 - remove extraneous obd dereference causing LASSERT failure (3334)
5363 - don't use get_cycles() when creating temp. files on the mds (3156)
5364 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5365 - handle EEXIST for set-stripe, set proper directory name (3336)
5367 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5368 - fix ksocknal_fmb_callback() error messages (2918)
5370 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5373 - don't copy lvb into (possibly NULL) reply on error (2983)
5374 - don't deref dentry after dput, don't free lvb on error (2922)
5375 - use the kms to determine writeback rpc length (2947)
5376 - increment oti_logcookies when osc is inactive (2948)
5377 - update client's i_blocks count via lvb messages (2543)
5378 - handle intent open/close of special files properly (1557)
5379 - mount MDS with errors=remount-ro, like obdfilter (2009)
5380 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5381 - don't use cancelling-locks' kms values (2947)
5382 - use highest lock extent for kms, not last one (2925)
5383 - don't dereference ERR_PTR() dentry in error handling path (3107)
5384 - fix thread race in portals_debug_dumplog() (3122)
5385 - create lprocfs device entries at setup instead of at attach (1519)
5386 - common AST error handler, don't evict client on completion race (3145)
5387 - zero nameidata in detach_mnt in 2.6 (3118)
5388 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5389 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5390 - initialize RPC timeout timer earlier for 2.6 (3219)
5391 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5392 - print nal/nid for unknown nid (3258)
5393 - additional checks for oscc recovery before doing precreate (3284)
5394 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5395 - don't crash in mdc_close for bad permissions on open (3285)
5396 - zero i_rdev for non-device files (3147)
5397 - clear page->private before handing to FS, better assertion (3119)
5398 - tune the read pipeline (3236)
5399 - fix incorrect decref of invalidated dentry (2350)
5400 - provide read-ahead stats and refine rpc in flight stats (3328)
5401 - don't hold journal transaction open across create RPC (3313)
5402 - update atime on MDS at close time (3265)
5403 - close LDAP connection when recovering to avoid server load (3315)
5404 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5405 - don't leak open file on MDS after open resend (3325)
5406 - serialize filter_precreate and filter_destroy_precreated (3329)
5407 - loop device shouldn't call sync_dev() for nul device (3092)
5408 - clear page cache after eviction (2766)
5409 - resynchronize MDS->OST in background (2824)
5410 - refuse to mount the same filesystem twice on same mountpoint (3394)
5411 - allow llmount to create routes for mounting behind routers (3320)
5412 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5413 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5414 - fix rare problem with rename on htree directories (3417)
5416 - allow default OST striping configuration per directory (1414)
5417 - fix compilation for qswnal for 2.6 kernels (3125)
5418 - increase maximum number of MDS request buffers for large systems
5419 - change liblustreapi to be useful for external progs like lfsck (3098)
5420 - increase local configuration timeout for slow disks (3353)
5421 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5423 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5426 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5427 - don't overwrite extent policy data in reply if lock was blocked (2901)
5428 - drop filter export grants atomically with removal from device (2663)
5429 - del obd_self_export from work_list in class_disconnect_exports (2908)
5430 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5431 - swab reply message in mdc_close, other PPC fixes (2464)
5432 - fix destroying of named logs (2325)
5433 - overwrite old logs when running lconf --write_conf (2264)
5434 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5435 - fix race in target_handle_connect (2898)
5436 - mds_reint_create() should take same inode create lock (2926)
5437 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5438 - don't close files for self_export to avoid uninitialized obd (2936)
5439 - allow MDS with the same name as client node (2939)
5440 - hold dentry reference for closed log files for unlink (2325)
5441 - reserve space for all logs during transactions (2059)
5442 - don't evict page beyond end of stripe extent (2925)
5443 - don't oops on a deleted current working directory (2399)
5444 - handle hard links to targets without a parent properly (2517)
5445 - don't dereference NULL lock when racing during eviction (2867)
5446 - don't grow lock extents when lots of conflicting locks (2919)
5448 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5451 - account for cache space usage on clients to avoid data loss (974)
5452 - lfsck support in lustre kernel code (2349)
5453 - reduce journal credits needed for BRW writes (2370)
5454 - orphan handling to avoid losing space on client/server crashes
5455 - ptlrpcd can be blocked, stopping ALL progress (2477)
5456 - use lock value blocks to assist in proper KMS, faster stat (1021)
5457 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5458 - recovery for initial connections (2355)
5459 - fixes for mds_cleanup_orphans (1934)
5460 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5461 - block all file creations until orphan recovery completes (1901)
