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
52 Severity : enhancement
54 Description: Adjust locks' extents on their first enqueue, so that at the time
55 they get granted, there is no need for another pass through the
56 queues since they are already shaped into the proper forms.
60 Description: Fix in ptlrpc_expire_one_request() to print the signed time
63 Severity : enhancement
65 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
68 Severity : enhancement
70 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
72 Severity : enhancement
74 Description: correctly handle big reply message.
75 Details : send LNet event if reply is bigger then buffer and adjust this buffer
80 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
82 Severity : enhancement
84 Description: Add support for OEL5.
88 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
92 Description: client could not reconnect to OST because of an active request.
93 Details : The client now retries to reconnect to the same server, if a
94 connect request failed with EBUSY or -EAGAIN. Test case 35b
95 conf-sanity.sh has been created.
99 Description: Don't allow make backward step on assignin osc next id.
100 Details : race between allocation next id and ll_sync thread can be cause
101 of set wrong osc next id and can be kill valid ost objects.
105 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
109 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
113 Description: Error handling in osc_statfs_interpret() has been improved.
114 Details : Check in osc_statfs_interpret() for EBADR.
118 Description: Do not send statfs() requests to OSTs disabled by administrator.
119 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
121 Severity : enhancement
123 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
125 Severity : enhancement
127 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
129 Severity : enhancement
131 Description: Enhancement for lfs(1) command to use numeric uid/gid.
135 Description: Improved cleanup procedure for conf-sanity test_46a.
139 Description: Do not start bulk transfer, if client tried to reconnect.
140 Details : Do not start bulk transfer, if client tried to reconnect,
141 but failed due to active rpcs. Check for exp_abort_active_req
142 has been implemented.
146 Description: Prevent inconsistences between linux and lustre mount structures.
147 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
148 Make the sleep interruptible.
150 Severity : enhancement
152 Description: provide server to client comms path
153 Details : server to client communications path via new LDLM_SET_INFO rpc
157 Description: truncate starts GFP_FS allocation under transaction causing deadlock
158 Details : ldiskfs_truncate calls grab_cache_page which may start page
159 allocation under an open transaction. This may lead to
160 calling prune_icache with consequent lustre reentrance.
164 Description: send correctly lsm on open replay
165 Details : MDS is trust to LSM size on replay open, but client can set wrong size
170 Description: enable client interface failover
171 Details : When a child reconnects from another NID, properly update export
172 nid hash position and ldlm reverse import.
174 Severity : enhancement
176 Description: Communicate OST degraded/readonly state via statfs to MDS
177 Details : Flags in the statfs returned from OSTs indicate whether the
178 OST is in a degraded RAID state, or if the filesystem has
179 turned read-only after a filesystem error is detected.
184 Description: don't panic if EPROTO was hit when reading symlink
185 Details : correctly handling request reference in error cases.
187 Severity : enhancement
189 Description: Add LustreNetLink, a kernel-userspace communcation path.
191 Severity : enhancement
193 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
198 Description: access to llog context before init.
199 Details : move handling CATALOGS file at osc layer and forbid access to llog
203 Frequency : rare, only when client is writing to OST on local node
205 Description: Avoid deadlock for local client writes
206 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
207 memory freeing context. This allows OST threads to set the
208 PF_MEMALLOC flag on task structures in order to allocate memory
209 from reserved pools and complete IO.
210 Use GFP_HIGHUSER for OST allocations for non-local client writes,
211 so that the OST threads generate memory pressure and allow
212 inactive pages to be reclaimed.
214 Severity : enhancement
216 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
221 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
222 Details : move ldlm namespace creation in setup phase to avoid grab
223 _lprocfs_lock with cli_sem held.
227 Description: Temporarily disable grant shrink.
228 Details : Disable the feature for debugging.
231 Frequency : only when formatting test filesystems
233 Description: Unable to run several mkfs.lustre on loop devices at the same time
234 Details : mkfs.lustre returns error 256 on the concurrent loop devices
235 formatting. The solution is to proper handle the error.
237 Severity : enhancement
239 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
241 Severity : enhancement
243 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
248 Description: resolve race between obd_disconnect and class_disconnect_exports
249 Details : if obd_disconnect will be called to already disconnected export he
250 forget release one reference and osc module can't unloaded.
252 Severity : enhancement
254 Description: Allow tuning service thread via /proc
255 Details : For each service a new
256 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
257 created that can be used to set min/max thread counts, and get the
258 current number of running threads.
263 Descriptoin: don't return error if have partially created objects for file.
264 Details : lov_update_create_set uses set->set_success as index for created
265 objects, so if some requests will be failed, they will have hole
266 at end of array and qos_shrink_lsm for allocate correct lsm.
268 Severity : enhancement
270 Description: Update OFED support to 1.4.1
275 Description: Reduce small size read RPC
276 Details : Set read-ahead limit for every file and only do read-ahead when
277 available read-ahead pages are bigger than 1M to avoid small size
280 Severity : enhancement
282 Description: move AT tunable parameters for more consistent usage
283 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
285 Severity : enhancement
287 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
288 Details : allow skip disconnected ost for send statfs request and hide error
294 Description: fix racy locking of mballoc block bitmaps causing BUG
295 Details : The locking of the mballoc buddy bitmap and the in-memory
296 block bitmap was using two different spin locks in some
297 cases. This made it possible to incorrectly access the
298 mballoc bitmap while another process was modifying it,
299 causing a sanity assertion to fail. While no on-disk corruption
300 was reported, there was some risk of this happening.
303 Frequency : rare, on llog test 6
305 Descriptoin: don't allow connect to already connected import
306 Details : allowing connect to already connected import is hide connecting problem.
311 Descriptoin: allow kill process which wait statahead result
312 Details : for some reasons 'ls' can stick in waiting result from statahead,
313 in this case need way for kill this process.
315 Severity : enhancement
317 Description: Add state history info file, enhance import info file
318 Details : Track import connection state changes in a new osc/mdc proc file;
319 add overview-type data to the osc/mdc import proc file.
321 Severity : enhancement
323 Description: MDS create should not wait for statfs RPC while holding DLM lock.
325 Severity : enhancement
327 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
332 Descriptoin: OST grant too much space to client even there are not enough space.
333 Details : Client will shrink its grant cache to OST if there are no write
334 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
335 this grant cache if there are already not enough avaible space
336 (left_space < total_clients * 32M).
339 Frequency : start MDS on uncleanly shutdowned MDS device
341 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
342 Details : stay in waiting mds<>ost recovery finished produce random bugs
343 due race between two ll_sync thread for one lov target. send
344 ACTIVATE event only if connect realy finished and import have
348 Frequency : rare, connect and disconnect target at same time
350 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
351 Details : don't call obd_disconnect under lov_lock. this long time
352 operation and can block ptlrpcd which answer to connect request.
357 Descriptoin: don't lose wakeup for imp_recovery_waitq
358 Details : recover_import_no_retry or invalidate_import and import_close can
359 both sleep on imp_recovery_waitq, but we was send only one wakeup
363 Frequency : always with long access acl
365 Descriptoin: mds can't pack reply with long acl.
366 Details : mds don't control size of acl but they limited by reint/getattr
369 Severity : enhancement
371 Description: Update to SLES10 kernel-2.6.16.60-0.33.
373 Severity : enhancement
375 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
378 Frequency : start MDS on uncleanly shutdowned MDS device
380 Descriptoin: aborting recovery hang on MDS
381 Details : don't throttle destroy RPCs for the MDT.
384 Frequency : on remount
386 Description: external journal device not working after the remount
387 Details : clear dev_rdonly flag for external journal devices in
393 Description: shutdown vs evict race
394 Details : client_disconnect_export vs connect request race.
395 if client will evicted at this time - we start invalidate
396 thread without referece to import and import can be freed
400 Frequency : rare, need acl's on inode.
402 Description: client can't handle ost additional correctly
403 Details : if ost was added after client connected to mds client can have
404 hit lnet_try_match_md ... to big messages to wide striped files.
405 in this case need teach client to handle config events about add
406 lov target and update client max ea size at that event.
408 Severity : enhancement
410 Description: Changelogs
411 Details : Changelogs are a lightweight mechanism to track filesystem
412 metadata and namespace changes. The changelog is recorded
413 permanently on the MDTs, and is periodically "consumed" / purged
414 when records are no longer needed.
416 Severity : enhancement
418 Description: compact fld format with extents
419 Details : Store range of seq rather than every seq in FLD. Seq
420 controller update FLD rather than clients. In Case of CMD, mdt0
421 has FLD, all other metadata server act as non persistent proxy
422 for FLD queries and cache fld entries in fld cache.
427 Description: don't skip ost target if they assigned to file
428 Details : Drop slow OSCs if we can, but not for requested start idx.
429 This means "if OSC is slow and it is not the requested
430 start OST, then it can be skipped, otherwise skip it only
431 if it is inactive/recovering/out-of-space.
435 Description: more cleanup in mds_lov
436 Details : not send LOV EA under replay, we can't know about they size at this
437 time. Don't allow client connect to mds before any ost connected,
438 for avoid problems with LOV EA size and returning EIO to client.
440 Severity : enhancement
442 Description: Interoperability at server side (Disk interoperability)
444 Severity : enhancement
446 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
448 Severity : enhancement
450 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
452 Severity : enhancement
454 Description: New client IO stack (CLIO).
456 Severity : enhancement
458 Description: Commit on sharing. Eliminate inter-client dependencies between
459 uncommitted transactions by doing transaction commits.
460 Thereby clients may recovery independently.
463 Frequency : Create a symlink file with a very long name
465 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
466 Details : If there is no extra space in the request for early cancels,
467 ldlm_req_handles_avail() returns 0 instead of a negative value.
469 Severity : enhancement
471 Description: Add /proc entry for import status
472 Details : The mdc, osc, and mgc import directories now have
473 an import directory that contains useful import data for debugging
476 Severity : enhancement
478 Description: Re-disable certain /proc logging
479 Details : Enable and disable client's offset_stats, extents_stats and
480 extents_stats_per_process stats logging on the fly.