5462 - client remove rq_connection from request struct (2423)
5463 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5464 - recovery timer race (2670)
5465 - mdc_close recovey bug (2532)
5466 - ptlrpc cleanup bug (2710)
5467 - mds timeout on local locks (2588)
5468 - namespace lock held during RPCs (2431)
5469 - handle interrupted sync write properly (2503)
5470 - don't try to handle a message that hasn't been replied to (2699)
5471 - client assert failure during cleanup after abort recovery (2701)
5472 - leak mdc device after failed mount (2712)
5473 - ptlrpc_check_set allows timedout requests to complete (2714)
5474 - wait for inflight reqs when ptlrpcd finishes (2710)
5475 - make sure unregistered services are removed from the srv_list
5476 - reset bulk XID's when resending them (caught by 1138 test)
5477 - unregister_bulk after timeout
5478 - fix lconf error (2694)
5479 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5480 - readahead locks pages, leaves pending causing memory pressure (2673)
5481 - increase OST request buffers to 4096 on large machines (2729)
5482 - fix up permission of existing directories in simple_mkdir (2661)
5483 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5484 - don't assign transno to errored transactions (2742)
5485 - don't delete objects on OST if given a bogus objid from MDS (2751)
5486 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5487 - if rq_no_resend, then timeout request after recovery (2432)
5488 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5489 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5490 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5491 - MDS recovery shouldn't delete objects that it has given out (2730)
5492 - if enqueue arrives after completion, don't clobber LVB (2819)
5493 - don't unlock pages twice when trigger_group_io returns error (2814)
5494 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5495 - don't write pages to disk if there was an error (1450)
5496 - don't ping imports that have recovery disabled (2676)
5497 - take buffered bytes into account when balancing socknal conn (2817)
5498 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5499 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5500 - remove little-endian swabbing of llog records (1987)
5501 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5502 - retry reposting request buffers if they fail (1191)
5503 - grow extent at grant time to avoid granting a revoked lock (2809)
5504 - lock revoke doesn't evict page if covered by a second lock (2765)
5505 - disable VM readahead to avoid reading outside lock extents (2805)
5507 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5508 - updated kernel patches for hp-2.4.20 kernel (2681)
5510 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5513 - fix truncated write corruption (2366)
5514 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5516 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5517 - ensure len(uuid) < 37 in lmc (1171)
5518 - fix ia64 OOPS in llog_test (2255)
5519 - zero end of page at obdfilter for partial page writes (2648)
5520 - don't leave stale dentries around after renames (bug 2428)
5521 - fix timeouts when evicting a client with a single lock held (2642)
5522 - set deadline for the initial HELLO message to drain (2634)
5523 - print out dotted-quad IP addresses in the socknal (2302)
5525 - additional debugging for MDS client eviction problem (2443)
5526 - fix mkfsoptions support for osts (2603, 2604)
5528 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5531 - add series for the vanilla 2.6.0 kernel
5532 - add series for the vanilla 2.4.24 kernel
5533 - add series for a cray x86/64 UL kernel drop
5534 - fix xattr patches for the vanilla 2.4.19 series
5536 - generate true UUIDs in lmc (1171)
5537 - have portals stack dumping break in UML (2466)
5538 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5539 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5540 - wake callback waiting threads on client eviction (2460)
5541 - Add --ptldebug and --subsystem to lmc (1719)
5542 - update assertion to allow safe interrupt allocation
5543 - set rq_no_resend for cancel requests (2432)
5544 - recalculate ptlrpcd timeout after resend (2494)
5545 - call vfs_rmdir when removing pending directories (2368)
5546 - fix renaming a file to itself (2429)
5547 - lmc creates a default one-stripe lov (2454)
5548 - expand procfs space to handle large clusters (2326)
5549 - increase UML stack to avoid overflow
5550 - update lconf's list of debug and subsystem masks
5551 - fix lfs find --obd (2510)
5552 - /proc tunable for disabling filter read caching (2591)
5553 - stop rpm packages from altering slapd.conf (2301)
5554 - disable nagle in the socknal under 0conf (2578)
5555 - choose mds inode size based on stripe count (2572)
5556 - fix kernel-source rpm problems (2516)
5558 - add --disable-doc to avoid pdf generation (2421)
5559 - update documentation, tests, type-os, comments
5560 - avoid format warnings on ia64
5561 - remove the TOE NAL
5562 - tiny code cleanups by removing unused fields
5564 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5567 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5568 - remove the most blatant lies from BUILDING file (2371)
5569 - change default debug level to reasonable production setting
5570 - reduce client side cache size to reduce cache flush time