483 Frequency : Only on FC kernels 2.6.22+
485 Description: oops in statahead
486 Details : Do not drop reference count for the dentry from VFS when lookup,
487 VFS will do that by itself.
489 Severity : enhancement
491 Description: Generic /proc file permissions
492 Details : Set /Proc file permissions in a more generic way to enable non-
493 root users operate on some /proc files.
497 Description: Hitting mdc_commit_close() ASSERTION
498 Details : Properly handle request reference release in
499 ll_release_openhandle().
503 Description: quota recovery deadlock during mds failover
504 Details : This patch includes att18982, att18236, att18237 in bz14840.
506 1. fix osts hang when mds does failover with quotaon
507 2. prevent watchdog storm when osts threads wait for the
512 Frequency : only patchless client
513 Description: add workaround for race between add/remove dentry from hash
515 Severity : enhancement
517 Description: Allow OST glimpses to return PW locks
521 Description: LBUG when llog conf file is full
522 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
526 Description: Prevent import from entering FULL state when server in recovery
530 Description: service mount cannot take device name with ":"
531 Details : Only when device name contains ":/" will mount treat it as
537 Description: replace ptlrpcd with the statahead thread to interpret the async
538 statahead RPC callback
542 Frequency : on recovery
543 Description: I/O failures after umount during fail back
544 Details : if client reconnected to restarted server we need join to recovery
545 instead of find server handler is changed and process self
546 eviction with cancel all locks.
548 Severity : enhancement
550 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
552 Severity : enhancement
554 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
556 Severity : enhancement
558 Description: Upcall on Lustre log has been dumped
559 Details : Allow for a user mode script to be called once a Lustre log has
560 been dumped. It passes the filename of the dumped log to the
561 script, the location of the script can be specified via
562 /proc/sys/lnet/debug_log_upcall.
567 Description: avoid idr_remove called for id which is not allocated.
568 Details : Move assigment s_dev for clustered nfs to end of initialization,
569 to avoid problem with error handling.
574 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
575 Details : When connection is reused this not moved from CONN_UNUSED_HASH
576 into CONN_USED_HASH and this prodice warning when put connection
577 again in unused hash.
580 Severity : enhancement
582 Description: File striping can now be set to use an arbitrary pool of OSTs.
584 Severity : enhancement
586 Description: Export bytes_read/bytes_write count on OSC/OST.
590 Description: Early reply size mismatch, MGC loses connection
591 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
592 the connect flags are properly negotiated.
597 Description: quotas are not honored with O_DIRECT
598 Details : all writes with the flag O_DIRECT will use grants which leads to
599 this problem. Now using OBD_BRW_SYNC to guard this.
603 Description: add quota statistics
604 Details : 1. sort out quota proc entries and proc code.
605 2. add quota statistics
607 Severity : enhancement
609 Description: enable quota support for HEAD.
613 Description: Properly propagate oinfo flags from lov to osc for statfs
614 Details : restore missing copy oi_flags to lov requests.
616 Severity : enhancement
618 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
619 : l_getgroups(8), lst(8), routerstat(8)
620 Details : included man pages for llobdstat(8), llstat(8),
621 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
623 Severity : enhancement
625 Description: configure's --enable-quota should check the
626 : kernel .config for CONFIG_QUOTA
627 Details : configure is terminated if --enable-quota is passed but
628 : no quota support is in kernel
632 Description: Remove portals compatibility
633 Details : Remove portals compatibility, not interoperable with releases
638 Description: Resolve device initialization race
639 Details : Prevent proc handler from accessing devices added to the
640 obd_devs array but yet be intialized.
642 Severity : enhancement
644 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
646 Severity : enhancement
648 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
653 Description: Using wrong pointer in osc_brw_prep_request
654 Details : Access to array[-1] can produce panic if kernel compiled with
655 CONFIG_PAGE_ALLOC enabled
659 Description: Client runs out of low memory
660 Details : Consider only lowmem when counting initial number of llap pages
664 Description: *optional* service tags registration
665 Details : if the "service tags" package is installed on a Lustre node
666 When the filesystem is mounted, a local-node service tag will
667 be created. See http://inventory.sun.com/ for more information
668 about the Service Tags asset management system.
672 Description: Kernel BUG tries to release flock
673 Details : Lustre does not destroy flock lock before last reference goes
674 away. So always drop flock locks when client is evicted and
675 perform unlock regardless of successfulness of speaking to MDS.
679 Description: add refcount protection for osc callbacks, avoid panic on shutdown
683 Description: sanity test 65a fails if stripecount of -1 is set
684 Details : handle -1 striping on filesystem in ll_dirstripe_verify
689 Description: ASSERTION(CheckWriteback(page,cmd)) failed
690 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
693 Severity : enhancement
695 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
699 Description: do not process already freed flock
700 Details : flock can possibly be freed by another thread before it reaches
701 to ldlm_flock_completion_ast.
705 Description: LBUG during stress test
706 Details : Need properly lock accesses the flock deadlock detection list.
710 Description: oops in page fault handler
711 Details : kernel page fault handler can return two special 'pages' in error
712 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
716 Description: timeout with invalidate import.
717 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
718 should be handled by ptlrpcd. This produce long age waiting and
719 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
721 Severity : enhancement
723 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
727 Description: enable MGS and MDT services start separately
728 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
729 a co-located MGS without starting the MGS, which is a complement
730 to 'nosvc' mount option.
734 Frequency : after recovery
735 Description: precreate to many object's after del orphan.
736 Details : del orphan st in oscc last_id == next_id and this triger growing
737 count of precreated objects. Set flag LOW to skip increase count
738 of precreated objects.
742 Frequency : rare, on clear nid stats
743 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
744 Details : when clean nid stats sometimes try destroy live entry,
745 and this produce panic in free.
749 Description: Stack overflow during MDS log replay
750 ease stack pressure by using a thread dealing llog_process.
754 Description: wait until IO finished before start new when do lock cancel.
755 Details : VM protocol want old IO finished before start new, in this case
756 need wait until PG_writeback is cleared until check dirty flag and
757 call writepages in lock cancel callback.
759 Severity : enhancement
761 Description: using special macro for print time and cleanup in includes.
765 Description: mds_mfd_close() ASSERTION(rc == 0)
766 Details : In mds_mfd_close(), we need protect inode's writecount change
767 within its orphan write semaphore to prevent possible races.
771 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
775 Frequency : rare, on shutdown ost
776 Description: don't hit live lock with umount ost.
777 Details : shrink_dcache_parent can be in long loop with destroy dentries,
778 use shrink_dcache_sb instead.
782 Description: don't panic with use echo client
783 Details : echo client pass NULL as client nid pointer and this produce null
788 Description: fix build on ppc32
789 Details : compile code with -m64 flag produce wrong object file for ppc32.
793 Description: add message levels for liblustreapi
797 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
798 Details : Move the 'good_osts' check before the 'total_bavail' check. This
799 will result in an -EAGAIN and in the exit call path we call
800 alloc_rr() which will with increasing aggressiveness attempt to
801 aquire precreated objects on the minimum number of required OSCs.
805 Description: Use old size assignment to avoid deadlock
806 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
807 the scheduling while holding a spinlock. We do not need locking
808 for size in ll_update_inode() because size is only updated from
809 the MDS for directories or files without objects, so there is no
810 other place to do the update, and concurrent access to such inodes
811 are protected by the inode lock.
815 Description: resolve "_IOWR redefined" build error on SLES10
819 Description: dump the memory debugging after all modules are unloaded to
820 suppress false negative in conf_sanity test 39
822 Severity : enhancement
824 Description: build kernel-ib packages for OFED 1.3 in our release cycle
829 Description: fix SLES kernel versioning
830 Details : the kernel version for our SLES 10 kernel did not include a "-"
831 before the "smp" at the end. while this was not a problem in
832 general, it did mean that software trying to use the kernel
833 version to try to detect a vendor specific kernel would fail.
834 this was most evident by the OFED build scripts.
838 Description: Don't update lov_desc members until making sure they are valid
839 Details : When updating lov_desc members via proc fs, need fix their
840 validities before doing the real update.
844 Description: don't put request into delay list while invalidate in flight.
845 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
846 import in flight. this produce timeout for invalidate and sometimes
847 can cause stale data.
851 Frequency : on ppc only
852 Description: not convert ost objects for directory because it's not exist.
853 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
857 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
859 Severity : enhancement
861 Description: Update to support 2.6.22.14 vanilla kernel.
865 Frequency : rare, on recovery
866 Description: read procfs can produce deadlock in some situation
867 Details : Holding lprocfs lock which send rpc can produce block for destroy
868 obd objects and this also block reconnect with -EALREADY. This
869 isn't fix all lprocfs bugs - but make it rare.
871 Severity : enhancement
873 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
876 Frequency : frequent on X2 node
878 Description: mdc_set_open_replay_data LBUG
879 Details : Set replay data for requests that are eligible for replay.
883 Description: lustre_mgs: operation 101 on unconnected MGS
884 Details : When MGC is disconnected from MGS long enough, MGS will evict the
885 MGC, and late on MGC cannot successfully connect to MGS and a lot
886 of the error messages complaining that MGS is not connected.
890 Frequency : on network error
891 Description: panic with double free request if network error
892 Details : mdc_finish_enqueue is finish request if any network error ocuring,
893 but it's true only for synchronus enqueue, for async enqueue
894 (via ptlrpcd) this incorrect and ptlrpcd want finish request
897 Severity : enhancement
899 Description: client-side metadata stat-ahead during readdir(directory readahead)
900 Details : perform client-side metadata stat-ahead when the client detects
901 readdir and sequential stat of dir entries therein
904 Frequency : on start mds
906 Description: Implement get_info(last_id) in obdfilter.
909 Frequency : occasional
911 Description: Correctly check stale fid, not start epoch if ost not support SOM
912 Details : open with flag O_CREATE need set old fid in op_fid3 because
913 op_fid2 was overwritten with new generated fid, but mds can answer
914 with one of these two fids and both is not stale. Setattr
915 incorrectly started an epoch and assume will be called
916 done_writing, but without SOM done_writing ever being called.