5571 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5572 - make TCP zerocopy and pinger support enabled by default (2476)
5573 - sync writes completed after process exits caused crashes (2319)
5574 - maintain correct mount count on the MDS (2356)
5575 - backout 1557, because 2316 wasn't really fixed
5576 - better file I/O statistics gathering in /proc
5577 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5578 - another kernel patch to fix zero-copy TCP function export
5579 - don't take duplicate lock when processing re-sent getattr (2420)
5580 - lctl uses obd_self_export instead of creating new conn (2353)
5581 - MDS/OST recovery case which requires object creation asserted (2425)
5582 - move lfs from /usr/sbin to /usr/bin in packages
5583 - fix race between mds_client_add and mds_client_free (2417)
5584 - use kmalloc instead of slabs in portals (2430)
5585 - don't create duplicate records when a failover MDS is present (2442)
5586 - remove unnecessary mount age check (2332)
5587 - don't remove directory inodes from locks prematurely (2451)
5588 - don't break if MDS service name is the same as hostname (2103)
5589 - fix races in client write RPC generation when cache full (2482)
5591 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5594 - remove now-unused request->rq_obd (278)
5595 - if an allocation fails, print out how much memory we've used (1933)
5596 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5597 - add the "configurable stack size" patch to most series files (1256)
5598 - ability to write large log records, for 100+ OST configs (2306)
5599 - fix NULL deref when filter_prep fails (2314)
5600 - fix operator precedence error in filter_sync
5601 - dynamic allocation of socknal TX descriptors (2315)
5602 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5603 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5604 - ensure that CWARN messages in Portals always get to the syslog
5605 - __init/__exit are not for prototype decls (ldlm_init/exit)
5606 - x86-64 compile warning fixes
5607 - fix gateway LMC keyword conflict (2318)
5608 - fix MDS lock inversions in getattr/reint paths (1844)
5609 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5610 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5611 - DLM race condition prevented some lock evictions (2328)
5612 - ENOMEM detection and retry on socknal sends (2230)
5613 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5614 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5615 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5616 - better error messages when a client is rejected during recovery (1505)
5617 - avoid cancelling locks which were never granted, after failure (2330)
5618 - fix i_sem/journal inversion in mds_client_add (2333)
5619 - fix truncate/getattr lock cycle deadlock (2334)
5620 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5621 - fix two rare exit paths which could leak an l_lock() ref (2321)
5622 - fencepost error in MDS/OST orphan recovery (2226)
5623 - make log record alignment 8 bytes (1988)
5624 - lstripe now fails when requested offset > ost_count (2237)
5625 - ensure that all kernel series have a complete list.h (1607)
5626 - fix crashes in special-file operations (2316)
5627 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5629 - allow configurable automake binary, for testing new versions
5630 - small update to the lfs documentation
5632 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5634 * fix negative export reference count in fsfilt_sync (2312)
5636 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5637 * release candidate 0.9.1
5639 - orphans are moved into the PENDING directory for possible recovery
5640 - replayed opens now open by fid for orphan/rename safety (1042)
5641 - last close of an orphan inode generates a transno (683)
5642 - chdir() and mount() now pin the directory entry (1020)
5643 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5644 - discard very old requests without processing them (1502)
5645 - remove some common, well-understood CERRORs (1505)
5646 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5647 - clear "grant" flags in OST replies until OST grant code lands (1644)
5648 - fix read performance by not clobbering i_blksize on client (1598)
5649 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5650 - make LRU size a /proc tunable, clears locks when reduced (707)
5651 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5652 - server locks take references on exports now (1558)
5653 - build fixes for 2.4.20-rh trees (1663)
5654 - return an error from lov_create if all OSCs are inactive (1751)
5655 - fix import levels when a reconnect happens without a timeout (1597)
5656 - exit early from mds_open if we get a lookup error (1749)
5657 - partial page read at EOF wouldn't wait for disk before sending (1642)
5658 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5659 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5660 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5661 - write/truncate lock inversion (1639)
5662 - Don't auto-load obdclass, portals modules during cleanup (1495)
5663 - fix timestamps from jumping to "now" (1763)
5664 - extra journal assertions (1648)
5665 - add an extra multiunlink test (1771)
5666 - fix read_record/write_record API (1776)
5667 - fix leak of offset_extent, possible incorrect i_size later (1772)
5668 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5669 - Debugging for the kqswnal_get_idle_tx problems (1820)
5670 - Allow recovery to be attempted multiple times (1536)