919 Frequency : rare, depends on device drivers and load
921 Description: MDS or OSS nodes crash due to stack overflow
922 Details : Code changes in 1.8.0 increased the stack usage of some functions.
923 In some cases, in conjunction with device drivers that use a lot
924 of stack the MDS (or possibly OSS) service threads could overflow
925 the stack. One change which was identified to consume additional
926 stack has been reworked to avoid the extra stack usage.
929 Frequency : occasional
931 Description: Do not fail import if osc_interpret_create gets -EAGAIN
932 Details : If osc_interpret_create got -EAGAIN it immediately exits and
933 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
934 oscc_has_objects and see OSC has no objests and call
935 oscc_internal_create to resend create request.
937 Severity : enhancement
939 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
941 Severity : enhancement
943 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
946 Frequency : very rare
948 Description: Fix replay if there is an un-replied request and open
949 Details : In some cases, older replay request will revert the
950 mcd->mcd_last_xid on MDS which is used to record the client's
953 Severity : enhancement
955 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
957 Severity : enhancement
959 Description: Add rhel5 support to HEAD.
961 Serverity : enhancement
963 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
968 Description: Don't allow skipping OSTs if index has been specified.
969 Details : Don't allow skipping OSTs if index has been specified, make
970 locking in internal create lots better.
974 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
975 Details : access to bitfield in structure is always rounded to long
976 and this produce problem with not atomic change any bit.
980 Description: Lustre make rpms failed.
981 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
982 builds Lustre rpms from tarball.
985 Frequency : rare on shutdown ost
987 Description: If llog cancel was not send before clean_exports phase, this can
988 produce deadlock in llog code.
989 Details : If llog thread has last reference to obd and call class_import_put
990 this produce deadlock because llog_cleanup_commit_master wait when
991 last llog_commit_thread exited, but this never success because was
992 called from llog_commit_thread.
996 Description: allow userland application know is lost one of stripes.
997 Details : fill lvb_blocks with error code on ost and return it to
998 application if error flag found.
1002 Description: NULL lov_tgts causing MDS oops
1003 Details : more safe checks for NULL lov_tgts for avoid oops.
1005 Severity : enhancement
1007 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1011 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1012 Details : If an OST is down the MDS will hang indefinitely in
1013 obd_statfs_rqset() waiting for the statfs data. While for
1014 MDS QOS usage of statfs, it should not stuck in waiting.
1016 Severity : enhancement
1018 Description: remote_acl support
1019 Details : Support ACL-based permission check for remote user.
1020 Support setfacl/getfacl for remote user with the utils
1021 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1022 the system "{s,g}etfacl" utils.
1024 Severity : enhancement
1026 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1028 Severity : enhancement
1030 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1034 Description: make mgs_setparam() handle fsname containing dash
1035 Details : fsname containing a dash does not work with lctl conf_param
1037 Severity : enhancement
1039 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1041 Severity : enhancement
1043 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1047 Frequency : rare, at shutdown
1048 Description: access already free / zero obd_namespace.
1049 Details : if client_disconnect_export was called without force flag set,
1050 and exist connect request in flight, this can produce access to
1051 NULL pointer (or already free pointer) when connect_interpret
1052 store ocd flags in obd_namespace.
1056 Frequency : only at startup
1057 Description: not alloc memory with spinlock held.
1058 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1059 if any spinlock held.
1061 Severity : enhancement
1063 Description: make lustre randomly fail allocating memory
1064 Details : Make lustre randomly failed allocating memory for testing purpose.
1066 Severity : enhancement
1068 Description: lost problems with lov objid file
1069 Details : Fixes some scability and access to not inited memory problems
1070 in work with lov objdid file.
1075 Description: lfs find does not continue on file error
1076 Details : Continue other files processing when a file/dir is absent.
1080 Description: Inconsistent usage of lustre_pack_reply()
1081 Details : Standardize the usage of lustre_pack_reply() such that it
1082 always generate a CERROR on failure.
1085 Frequency : occasional
1087 Description: MDS hang or stay in waiting lock
1088 Details : If client receive lock with CBPENDING flag ldlm need send lock
1089 cancel as separate rpc, to avoid situation when cancel request
1090 can't processed due all i/o threads stay in wait lock.
1094 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1098 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1102 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1106 Description: Update to RHEL4 latest kernel.
1108 Severity : enhancement
1110 Description: Build SLES10 patchless client fails
1111 Details : The configure was broken by run ./configure with
1112 --with-linux-obj=.... argument for patchless client. When the
1113 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1114 can't search header adequately. Use absolute path such as
1115 -I($LINUX)/include instead.
1119 Description: interrupt oig_wait produce painc on resend.
1120 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1121 private set, and this produce situation when rq_ptlrpcd_data not
1122 copyed to new allocated request and triggered LBUG on assert
1123 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1124 ptlrpcd set if request is changed and can be safety changed to use
1129 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1130 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1131 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1132 these macros and replace them with their difinitions instead.
1136 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1137 Details : Update which_patch & target file for SLES10 latest kernel.
1140 Description: Remove client patches from SLES 10 kernel.
1141 Details : This causes SLES 10 clients to behave as patchless clients
1142 even on a Lustre-patched (server) kernel.
1144 Severity : enhancement
1146 Description: self-adjustable client's lru lists
1147 Details : use adaptive algorithm for managing client cached locks lru
1148 lists according to current server load, other client's work
1149 pattern, memory activities, etc. Both, server and client
1150 side namespaces provide number of proc tunables for controlling
1153 Severity : enhancement
1155 Description: light-weight GSS support
1156 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1157 and reduce performance overhead.
1159 Severity : enhancement
1161 Description: Linux keyring support
1162 Details : Support using service of Linux keyring for Lustre GSS internal
1163 context refresh/cache mechanism.
1167 Description: Fix errors in lfs documentation
1168 Details : Fixes man pages
1172 Description: don't use GFP_* in generic Lustre code.
1173 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1177 Description: obdclass is limited by single OBD_ALLOC(idarray)
1178 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1182 Description: warnings with build patchless client with vanila 2.6.19 and up
1183 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1184 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1188 Frequency : Only for SLES
1189 Description: sanity_quota fail test_1
1190 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1191 file, which makes TSTID[2] inunique.
1195 Frequency : Only for SLES9
1196 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1197 Details : Update target/ChangeLog/which_patch .
1201 Frequency : Only for test_10 in sanity-quota.sh
1202 Description: a bug in quota test code
1203 Details : checking a test flag in a wrong place
1207 Frequency : Only for test_13 in sanity-quota.sh
1208 Description: a bug in quota test script
1209 Details : shouldn't check overfull allocated quota
1213 Description: jbd statistics
1214 Details : Port older jbd statistics patch for sles10
1216 Severity : enhancement
1218 Description: Add printing inode info into message about error in writepage.
1223 Description: reply_lock_interpret crash due to race with it and lock cancel.
1224 Details : Do not replay locks that are being cancelled. Do not reference
1225 locks by their address during replay, just by their handle.
1229 Frequency : When flocks are used.
1230 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1231 Details : Flock locks might destroy just granted lock if it could be merged
1232 with another existing flock, this is done in completion handler,
1233 so teach ldlm_cli_enquque_fini that this is a valid case for
1239 Description: Oops in read and write path when failing to allocate lock.
1240 Details : Check if lock allocation failed and return error back.
1244 Description: lstripe command fails for valid OST index
1245 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1246 instead of lov->desc.ld_active_tgt_count.
1250 Description: sanity.sh failed test 103
1251 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1256 Description: df doesn't work properly if diskfs blocksize != 4K
1257 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1261 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1262 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1264 Severity : enhancement
1266 Description: Adaptive timeouts
1267 Details : RPC timeouts adapt to changing server load and network
1268 conditions to reduce resend attempts and improve recovery time.
1272 Description: llapi_file_create() does not allow some changes
1273 Details : add llapi_file_open() that allows specifying the mode and
1274 open flags, and also returns an open file handle.
1278 Description: merge and cleanup kernel patches.
1279 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1283 Description: Add journal checksum support.(Kernel part)
1284 Details : The journal checksum feature adds two new flags i.e
1285 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1286 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1287 indicates that the commit block contains the checksum for
1288 the blocks described by the descriptor blocks. Now commit
1289 record can be sent to disk without waiting for descriptor
1290 blocks to be written to disk. This behavior is controlled
1291 using JBD2_FEATURE_ASYNC_COMMIT flag.
1295 Description: OSS needs mutliple precreate threads
1296 Details : Add ability to start more than one create thread per OSS.
1300 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1301 Details : execute lfs setstripe on client
1305 Description: mds_obd_create error creating tmp object
1306 Details : When the user sets quota on root, llog will be affected and can't
1307 create files and write files.
1310 Frequency : Always on ia64 patchless client, and possibly others.
1312 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1313 Details : This allows the patchless client to be loaded on architectures
1314 without this export.
1318 Description: RedHat Update kernel for RHEL5
1319 Details : Add the kernel config file for RHEL5.
1323 Description: RedHat Update kernel for RHEL5
1324 Details : Modify the kernel config file more closer RHEL5.
1328 Description: Build failure against Centos5 (RHEL5)
1329 Details : Define PAGE_SIZE when it isn't present.
1333 Description: test_7 on sanity-quota.sh can't be used on separate machines
1334 Details : add facet to handle it
1338 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1339 Details : with NFS, the anon dentry's parent was set to itself in
1340 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1341 corresponding dentry other than use rec->ur_name.