5671 - Write out MDS last_rcvd file after it is first created (1600)
5672 - Fix tx_descriptor leak in failed transmit situations (1827)
5673 - ext3 journaling fixes for assertion failure after IO error (1871)
5674 - class_export_put() on freed export after completion AST error (1896)
5675 - Fix revalidate looping in VFS (1322)
5676 - Don't access a freed export during MDS_REINT timeout (1521)
5677 - Add open-unlink recovery support on the MDS (1673,1764)
5678 - Return an error if no MDS data was read from last_rcvd (1946)
5679 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5680 - Don't setup a disk device that doesn't match exported UUID (317)
5681 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5682 - avoid committing NULL handle in force close
5683 - local.sh is now a one-stripe LOV configuration
5684 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5685 - fix ext3 htree duplicate directory entry corruption (1516)
5686 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5687 - update to new LOV EA format (2097)
5688 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5690 2003-06-15 Phil Schwan <phil@clusterfs.com>
5693 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5694 - per-import/export recovery handling (958, 931, 959)
5695 - multiple last-rcvd slots, for serving multiple FSes (949)
5696 - connections are again shared between multiple imp/exports (963, 964)
5697 - "umount -f" would hang if any requests needed to be sent (393, 978)
5698 - avoid pinning large req buffer by copying for queued messages (989)
5699 - add "uuid" to "lctl device" command to help upcalls (991)
5700 - "open" RPCs with transnos would confuse recovery counters (1037)
5701 - do proper endian conversion of all wire messages (288, 340, 891)
5702 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5703 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5704 - fix ptlrpc_connection leak in target_handle_connect (1174)
5705 - fix import refcounting bug in OST and MDS cleanup (1134)
5706 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5707 - fix very unlikely obd_types race condition (501)
5708 - remove osc_open hack for echo_client (1187)
5709 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5710 - a failure in read_inode2 leads to deadlock (1139)
5711 - cancel ack-locks as soon as transaction is committed (1072)
5712 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5713 - make sure to commitrw after any preprw to avoid deadlock (1162)
5714 - failing to execute a file in a lustre FS would lock inode (1203)
5715 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5716 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5717 - obd_uuid2tgt would walk off the end of the list (1255)
5718 - on IA64 the osc would give portals incorrect bulk size (1258)
5719 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5720 - fix lock inversion caused by new llite matching code (1282)
5721 - limit the number of dirty pages on a client to 10MB (1286)
5722 - timed out locks were not being corrected cancelled (1289)
5723 - fix O_DIRECT above 4GB on IA-32 (1292)
5724 * major user-visible changes
5725 - fail out/fail over policy now controlled by the upcall (993)
5727 - add OBD_PING to check server availability and failure (954)
5728 - lustre messages are now sent in sending host order (288, 340, 891)
5729 - add eadatalen to MDS getattr reply (340)
5730 - OST read replies may contain second buffer, with per-page status (593)
5732 2003-03-11 Phil Schwan <phil@clusterfs.com>
5735 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5736 - failed open()s could cause deadlock; fixed (867, 869)
5737 - stop cancelling OST locks when files are closed (481)
5738 - overlapping XID spaces caused network corruption (851, 853)
5739 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5740 - setattr_raw added, to do single-RPC, server-side setattrs
5741 - lmc/lconf syntax change for OST UUIDs
5742 - fix crashy race condition between ptlrpc_free_req and osc_close
5743 - don't use request in mdc_enqueue if we hit a timeout (889)
5744 - don't set the inode i_size for regular files from the MDS (896)
5745 - handle out of order completion AST (842)
5746 - don't LBUG if a lock request times out after receiving AST (913)
5747 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5748 - if a bad lock AST arrives, send an error instead of dropping entirely
5749 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5750 - fix leak in bulk IO when only partially completed (899, 900, 926)
5751 - fix O_DIRECT for ia64 (55)
5752 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5753 - C-z after timeout could hang a process forever; fixed (977)
5755 - client-side I/O cache (678, 924, 929, 941, 970)
5757 - READPAGE and SETATTRs which don't take server-side locks get
5760 2003-02-11 Phil Schwan <phil@clusterfs.com>
5763 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5764 - Fix fsfilt_extN_readpage() to read a full page of directory
5765 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5766 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5767 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5768 - Most intent operations were removed, in favour of a new RPC mode
5769 that does a single RPC to the server and bypasses most of the VFS
5770 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5771 - Aggressively cancel local locks on DLM servers
5772 - mds_reint_unlink sends EA to the client if it's the last nlink.
5773 client uses that EA to unlink OST objects.
5774 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5775 - recursive symlinks were fixed (439)
5776 - fixed NULL deref in DEBUG_REQ
5777 - filter_update_lastobjid no longer calls sync, which annoyed extN
5778 - fixed multi-client small-writes to a single file problem (445)
5779 - fixed mtime updates during file writes (607)
5780 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5781 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5782 - fixed Linux OST size reporting problem (444, 656)
5783 - OST now updates object mtime with writes or setattr (607, 619)
5784 - client verifies file size before zeroing page past EOF (445)
5785 - OST now writes last allocated objid to disk with allocation (108)
5786 - LOV on echo now works (409)
5788 - mds_reint_unlink sends a new buffer, with the EA included. this
5789 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5790 set if a regular file was being unlinked, and it was the last link
5791 - use PtlGet from the target for bulk writes (315)
5792 - OST now updates object mtime with writes or setattr (607, 619)
5793 - LDLM now has a grant-time callback to revalidate locked items, if
5795 - Many MDS operations were reorganized to combat race conditions
5797 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5798 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5800 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5803 - Fully reactivate OST imports after reconnection (512, others)
5804 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5805 - More graceful error handling for truncating on dead OST (515)
5806 - Don't error out unless we're actually accessing dead stripes (474)
5807 - Fix garbage sizes when stripes are missing (410)
5808 - LRU counters were broken, causing constant lock purge (433, 432)
5809 - garbage on read from stripes with failed OSTs (441)
5810 - mark OSCs as active before reconnecting during recovery (438)
5811 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5812 - lfind did not preserve OST order in output (443)
5813 - symlinks cause hung clients, incorrect data (439)
5814 - stop dereferencing request after dropping refcount (457)
5815 - don't LASSERT(spin_is_locked) on non-SMP (455)
5816 - fixes for many rename() bugs
5817 - fstat didn't correctly synchronize attributes (399)
5818 - server must handle lock cancellation during blocking AST prep (487)
5819 - bulk descriptors were free()d too soon (511)
5820 - fix paths in lconf, which would load incorrect modules (451, 507)
5821 - fix confusing lconf 'host not found' error message (386)
5822 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5823 - fix race condition in mdc_blocking_ast() for inode access (526)
5824 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5825 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5826 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5827 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5828 - fix lstripe in several ways (kernel side) (527)
5829 - fix request leak in ldlm_cli_enqueue (262)
5830 - incorrect OSC was marked inactive after OST failure
5831 - call mds_fs_cleanup before unmounting filesystem (524)
5832 - fix races between taking ns_lock and ldlm_lock_change_resource
5833 - fix races updating LOV export open file list
5834 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5835 - fix recovery NULL deref in ldlm_cli_cancel_unused
5836 - fix some DLM races by using new hash table for lock handles (419)
5837 - permit the client to specify desired inodes, at replay
5838 - duplicate requests when we queue them for replay reintegration
5839 - fix last_rcvd offset calculation
5840 - sync after each recovered transaction, so we always make progress
5841 - never, not always, ERESTART requests without transnos
5842 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5844 - skip replay if the MDS says that the client is already connected
5845 - don't check for a recovery-enabled export to match lctl's UUID
5846 - don't INC_USE_COUNT for phantom exports
5847 - don't crash when cleaning up phantom exports (567)
5848 - don't double-finish or set replay data for errored mdc_open requests
5849 - abort requests when they time out, so we don't get old replies
5850 - send/receive replies for AST messages again
5851 - if the client says that it doesn't have the lock, cancel it on the
5853 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5854 mark it as destroyed, it will all work out when decref is called