1343 Severity : enhancement
1345 Description: lfs setstripe enhancement
1346 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1350 Description: enable data checksumming by default
1351 Details : enable checksum by default, allow --disable-checksum
1352 configure option and "-o nochecksum" mount option. Checksums
1353 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1357 Description: lustre support for RHEL5
1358 Details : Add support for RHEL5.
1362 Description: Client eviction due to failover config
1363 Details : after a connection loss, the lustre client should attempt to
1364 reconnect to the last active server first before trying the
1365 other potential connections.
1369 Description: when mds and osts use different quota unit(32bit and 64bit),
1370 quota will be released repeatly.
1371 Details : void sending multiple quota reqs to mds, which will keep the
1372 status between the reqs.
1376 Description: osts not allocated evenly to files
1377 Details : change the condition to increase offset_idx
1380 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1382 Description: Data corruption for OSTs that are formatted larger than 2TB
1384 Details : When generating the bio request for lustre file writes the
1385 sector number would overflow a temporary variable before being
1386 used for the IO. The data reads correctly from Lustre (which
1387 will overflow in a similar manner) but other file data or
1388 filesystem metadata may be corrupted in some cases.
1392 Description: Tune the kernel for good SCSI performance.
1393 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1394 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1399 Description: rewrite ext2-derived code in obdclass/uuid.c
1400 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1401 from scratch preserving functionality.
1405 Description: "lfs find -obd UUID" prints directories
1406 Details : "lfs find -obd UUID" will return all directory names instead
1407 of just file names. It is incorrect because the directories
1408 do not reside on the OSTs.
1412 Description: use i_size_read and i_size_write in 2.6 port
1413 Details : replace inode->i_size access with i_size_read/write()
1416 Frequency : when removing large files
1418 Description: scheduling issue during removal of large Lustre files
1419 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1420 It causes scheduling issues when removing large files (17TB in the
1424 Frequency : only with liblustre clients on XT3
1426 Description: evictions taking too long
1427 Details : allow llrd to evict clients directly on OSTs
1430 Frequency : only on ppc
1432 Description: /proc/fs/lustre/devices broken on ppc
1433 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1434 We should make sure the type of 'index' is loff_t and then cast
1435 explicitly as needed below. Do not assign an explicitly cast
1439 Frequency : only for rhel5
1441 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1442 Details : Modify the target file & which_kernel.
1444 Severity : enhancement
1446 Description: omit set fsid for export NFS
1447 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1448 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1452 Frequency : Always, for kernels after 2.6.16
1453 Description: Fix warning idr_remove called for id=.. which is not allocated.
1454 Details : Last kernels save old s_dev before kill super and not allow
1455 to restore from callback - restore it before call kill_anon_super.
1459 Description: buffer overruns could theoretically occur
1460 Details : llapi_semantic_traverse() modifies the "path" argument by
1461 appending values to the end of the origin string, and a
1462 overrun may occur. Adding buffer overrun check in liblustreapi.
1466 Description: Fix error on 'ls .' at the top of the Lustre mount.
1467 Details : Don't revalidate dentry if it is a root dentry.
1471 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1472 Details : Modify vm-tunables-rhel4.patch.
1476 Description: Kernel config for 2.6.18-vanilla.
1477 Details : Modify targets/2.6-vanilla.target.in.
1478 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1479 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1480 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1481 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1485 Description: improve handling recoverable errors
1486 Details : if request processig with error which can be recoverable on server
1487 request should be resend, otherwise page released from cache and
1492 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1493 Details : Modify target file & which_patch.
1494 A flaw was found in the IA32 system call emulation provided
1495 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1496 value could be stored in the %RAX register, which could trigger an
1497 out-of-bounds system call table access. An untrusted local user
1498 could exploit this flaw to run code in the kernel
1499 (ie a root privilege escalation). (CVE-2007-4573).
1503 Description: O_DIRECT bypasses client statistics.
1504 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1505 read_ahead_stats were updated. copied the stats section from
1506 osc_send_oap_rpc() into async_internal().
1510 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1514 Description: change order of libsysio includes
1515 Details : '#include sysio.h' should always come before '#include xtio.h'
1519 Description: handle "serious error: objid * already exists" more gracefully
1520 Details : If LAST_ID value on disk is smaller than the objects existing in
1521 the O/0/d* directories, it indicates disk corruption and causes an
1522 LBUG(). If the object is 0-length, then we should use the existing
1523 object. This will help to avoid a full fsck in most cases.
1525 Severity : enhancement
1527 Description: adapt the lustre_config script to support the upgrade case
1528 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1529 server targets to 1.6 in parallel.
1533 Description: To avoid grant space > avaible space when the disk is almost
1534 full. Without this patch you might see the error "grant XXXX >
1535 available" or some LBUG about grant, when the disk is almost
1537 Details : In filter_check_grant, for non_grant cache write, we should
1538 check the left space by if (*left > ungranted + bytes), instead
1539 of (*left > ungranted), because only we are sure the left space
1540 is enough for another "bytes", then the ungrant space should be
1541 increase. In client, we should update cl_avail_grant only there
1542 is OBD_MD_FLGRANT in the reply.
1546 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1547 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1548 described in CVE-2007-4573. This problem could allow untrusted
1549 local users to gain root access.
1552 Frequency : when using O_DIRECT and quotas
1554 Description: Incorrect file ownership on O_DIRECT output files
1555 Details : block usage reported by 'lfs quota' does not take into account
1556 files that have been written with O_DIRECT.
1561 Description: touch file failed when fs is not full
1562 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1563 otherwise we can get ENOSP while fs is not full.
1567 Description: parallel lock callbacks
1568 Details : Instead of sending blocking and completion callbacks as separated
1569 requests, adding them to a set and sending in parallel.
1572 Frequency : only for Cray XT3
1573 Bugzilla : 12829/13455
1574 Description: Changing primary group doesn't change the group lustre assigns to
1576 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1577 overridden with the primary group provided by the group upcall,
1578 whereas the supplied fsgid can be trusted if it is in the list of
1579 supplementary groups returned by the group upcall.
1581 Severity : enhancement
1583 Description: Allow masking D_WARNING, D_ERROR messages from console
1584 Details : Console messages can now be disabled via lnet.printk.
1588 Description: User code with malformed file open parameter crashes client node
1589 Details : Before packing join_file req, all the related reference should be
1590 checked carefully in case some malformed flags cause fake
1591 join_file req on client.
1595 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1596 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1597 mode is equal to the granted mode. However, it should also check
1598 that LDLM_FL_CP_REQD is not set.
1602 Description: Heavy nfs access might result in deadlocks
1603 Details : After ELC code landed, it is now improper to enqueue any mds
1604 locks under och_sem, because enqueue might want to decide to
1605 cancel open locks for same inode we are holding och_sem for.
1609 Description: Client eviction while running blogbench
1610 Details : A lot of unlink operations with concurrent I/O can lead to a
1611 deadlock causing evictions. To address the problem, the number of
1612 oustanding OST_DESTROY requests is now throttled to
1613 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1614 callbacks are processed in priority.
1618 Description: enable ACLs on MDS by default
1619 Details : ACLs must be enabled on MDS by default.
1622 Frequency : PPC/PPC64 only
1624 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1625 Details : fix duplicated definitions between asm-ppc64/types.h and
1626 lustre_types.h on PPC.
1629 Frequency : PPC/PPC64 only
1631 Description: asm-ppc/segment.h does not exist
1632 Details : fix compile issue on PPC.
1636 Description: better handle error messages in extents code
1639 Frequency : RHEL4 only
1641 Description: mkfs is very slow on IA64/RHEL4
1642 Details : A performance regression has been discovered in the MPT Fusion
1643 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1644 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1645 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1647 Severity : enhancement
1649 Description: SNMP support enhancement
1650 Details : Adding total number of sampled request for an MDS node in snmp
1653 Severity : enhancement
1655 Description: Optimize ldlm waiting list processing for PR extent locks
1656 Details : When processing waiting list for read extent lock and meeting read
1657 lock that is same or wider to it that is not contended, skip
1658 processing rest of the list and immediatelly return current
1659 status of conflictness, since we are guaranteed there are no
1660 conflicting locks in the rest of the list.
1664 Description: Time out and refuse to reconnect
1665 Details : When the failover node is the primary node, it is possible
1666 to have two identical connections in imp_conn_list. We must
1667 compare not conn's pointers but NIDs, otherwise we can defeat
1668 connection throttling.
1672 Description: port llog fixes from b1_6 into HEAD
1673 Details : Port llog reference couting and some llog cleanups from b1_6
1674 (bug 10800) into HEAD, for protect from panic and access to already
1675 free llog structures.
1679 Description: Detect stride IO mode in read-ahead
1680 Details : When a client does stride read, read-ahead should detect that and
1681 read-ahead pages according to the detected stride pattern.
1685 Description: data checksumming impacts single node performance
1686 Details : add support for several checksum algorithm. Currently, only CRC32
1687 and Adler-32 are supported. The checksum type can be changed on
1688 the fly via /proc/fs/lustre/osc/*/checksum_type.
1692 Description: use adler32 for page checksums
1693 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1698 Description: build for x2 fails
1699 Details : fix compile issue on Cray systems.
1703 Description: Properly match for duplicate locks
1704 Details : Due to different lock order from skiplists code, we need to
1705 traverse entire list for now
1708 Frequency : only on PPC/SLES10
1710 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1711 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1712 be defined. Add a hack in mkfs_lustre.c to work around this compile
1717 Description: LASSERT on MDS when client holding flock lock dies
1718 Details : ldlm pool logic depends on number of granted locks equal to
1719 number of released locks which is not true for flock locks, so
1720 just exclude such locks from consideration.
1724 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1725 Details : Use fsfilt_sync() for both the whole filesystem sync and
1726 individual file sync to eliminate dangerous inode locking
1727 with I_LOCK that can lead to a deadlock.
1731 Description: performance in 1.6.3
1732 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1733 sure that 1MB requests are not fragmented by the block layer.
1735 Severity : enhancement
1737 Description: organize the server-side client stats on per-nid basis
1738 Details : Change the structure of stats under obdfilter and mds to
1746 The "uuid"s file would list the uuids of _active_ exports.
1747 And the clear entry is to clear all stats and stale nids.
1749 Severity : enhancement
1751 Description: eliminate client locks in face of contention
1752 Details : file contention detection and lockless i/o implementation
1753 for contended files.
1757 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1760 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1762 Description: MDS may lose file striping (and hence file data) in some cases
1763 Details : If there are additional extended attributes stored on the MDS,
1764 in particular ACLs, SELinux, or user attributes (if user_xattr
1765 is specified for the client mount options) then there is a risk
1766 of attribute loss. Additionally, the Lustre file striping
1767 needs to be larger than default (e.g. striped over all OSTs),
1768 and an additional attribute must be stored initially in the
1769 inode and then increase in size enough to be moved to the
1770 external attribute block (e.g. ACL growing in size) for file
1775 Description: skiplist implementation simplification
1776 Details : skiplists are used to group compatible locks on granted list
1777 that was implemented as tracking first and last lock of each
1778 lock group the patch changes that to using doubly linked lists
1782 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1783 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1784 because LASSERT on both the data supplied by a client, and the
1785 data on disk is dangerous and incorrect.
1787 Severity : enhancement
1789 Description: Slow truncate/writes to huge files at high offsets.
1790 Details : Directly associate cached pages to lock that protect those pages,
1791 this allows us to quickly find what pages to write and remove
1792 once lock callback is received.
1796 Description: more ldlm soft lockups
1797 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1798 starve other threads from the resource lock for a long time in
1799 case of long waiting queue, so change the debug level from
1800 D_OTHER to the less frequently used D_INFO.
1802 Severity : enhancement
1804 Description: add -gid, -group, -uid, -user options to lfs find
1808 Description: Hung threads in invalidate_inode_pages2_range
1809 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1810 once one is completed. As a result, some RPCs are stuck in the
1811 queue and are never sent.
1815 Description: filter threads hungs on waiting journal commit
1816 Details : Cleanup filter group llog code, then only filter group llog will
1817 be only created in the MDS/OST syncing process.
1821 Description: Procfs and llog threads access destoryed import sometimes.
1822 Details : Sync the import destoryed process with procfs and llog threads by
1823 the import refcount and semaphore.
1825 Severity : enhancement
1827 Description: openlock cache of b1_6 port to HEAD
1832 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1833 Details : If insertion of an extent fails, then discard the inode
1834 preallocation and free data blocks else it can lead to duplicate
1839 Description: don't always update ctime in ext3_xattr_set_handle()
1840 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1841 In some cases the ctime should not be updated, for example for
1842 2.0->1.8 compatibility it is necessary to delete an xattr and it
1843 should not update the ctime.
1847 Bugzilla : 15713/16362
1848 Description: Assertion in iopen_connect_dentry in 1.6.3
1849 Details : looking up an inode via iopen with the wrong generation number can
1850 populate the dcache with a disconneced dentry while the inode
1851 number is in the process of being reallocated. This causes an
1852 assertion failure in iopen since the inode's dentry list contains
1853 both a connected and disconnected dentry.
1857 Description: assertion failure in ldlm_handle2lock()
1858 Details : fix a race between class_handle_unhash() and class_handle2object()
1859 introduced in lustre 1.6.5 by bug 13622.
1864 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1865 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1870 Description: oops in sd_iostats_seq_show()
1871 Details : unloading/reloading the scsi low level driver triggers a kernel
1872 bug when trying to access the sd iostat file.
1877 Description: Kernel panics during QLogic driver reload
1878 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1879 patch, causing memory corruption.
1884 Description: journal_dev option does not work in b1_6
1885 Details : pass mount option during pre-mount.
1887 Severity : enhancement
1889 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1890 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1891 extent information of a file. It can be used to map logical blocks
1892 in a file to physical blocks in the block device.
1896 Description: LDLM soft lockups - improvement
1897 Details : It is be possible to send the lock handle along with each read
1898 or write request because the client is already doing a lock match
1899 itself so there isn't any reason the OST should have to re-do that
1903 Frequency : only on Cray X2
1905 Description: X2 build failures
1906 Details : fix build failures on Cray X2.
1910 Description: xid & resent requests
1911 Details : Initialize RPC XID from clock at startup (randomly if clock is
1914 Severity : enhancement
1916 Description: Add lustre_start utility to start or stop multiple Lustre servers
1921 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1922 Details : In case of memory pressure, list_del() can be called twice on
1923 req->rq_history_list, causing a kernel oops.
1927 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1928 Details : fix stack overflow in the distributed lock manager by defering
1929 export eviction after a failed AST to the elt thread instead of
1930 handling it in the dlm interpret routine.
1934 Description: Convert some comments to new format.
1935 Details : Update documenting comments to match doxygen conventions.
1939 Description: Grammar fixes.
1940 Details : A couple of trivial spelling fixes.
1944 Description: OSD_COUNTERS-mandatory
1945 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1946 to simplify the code.
1950 Description: simplify cmm_device freeing logic.
1951 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1955 Description: Add lockdep support to dt_object_operations locking interface.
1956 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1957 parameter indicating lock ordering. Update mdd code to use new
1962 Description: Introduce failloc constants for lockless IO tests.
1963 Details : Add two new failloc constants to test lockless IO. Only one of
1964 them in implemented---another is checked in yet to be landed
1969 Description: Add lockdep support for inode mutex.
1970 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1971 in the situations where Lustre has to lock more than one inode
1976 Description: Add optional invariants checking support.
1977 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1978 configure switch. This macro is to be used for consistency and
1979 sanity checks that are too expensive to be left in `production'
1984 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1985 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1990 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1991 Details : Use spin_lock_nested() in (the only) situation where more than
1992 one ldlm_lock is locked simultaneously. Also, fix possible
1993 dead-lock in ldlm_lock_change_resource() by enforcing particular
1998 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1999 Details : Instead of specifying each ldlm_lock call-back through separate
2000 parameter, wrap them into struct ldlm_callback_suite.
2004 Description: Kill join_lru obd method and its callers.
2005 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2006 in memory---a requirement implemented through ->o_join_lru() obd
2007 method in HEAD. Remove this method and its users.
2011 Description: Add asynchronous ldlm ENQUEUE completion handler.
2012 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2013 case that stock ldlm_completion_ast() cannot handle as it waits
2014 until lock is granted. Introduce new ldlm_completion_ast_async()
2015 for this. Also comment ldlm_completion_ast().
2019 Description: ldlm_error <-> errno conversion.
2020 Details : Add functions to map (rather arbitrary) between LDLM error codes
2021 and standard errno values. CLIO needs this to prevent LDLM specific
2022 constants from escaping ldlm and osc.
2026 Description: Kill unused ldlm_handle2lock_ns() function.
2027 Details : Kill unused ldlm_handle2lock_ns() function.
2031 Description: Add lu_ref support to ldlm_lock
2032 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2033 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2034 and ldlm_resource. LDLM interface has to be changed, because code that
2035 releases a reference on a lock, has to "know" what reference this is.
2036 In the most frequent case
2038 lock = ldlm_handle2lock(handle);
2040 LDLM_LOCK_PUT(lock);
2042 no changes are required. When any other reference (received _not_ from
2043 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2044 instead of LDLM_LOCK_PUT().
2046 Arguably, changes are pervasive, and interface requires some discipline
2047 for proper use. On the other hand, it was very instrumental in finding
2048 a few leaked lock references.
2052 Description: Add ldlm_lock_addref_try().
2053 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2054 attempts to addref a lock that might be being canceled
2059 Description: Add ldlm_weigh_callback().
2060 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2061 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2062 hard-coded `number of pages' logic.
2066 Description: Add lockdep annotations to llog code.
2067 Details : Use appropriately tagged _nested() locking calls in the places
2068 where llog takes more than one ->lgh_lock lock.
2072 Description: Add loi_kms_set().
2073 Details : Wrap kms updates into a helper function.
2077 Description: Constify instances of struct lsm_operations.
2078 Details : Constify instances of struct lsm_operations.
2082 Description: lu_conf support.
2083 Details : On a server, a file system object is uniquely identified
2084 by a fid, which is sufficient to locate and load all object
2085 state (inode). On a client, on the other hand, more data are
2086 necessary instantiate an object. Change lu_object_find() and
2087 friends to take additional `lu_conf' argument describing object.
2088 Typically this includes layout information.
2092 Description: lu_context fixes.
2093 Details : Introduce new lu_context functions that are needed on the client
2094 side, where some system threads (ptlrpcd) are shared by multiple
2095 modules, and so cannot be stopped during module shutdown.
2099 Description: Add start and stop methods to lu_device_type_operations.
2100 Details : Introduce two new methods in lu_device_type_operations, that are
2101 invoked when first instance of a given type is created and last one
2102 is destroyed respectively. This is need by CLIO.
2106 Description: Add lu_ref support to struct lu_device.
2107 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2108 track leaked references.
2112 Description: Introduce lu_kmem_descr.
2113 Details : lu_kmem_descr and its companion interface allow to create
2114 and destroy a number of kmem caches at once.
2118 Description: Fix lu_object finalization race.
2119 Details : Fix a race between lu_object_find() finding an object and its
2120 concurrent finalization. This race is (most likely) not possible
2121 on the server, but might happen on the client.
2125 Description: Introduce lu_ref interface.
2126 Details : lu_ref is a debugging module allowing to track references to
2127 a given object. It is quite cpu expensive, and has to be
2128 explicitly enabled with --enable-lu_ref. See usage description
2133 Description: Factor lu_site procfs stats into a separate function.
2134 Details : Separate lu_site stats printing code into a separate function
2135 that can be reused on a client.
2139 Description: Constify instances of struct {lu,dt,md}_device_operations.
2140 Details : Constify instances of struct {lu,dt,md}_device_operations.
2144 Description: Introduce struct md_site and move meta-data specific parts of
2145 struct lu_site here.