5855 - fix module use counts (22, 581)
5857 - ASTs now expect a reply (server cancels lock on error reply)
5859 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5862 - fix many simultaneous client startup (392)
5863 - fix dentry->d_it clobbering
5864 - credentials weren't being shipped for readdir/getattr operations
5865 - remove invalid assertions triggered during some concurrent MD
5867 - proper Lustre versions added (336, 389)
5868 - fix memory leak for create error case (398)
5869 - fix LOV locking bug that would get cli/srv out of sync
5870 - fix echo client over LOV (409)
5871 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5872 - fix double-O_EXCL intent crash (424)
5873 - avoid sending multiple lock CANCELs (352)
5875 - MDS can do multi-client recovery (modulo bugs in new code)
5877 - many updates, edits, cleanups
5879 2002-11-18 Phil Schwan <phil@clusterfs.com>
5882 - fix null d_it dereference (346)
5883 - fix full OST/dbench hang (333)
5884 - fix permission problem with file removal (286)
5885 - fix removal of OSCs from LOV when they fail
5886 - fix NULL deref during bulk timeout (214)
5887 - fix problems related to multiple filesystems on one MDS (241)
5888 - fixed serious subtle metadata locking bugs
5889 - free locks on clients when inodes are removed due to memory
5891 - fix inode pointer in lock data (285)
5892 - partial support for multiple MDS on a single host (241)
5893 - data locks weren't cancelled at clear_inode time (290, 311)
5894 - intent locks could lead to unbounded lock growth (205)
5895 - added a maximum lock count, an LRU list, and a flusher
5896 - fix multiple rename (365)
5897 - properly abstracted the echo client
5898 - OSC locked 1 byte too many; fixed
5899 - rewrote brw callback code:
5900 - fixed recovery bugs related to LOVs (306)
5901 - fixed too-many-pages-in-one-write crash (191)
5902 - fixed (again) crash in sync_io_timeout (214)
5903 - probably fixed callback-related race (385)
5905 - Add capability to MDS protocol
5906 - LDLM cancellations and callbacks on different portals
5908 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5911 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5912 - defer open object destruction to close time (601981, 138)
5913 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5914 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5915 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5916 - don't call obd_setattr in ll_file_release for destroyed objects
5917 * protocol change to lustre_msg: move |version| and add |flags|
5918 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5919 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5920 * added replay of create, unlink, link and rename operations during
5921 MDS failover; recovery should be much more robust now
5922 * remove failed OSCs from LOVs (only lov_create uses this so far)
5923 * the lustre-HOWTO was brought (more) up to date (582544)
5925 2002-10-23 Phil Schwan <phil@clusterfs.com>
5928 - in-use dentries weren't being reused properly (617851)
5929 - prevent multiple LDLM setup (599178)
5930 - fix LOV size calculations for truncate (617853)
5931 - fix client handling of MDS intent errors (POSIX)
5932 - fix permission bug in lovstripe.c test (624321)
5933 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5934 - truncate past end of file could corrupt data
5935 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5936 - a race in recovery could return ETIMEDOUT to apps (623947)
5937 - building outside the source directory was fixed
5938 * the lustre-HOWTO was brought (more) up to date (582544)
5939 * major progress was made on recovery functionality
5941 2002-10-10 Phil Schwan <phil@clusterfs.com>
5944 - recovery deadlock fix
5945 - rm -rf causes LBUG fix (617817)
5946 - file open by multiple tasks fix (618962)
5947 - directory permissions bugs (602707 and 620007)
5948 - journal_stop fixed with locking (611313)
5949 - O_APPEND failures resolved (618273, perhaps 614459)
5950 - lconf PATH fix (619770)
5951 - IA64 build fix (621450)
5952 - RPC buffer sizes scale with amount of memory
5954 2002-10-01 Phil Schwan <phil@clusterfs.com>
5957 - locks would be cancelled without throwing away data pages,
5958 resulting in inconsistent data (605627)
5959 - inode attributes were not always being refreshed (605627, 612449)
5960 - lconf now continues to cleanup after lctl reports an error
5961 - MDS now enforces user permissions (602707)
5962 - lprocfs cleanup fixed, but not yet enabled (614157)