2146 Details : Move md-specific fields out of struct lu_site into special struct
2147 md_site, so that lu_site can be used on a client.
2151 Description: Kill mdd_lov_destroy().
2152 Details : Remove unused mdd code.
2156 Description: Add st_block checking to multistat.c.
2157 Details : Add st_block checking to multistat.c.
2161 Description: Add lu_ref support to struct obd_device.
2162 Details : Add lu_ref tracking to obd_device.
2166 Description: Kill obd_set_fail_loc().
2167 Details : Remove unused code.
2171 Description: Add special type for ptlrpc_request interpret functions.
2172 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2173 there. Actual usage will be in CLIO.
2177 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2178 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2179 doesn't work with lockdep.
2183 Description: Add rwv.c test program.
2184 Details : New testing program exercising readv(2) and writev(2) (Qian).
2188 Description: Add sendfile.c test program.
2189 Details : New testing program exercising sendfile(2) (Jay).
2193 Description: Ratelimit a message that can be very frequent.
2194 Details : Ratelimit a memory allocation failure message that can
2199 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2200 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2204 Description: Liblustre build fixes.
2205 Details : Liblustre build fixes.
2209 Description: libcfs: add cfs_{need,cond}_resched() interface.
2210 Details : libcfs: add cfs_{need,cond}_resched() definition and
2211 implementations for Linux, NT, and liblustre.
2213 Severity : enhancement
2215 Description: More exported tunables for mballoc
2216 Details : Add support for tunable preallocation window and new tunables for
2217 large/small requests
2221 Description: Detect corruption of block bitmap and checking for preallocations
2222 Details : Checks validity of on-disk block bitmap. Also it does better
2223 checking of number of applied preallocations. When corruption is
2224 found, it turns filesystem readonly to prevent further corruptions.
2228 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2229 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2230 window to 0. If it does hit cache miss, and read-pattern is still
2231 stride-io mode, does not reset the stride window, but also does not
2232 increase the stride window length in this case.
2236 Frequency : only for big-endian servers
2237 Description: Check if system is big-endian while mounting fs with extents feature
2238 Details : Mounting a filesystem with extents feature will fail on big-endian
2239 systems since ext3-based ldiskfs is not supported on big-endian
2240 systems. This can be over-riden with "bigendian_extents" mount option.
2242 Severity : enhancement
2244 Description: The root squash functionality
2245 Details : A security feature, which is to prevent users from being able
2246 to mount lustre on their desktop, run as root, and delete
2247 all of the files in the filesystem. The goal is accomplished by
2248 remapping user id (UID) and group id (GID) of the root user to
2249 a UID and GID specified by the system administartor via Lustre
2250 configuration management server (MGS). The functionality also
2251 allows to specify sets of clients for which the remapping does
2256 Description: Excessive recovery window
2257 Details : With AT enabled, the recovery window can be excessively long (6000+
2258 seconds). To address this problem, we no longer use
2259 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2260 timeout no longer depends on the service time, it is set to
2261 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2262 time via pb_service_time.
2266 Description: Watchdog triggered on MDS failover
2267 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2268 the OSTs know that the MDS "UUID" can be reused for the same export
2269 from a different NID, so we do not need to wait for the export to be
2273 Frequency : rare, only if using MMP with Linux RAID
2275 Description: MMP doesn't work with Linux RAID
2276 Details : While using HA for Lustre servers with Linux RAID, it is possible
2277 that MMP will not detect multiple mounts. To make this work we
2278 need to unplug the device queue in RAID when the MMP block is being
2279 written. Also while reading the MMP block, we should read it from
2280 disk and not the cached one.
2282 Severity : enhancement
2284 Description: open file using fid
2285 Details : A file can be opened using just its fid, like
2286 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2289 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2292 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2293 Details : add proper locking for creation of jbd_4k slab cache
2297 Description: MMP check in ext3_remount() fails without displaying any error
2298 Details : When multiple mount protection fails during remount, proper error
2301 Severity : enhancement
2303 Description: Allow stripe size to be up to 4G-64k
2304 Details : Fix math logic to allow large stripe sizes.
2308 Description: add check for >8TB ldiskfs filesystems
2309 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2310 Don't allow >8TB ldiskfs filesystems to be mounted without
2311 force_over_8tb mount option
2316 Descriptoin: OSS complains "too many blocks" and crashes
2317 Details : In filter_commitrw_write():
2318 1. lquota_chkquota();
2319 2. obdo_from_inode();
2320 3. lquota_pending_commit();
2321 "2." has changed uid and gid in this case, then caused
2322 lquota_chkquota() and lquota_pending_commit() to use
2323 different uid and gid. That is the root of the bug.
2327 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2328 Details : Detect and print out the 'stripe count = -1' case properly.
2332 Description: Out or order replies might be lost on replay
2333 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2334 request with tid smaller then one currently being added, add it
2335 to the start, not end of the list.
2337 --------------------------------------------------------------------------------
2339 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2341 * Support for kernels:
2342 2.6.5-7.283 (SLES 9),
2343 2.6.9-55.EL (RHEL 4),
2344 2.6.16.46-0.14 (SLES 10),
2345 2.6.18.8 vanilla (kernel.org)
2346 * Client support for unpatched kernels:
2347 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2348 2.6.16 - 2.6.22 vanilla (kernel.org)
2349 * Due to recently discovered recovery problems, we do not recommend
2350 using patchless RHEL 4 clients with this or any earlier release.
2351 * Recommended e2fsprogs version: 1.39.cfs8
2352 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2353 * Starting with this release, the ldiskfs backing filesystem required
2354 by Lustre is now in its own package, lustre-ldiskfs. This package
2355 should be installed. It is versioned separately from Lustre and
2356 may be released separately in future.
2360 Description: block reactivating mgc import until all deactivates complete
2361 Details : Fix race when failing back MDT/MGS to itself (testing)
2364 Frequency : at statup only
2366 Description: mds_lov_synchronize race leads to various problems
2367 Details : simultaneous MDT->OST connections at startup can cause the
2368 sync to abort, leaving the OSC in a bad state.
2370 Severity : enhancement
2372 Description: add optional extra BUILD_VERSION info
2373 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2374 to override the lustre version.
2377 Frequency : 2.6.18 servers only
2379 Description: ll_kern_mount() doesn't release the module reference
2380 Details : The ldiskfs module reference count never drops down to 0
2381 because ll_kern_mount() doesn't release the module reference.
2386 Description: server LBUG when using old ost_num_threads parameter
2387 Details : Accept the old ost_num_threads parameter but warn that it
2388 is deprecated, and fix an off-by-one error that caused an LBUG.
2393 Description: Transient SCSI error results in persistent IO issue
2394 Details : iobuf->dr_error is not reinitialized to 0 between two
2398 Frequency : sometimes when underlying device returns I/O errors
2400 Description: OSTs not going read-only during write failures
2401 Details : OSTs are not remounted read-only when the journal commit threads
2402 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2403 instead of the ext3 wrappers.
2407 Description: poor connect scaling with increasing client count
2408 Details : Don't run filter_grant_sanity_check for more than 100 exports
2409 to improve scaling for large numbers of clients.
2412 Frequency : SLES10 only
2414 Description: sanity-quota.sh quotacheck failed: rc = -22
2415 Details : Quotas cannot be enabled on SLES10.
2418 Frequency : liblustre clients only
2420 Description: getdirentries does not give error when run on compute nodes
2421 Details : getdirentries does not fail when the size specified as an argument
2422 is too small to contain at least one entry
2424 Severity : enhancement
2426 Description: Add LNET router traceability for debug purposes
2427 Details : If a checksum failure occurs with a router as part of the
2428 IO path, the NID of the last router that forwarded the bulk data
2429 is printed so it can be identified.
2434 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2435 Details : Due to a race condition, liblustre clients were occasionally
2436 evicted incorrectly.
2438 Severity : enhancement
2440 Description: lfs setstripe use optional parameters instead of postional
2443 Severity : enhancement
2445 Description: Nanosecond timestamp support for ldiskfs
2446 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2447 resolution timestamps. There is not yet support for this at
2448 the Lustre filesystem level.
2451 Frequency : during server recovery
2453 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2454 Details : request with rq_no_resend flag not awake l_wait_event if they get
2458 Frequency : nfs export on patchless client
2460 Description: connectathon hang when test nfs export over patchless client
2461 Details : Disconnected dentry cannot be found with lookup, so we do not need
2462 to unhash it or make it invalid
2465 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2466 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2467 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2472 Description: open req refcounting wrong on reconnect
2473 Details : If reconnect happened between getting open reply from server and
2474 call to mdc_set_replay_data in ll_file_open, we will schedule
2475 replay for unreferenced request that we are about to free.
2476 Subsequent close will crash in variety of ways.
2477 Check that request is still eligible for replay in
2478 mdc_set_replay_data().
2483 Description: disable writes to filesystem when reading health_check file
2484 Details : the default for reading the health_check proc file has changed
2485 to NOT do a journal transaction and write to disk, because this
2486 can cause reads of the /proc file to hang and block HA state
2487 checking on a healthy but otherwise heavily loaded system. It
2488 is possible to return to the previous behaviour during configure
2489 with --enable-health-write.
2491 Severity : enhancement
2493 Description: 64-bit inode version
2494 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2495 to the inode. This will be required for version-based recovery.
2500 Description: MDS fails to start if a duplicate client export is detected
2501 Details : in some rare cases it was possible for a client to connect to
2502 an MDS multiple times. Upon recovery the MDS would detect this
2503 and fail during startup. Handle this more gracefully.
2505 Severity : enhancement
2507 Description: Add -o localflock option to simulate old noflock
2509 Details : This will achieve local-only flock/fcntl locks
2515 Description: log_commit_thread vs filter_destroy race leads to crash
2516 Details : Take import reference before releasing llog record semaphore
2521 Description: Wrong request locking in request set processing
2522 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2523 imp_delayed_list, in this place should be used imp_lock.
2526 Frequency : when reconnection
2528 Description: Grant Leak when osc reconnect to OST
2529 Details : When osc reconnect ost, OST(filter) should check whether it
2530 should grant more space to client by comparing fed_grant and
2531 cl_avail_grant, and return the granted space to client instead
2532 of "new granted" space, because client will call osc_init_grant
2533 to update the client grant space info.
2536 Frequency : when client reconnect to OST
2538 Description: Grant Leak when osc do resend and replay bulk write
2539 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2540 bulk write request, because the grant info while be sync between
2541 OSC and OST when reconnect, and we should ignore the grant info
2542 these of resend/replay write req.
2547 Description: Grant space more than avaiable left space sometimes.
2548 Details : When then OST is about to be full, if two bulk writing from
2549 different clients came to OST. Accord the avaliable space of the
2550 OST, the first req should be permitted, and the second one
2551 should be denied by ENOSPC. But if the seconde arrived before
2552 the first one is commited. The OST might wrongly permit second
2553 writing, which will cause grant space > avaiable space.
2556 Frequency : when client is evicted
2558 Description: Grant might be wrongly erased when osc is evicted by OST
2559 Details : when the import is evicted by server, it will fork another
2560 thread ptlrpc_invalidate_import_thread to invalidate the
2561 import, where the grant will be set to 0. While the original
2562 thread will update the grant it got when connecting. So if
2563 the former happened latter, the grant will be wrongly errased
2564 because of this race.
2569 Description: Checking Stale with correct fid
2570 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2571 to check whether it is stale, because sometimes, we want the
2572 enqueue happened anyway, and op_data.fid2 will not be initialized.
2574 Severity : enhancement
2576 Description: update patchless client
2577 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2580 Frequency : only with 2.4 kernel
2582 Description: random memory corruption
2583 Details : size of struct ll_inode_info is to big for union inode.u and this
2584 can be cause of random memory corruption.
2589 Description: Memory leak in recovery
2590 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2591 It should also check obd_fail before fsfilt_start, otherwise if
2592 fsfilt_start return -EROFS,(failover mds during mds recovery).
2593 then the req will return with repmsg->transno = 0 and rc = EROFS.
2594 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2595 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2596 be freed no matter whether fsfilt_commit success or not.
2599 Frequency : only with huge count clients
2601 Description: Prevents from taking the superblock lock in llap_from_page for
2603 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2604 for llap_from_page call in ll_removepage prevents from taking the
2605 superblock lock for a soon died page.
2610 Description: Not check open intent error before release open handle
2611 Details : in some rare cases, the open intent error is not checked before
2612 release open handle, which may cause
2613 ASSERTION(open_req->rq_transno != 0), because it tries to release
2614 the failed open handle.
2619 Description: Set cat log bitmap only after create log success.
2620 Details : in some rare cases, the cat log bitmap is set too early. and it
2621 should be set only after create log success.
2625 Description: Accessing a block bevice can re-enable I/O when Lustre is
2626 tearing down a device.
2627 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2632 Description: service threads may hog cpus when there are a lot of requests
2633 Details : Insert cond_resched to give other threads a chance to use some CPU
2638 Description: the cat log was not initialized in recovery
2639 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2640 unlink log on mds will not be initialized until mds post
2641 recovery. And also in mds post recovery, the unlink log will
2642 initialization will be done asynchronausly, so there will be race
2643 between add unlink log and unlink log initialization.
2647 Description: brw_stats were being printed incorrectly
2648 Details : brw_stats were being printed as log2 but all of them were not
2649 recorded as log2. Also remove some code duplication arising from
2650 filter_tally_{read,write}.
2654 Frequency : rare, only in recovery.
2655 Description: ASSERTION(req->rq_type != LI_POISON) failed
2656 Details : imp_lock should be held while iterating over imp_sending_list for
2657 prevent destroy request after get timeout in ptlrpc_queue_wait.
2661 Description: replay-single.sh test 52 fails
2662 Details : A lock's skiplist need to be cleanup when it being unlinked
2663 from its resource list.
2667 Description: Short directio read returns full requested size rather than
2669 Details : Direct I/O operations should return actual amount of bytes
2670 transferred rather than requested size.
2672 Severity : enhancement
2674 Description: metadata RPC reduction (e.g. for rm performance)
2675 Details : decrease the amount of synchronous RPC between clients and servers
2676 by canceling conflicing lock before the operation on the client
2677 and packing thier handles into the main operation RPC to server.
2679 Severity : enhancement
2681 Description: Async OSC create to avoid the blocking unnecessarily.
2682 Details : If a OST has no remain object, system will block on the creating
2683 when need to create a new object on this OST. Now, ways use
2684 pre-created objects when available, instead of blocking on an
2685 empty osc while others are not empty. If we must block, we block
2686 for the shortest possible period of time.
2690 Description: improve handling recoverable errors
2691 Details : if request processig with error which can be recoverable on server
2692 request should be resend, otherwise page released from cache and
2695 Severity : enhancement
2697 Description: refine locking for avoid write wrong info into lov_objid file
2698 Details : fix possible races with add new target and write/update data in
2701 --------------------------------------------------------------------------------
2703 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2708 Frequency : on some architectures
2710 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2711 Details : Uninitialized flags sometimes cause configuration commands to
2715 Frequency : patchless clients only
2717 Description: missing __iget() symbol export
2718 Details : The __iget() symbol export is missing. To avoid the need for
2719 this on patchless clients the deathrow inode reaper is turned
2720 off, and we depend on the VM to clean up old inodes. This
2721 dependency was during via the fix for bug 12181.
2723 --------------------------------------------------------------------------------
2725 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2727 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2728 INTEROPERATE with older versions automatically. In many cases a
2729 special upgrade step is needed. Please read the
2730 user documentation before upgrading any part of a live system.
2731 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2732 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2733 * WARNING: Lustre configuration and startup changes are required with
2734 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2738 Severity : enhancement
2740 Description: MountConf
2741 Details : Lustre configuration is now managed via mkfs and mount
2742 commands instead of lmc and lconf. New obd types (MGS, MGC)
2743 are added for dynamic configuration management. See
2744 https://mail.clusterfs.com/wikis/lustre/MountConf for
2747 Severity : enhancement
2749 Description: dynamic OST addition
2750 Details : OSTs can now be added to a live filesystem
2752 Severity : enhancement
2754 Description: startup order invariance
2755 Details : MDTs and OSTs can be started in any order. Clients only
2756 require the MDT to complete startup.
2758 Severity : enhancement
2760 Description: parallel, asynchronous orphan cleanup
2761 Details : orphan cleanup is now performed in separate threads for each
2762 OST, allowing parallel non-blocking operation.
2764 Severity : enhancement
2766 Description: optimized stripe assignment
2767 Details : stripe assignments are now made based on ost space available,
2768 ost previous usage, and OSS previous usage, in order to try
2769 to optimize storage space and networking resources.
2771 Severity : enhancement
2773 Description: Permanently set tunables
2774 Details : All writable /proc/fs/lustre tunables can now be permanently
2775 set on a per-server basis, at mkfs time or on a live system.
2777 Severity : enhancement
2779 Description: Lustre message v2
2780 Details : Add lustre message format v2.
2782 Severity : enhancement
2784 Description: client OST exclusion list
2785 Details : Clients can be started with a list of OSTs that should be
2786 declared "inactive" for known non-responsive OSTs.
2790 Description: SPEC SFS validation failure on NFS v2 over lustre.
2791 Details : Changes the blocksize for regular files to be 2x RPC size,
2792 and not depend on stripe size.
2794 Severity : enhancement
2796 Description: Multiple MD RPCs in flight.
2797 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2798 To avoid overly-overloading MDS, introduce a limit on number of
2799 MDS RPCs in flight for a single client and add /proc controls
2800 to adjust this limit.
2802 Severity : enhancement
2804 Description: client read/write statistics
2805 Details : Add client read/write call usage stats for performance
2806 analysis of user processes.
2807 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2808 file access. extents_stats shows chunk size distribution.
2809 extents_stats_per_process show chunk size distribution per
2812 Severity : enhancement
2814 Description: mds statistics
2815 Details : Add detailed mds operations statistics in
2816 /proc/fs/lustre/mds/*/stats.
2820 Description: Failure of copying files with lustre special EAs.
2821 Details : Client side always return success for setxattr call for lustre
2822 special xattr (currently only "trusted.lov").
2827 Description: Refcount LNET uuids
2828 Details : The global LNET uuid list grew linearly with every startup;
2829 refcount repeated list entries instead of always adding to
2832 Severity : enhancement
2834 Description: Dynamic service threads
2835 Details : Within a small range, start extra service threads
2836 automatically when the request queue builds up.
2839 Frequency : mixed-endian client/server environments
2841 Description: mixed-endian crashes
2842 Details : The new msg_v2 system had some failures in mixed-endian
2845 Severity : enhancement
2847 Description: Easy OST removal
2848 Details : OSTs can be permanently deactivated with e.g. 'lctl
2849 conf_param lustre-OST0001.osc.active=0'
2851 Severity : enhancement
2853 Description: MGS proc entries
2854 Details : Added basic proc entries for the MGS showing what filesystems
2857 Severity : enhancement
2859 Description: provide MGS failover
2860 Details : Added config lock reacquisition after MGS server failover.
2862 Severity : enhancement
2864 Description: add Linux 2.4 support
2865 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2869 Description: a large application tries to do I/O to the same resource and dies
2870 in the middle of it.
2871 Details : Check the req->rq_arrival time after the call to
2872 ost_brw_lock_get(), but before we do anything about
2873 processing it & sending the BULK transfer request. This
2874 should help move old stale pending locks off the queue as
2875 quickly as obd_timeout.