5963 - fixed infinite server hang, should a client not respond to an AST
5964 - avoid going into recovery if user calls readlink() with a buffer
5965 that's too small (613941)
5966 - AST RPCs no longer require replies (614867) -- this may be changed
5967 - don't crash server if client sends an IOV that's too big (611336)
5968 - fixed lock conversion deadlock (611892)
5969 - fixed the following of symlinks (614622)
5970 * recovery: the server can remove locks from a client that dies, other
5971 clients can make progress
5972 * more extN patch fixes
5973 * compile-time configurable ptlrpc buffer allocations
5975 - collaborative read cache document
5976 - Lustre Lite Performance CDR document-in-progress
5978 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5981 - fix typo in patch-2.4.18
5983 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5986 - clear ptlrpc request each time in handle_incoming_request()
5987 - unlink of files now destroys the object on the OST
5989 2002-09-19 Peter Braam <braam@clusterfs.com>
5991 * add hard link support
5992 * change obdfile creation method
5993 * kernel patch changed
5995 2002-09-19 Peter Braam <braam@clusterfs.com>
5998 - stack overflow bug in extN fixed
6000 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6002 * documentation updates
6003 - add man pages for config tools
6004 - update tests/README to describe testing with new config tools
6005 - finish metadata API descriptions
6006 * bug fixes and cleanups
6007 - statfs workaround for 16TB limit
6008 - LOV stripe allocation improved, can stripe on subset of OSTs
6009 - LOV file size/IO offset was wrong for files > 4GB in size
6010 - object EA data was being dropped, caused files to be unreadable
6011 - memory overflow with non-LOV OST caused memory corruption
6012 - fixed regression tests to work with new config tools, obdfilter
6013 - fixed bug when directory size became larger than 1 block
6014 - fixed bug (for single client case) when PWD was deleted
6015 - invalidate local directory pages when doing intent-based ops
6016 - avoid LDLM oops when lock callback contained bad data
6018 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6020 * documentation updates
6021 * bug fixes and cleanups
6022 - configuration tools
6025 - 64-bit compile warnings
6026 - 64-bit internal statfs data
6028 * test_brw on persistent OST devices
6030 * lprocfs (disabled)
6032 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6034 * documentation updates
6035 * bug fixes and cleanups
6036 * configuration tools
6038 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6041 * many small fixes to 0_5_4
6042 * io/network handling
6043 * thinkos in MDS operations
6045 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6048 * crucial basic fixes to 0.5.3
6049 * IOR, Iozone work over Elan
6052 2002-08-07 Phil Schwan <phil@clusterfs.com>
6053 * version 0_5_3, our first alpha
6054 * we use the new Portals iovs
6055 * documentation updates
6056 * bug fixes and cleanups
6057 * small changes in the DLM wire protocol
6059 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6060 * version 0_5_1 with some initial stability,
6061 * locking on MD and file I/O.
6062 * documentation updates
6063 * several bug fixes since 0.5.0
6064 * small changes in wire protocol
6066 2002-07-18 Phil Schwan <phil@clusterfs.com>
6068 * delivered as Lustre Light Alpha
6069 * fixed a crash after handling invalid MDS requests
6070 * fixed directory pages for architectures with non-4k pages sizes
6072 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6073 * release version v0_4_4
6074 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6076 * Fixes a number of interruption problems with OST operations.
6077 * Update documentation for portals header changes
6078 * Move all wire protocol structs/defines to lustre_idl.h
6079 * Fixes symlink length bug.
6080 * Add tcpdump to repository.
6082 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6083 * release version v0_4_3
6084 * Fixes statfs for inodes on extN.
6085 * Fixes bug in runtests which would delete /etc/hosts.
6086 * Use 64-bit object IDs wherever possible (not into VFS though)
6087 Remove ost_get_info, which is unused by lustre, and out of date.
6089 2002-07-03 Peter Braam <braam@clusterfs.com>
6090 * release version v0_4_2 Fixes a lookup error (type not passed)
6091 * move forward to head of Portals
6092 * move forward to latest Lustre kernel
6094 2002-06-25 Peter Braam <braam@clusterfs.com>
6095 * release version v0_4_1. Hopefully stable on single node use.