2878 Frequency : when an incorrect nid is specified during startup
2880 Description: ptlrpc connect to non-existant node causes kernel crash
2881 Details : LNET can't be re-entered from an event callback, which
2882 happened when we expire a message after the export has been
2883 cleaned up. Instead, hand the zombie cleanup off to another
2886 Severity : enhancement
2888 Description: plain/inodebits lock performance improvement
2889 Details : Grouping plain/inodebits in granted list by their request modes
2890 and bits policy, thus improving the performance of search through
2894 Frequency : only if OST filesystem is corrupted
2896 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2897 Details : for a short time RPCs with bulk IO are in the replay list,
2898 but replay of bulk IOs is unimplemented. If the OST filesystem
2899 is corrupted due to disk cache incoherency and then replay is
2900 started it is possible to trip an assertion. Avoid putting
2901 committed RPCs into the replay list at all to avoid this issue.
2904 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2906 Description: System hang on startup
2907 Details : This bug allowed the liblustre (e.g. catamount) client to
2908 return to the app before handling all startup RPCs. This
2909 could leave the node unresponsive to lustre network traffic
2910 and manifested as a server ptllnd timeout.
2912 Severity : enhancement
2914 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2915 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2916 boolean to control whether to print peer debug info when a
2917 client's RPC times out.
2920 Frequency : only for kernels with patches from Lustre below 1.4.3
2922 Description: Remove old rdonly API
2923 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2926 Frequency : only for devices with external journals
2928 Description: Set external device read-only also
2929 Details : During a commanded failover stop, we set the disk device
2930 read-only while the server shuts down. We now also set any
2931 external journal device read-only at the same time.
2934 Frequency : when upgrading from 1.4 while trying to change parameters
2936 Description: The wrong (new) MDC name was used when setting parameters for
2937 upgraded MDT's. Also allows changing of OSC (and MDC)
2938 parameters if --writeconf is specified at tunefs upgrade time.
2941 Frequency : when setting specific ost indicies
2943 Description: QOS code breaks on skipped indicies
2944 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2945 created with --index need not be sequential.
2947 Severity : enhancement
2949 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2950 Details : The uninit_groups feature works in conjunction with the kernel
2951 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2952 up the pass1 processing of e2fsck. This is a read-only feature
2953 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2954 mount filesystems that have had this feature enabled.
2956 Severity : enhancement
2958 Description: Improve multi-block allocation algorithm to avoid fragmentation
2959 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2960 allocation locality and avoid filesystem fragmentation.
2962 ------------------------------------------------------------------------------
2964 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2966 * Support for kernels:
2967 2.4.21-47.0.1.EL (RHEL 3)
2968 2.6.5-7.283 (SLES 9)
2969 2.6.9-42.0.10.EL (RHEL 4)
2970 2.6.12.6 vanilla (kernel.org)
2971 2.6.16.27-0.9 (SLES 10)
2972 * Recommended e2fsprogs version: 1.39.cfs5
2974 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2978 Frequency : occasional, depends on client load and configuration
2979 Bugzilla : 12181, 12203
2980 Description: data loss for recently-modified files
2982 Details : In some cases it is possible that recently written or created
2983 files may not be written to disk in a timely manner (this should
2984 normally be within 30s unless client IO load is very high).
2985 The problem appears as zero-length files or files that are a
2986 multiple of 1MB in size after a client crash or client eviction
2987 that are missing data at the end of the file.
2989 This problem is more likely to be hit on clients where files are
2990 repeatedly created and unlinked in the same directory, clients
2991 have a large amount of RAM, have many CPUs, the filesystem has
2992 many OSTs, the clients are rebooted frequently, and/or the files
2993 are not accessed by other nodes after being written.
2995 The presence of the problem can be detected by looking at
2996 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2997 smaller than the second (nr_unused) then dirty files will not
2998 be flushed automatically to disk. "sync; sleep 10" should be
2999 run several times on the node before unmounting it to update
3000 Lustre (this is also safe to run on nodes without this problem).
3002 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3003 that can cause this same problem, so customers using that kernel
3004 also need to update the kernel in addition to Lustre. In order
3005 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3007 It is normal that files written just before a client crash (less
3008 than 30s) may not yet have been flushed to disk, even for local
3012 Frequency : frequent on thin XT3 nodes
3014 Description: UUID collision on thin XT3 Linux nodes
3015 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3016 randomly, since we relied on an insufficiently-seeded PRNG.
3021 Description: OSS hangs after "All ost request buffers busy"
3022 Details : A deadlock between quota and journal operations caused OSS
3023 hangs after printing "All ost request buffers busy."
3026 Frequency : always on liblustre builds
3028 Description: Cleanup compiler warnings on liblustre
3031 Frequency : always on liblustre builds on XT3
3033 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3037 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3038 > 32000 subdirectories
3039 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3040 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3041 subdirectory count crosses 32000. This will aid e2fsck to
3042 correctly handle more than 32000 subdirectories.
3045 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3047 Description: System hang on startup
3048 Details : This bug allowed the liblustre (e.g. catamount) client to
3049 return to the app before handling all startup RPCs. This
3050 could leave the node unresponsive to lustre network traffic
3051 and manifested as a server ptllnd timeout.
3053 Severity : enhancement
3055 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3056 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3057 boolean to control whether to print peer debug info when a
3058 client's RPC times out.
3063 Description: make O_SYNC working on 2.6 kernels
3064 Details : 2.6 kernels use different method for mark pages for write,
3065 so need add a code to lustre for O_SYNC work.
3070 Description: Failure to close file and release space on NFS
3071 Details : Put inode details into lock acquired in ll_intent_file_open.
3072 Use mdc_intent_lock in ll_intent_open to properly
3073 detect all kind of errors unhandled by mdc_enqueue.
3078 Description: proc file read during shutdown sometimes raced obd removal,
3080 Details : Add lock to prevent obd access after proc file removal.
3083 Frequency : Only for files larger than 4GB on 32-bit clients.
3085 Description: improperly doing page alignment of locks
3086 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3087 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3092 Description: RPCs being resent when they shouldn't be
3093 Details : Some RPCs that should not be resent are being resent. This
3094 can cause inconsistencies in the RPC state machine. Do not
3095 resend such requests.
3098 Frequency : rare, only with NFS export
3100 Description: Crash on NFS re-export node
3101 Details : under very unusual load conditions an assertion is hit in
3102 ll_intent_file_open()
3105 Frequency : only if OST filesystem is corrupted
3107 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3108 Details : for a short time RPCs with bulk IO are in the replay list,
3109 but replay of bulk IOs is unimplemented. If the OST filesystem
3110 is corrupted due to disk cache incoherency and then replay is
3111 started it is possible to trip an assertion. Avoid putting
3112 committed RPCs into the replay list at all to avoid this issue.
3117 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3118 Details : Large single O_DIRECT read and write calls can fail to allocate
3119 a sufficiently large buffer to process the request. In case of
3120 allocation failure the allocation is retried with a smaller
3121 buffer and broken into smaller requests.
3123 Severity : enhancement
3125 Description: Add -o localflock option to simulate old noflock behaviour.
3126 Details : This will achieve local-only flock/fcntl locks coherentness.
3131 Description: versioning check is incomplete
3132 Details : Checking the version difference of client vs. server, report
3133 error if the gap is too big.
3138 Description: add support PG_writeback bit
3139 Details : add support for PG_writeback bit for Lustre, for more carefull
3140 work with page cache in 2.6 kernel. This also fix some deadlocks
3141 and remove hack for work O_SYNC with 2.6 kernel.
3143 Severity : enhancement
3145 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3146 Details : The uninit_groups feature works in conjunction with the kernel
3147 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3148 up the pass1 processing of e2fsck. This is a read-only feature
3149 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3150 mount filesystems that have had this feature enabled.
3152 Severity : enhancement
3154 Description: Improve multi-block allocation algorithm to avoid fragmentation
3155 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3156 allocation locality and avoid filesystem fragmentation.
3158 ------------------------------------------------------------------------------
3160 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3162 * Support for kernels:
3163 2.6.9-42.0.3.EL (RHEL 4)
3164 2.6.5-7.276 (SLES 9)
3165 2.4.21-47.0.1.EL (RHEL 3)
3166 2.6.12.6 vanilla (kernel.org)
3167 2.6.16.21-0.8 (SLES10)
3168 * Recommended e2fsprogs version: 1.39.cfs2-0
3170 * The backwards-compatible /proc/sys/portals symlink has been removed
3171 in this release. Before upgrading, please ensure that you change
3172 any configuration scripts or /etc/sysctl.conf files that access
3173 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3174 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3175 in advance of the upgrade on any system running Lustre 1.4.6 or
3176 newer, since /proc/sys/lnet was added in that version.
3177 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3181 Frequency : only when quota is used
3183 Description: avoid scanning export list for quota master
3184 Details : Change the algorithms to avoid scanning export list in order
3185 to improve the efficiency.
3188 Frequency : MDS failover only, very rarely
3190 Description: "went back in time" messages on mds failover
3191 Details : The greatest transno may be lost when the current operation
3192 finishes with an error (transno==0) and the client's last_rcvd
3193 record is over-written. Save the greatest transno in the
3194 mds_last_transno for this case.
3197 Frequency : always for specific kernels and striping counts
3199 Description: client may get "Matching packet too big" without ACL support
3200 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3201 when trying to access files in certain configurations. The
3202 clients should in fact be denied when mounting because they do
3203 not understand ACLs.
3206 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3208 Description: many clients connecting with IO in progress causes connect timeouts
3209 Details : Avoid synchronous journal commits to avoid delays caused by many
3210 clients connecting/disconnecting when bulk IO is in progress.
3211 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3212 OST_IO_PORTAL to avoid delays behind potentially many pending
3216 Frequency : occasionally with multiple writers to a single file
3218 Description: shared writes to file may result in wrong size reported by stat()
3219 Details : Allow growing of kms when extent lock is cancelled
3222 Frequency : always with random mmap IO to multi-striped file
3224 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3225 Details : Only if the hole is at the end of OST object so that kms is too
3226 small. Fix is to increase kms accordingly in ll_nopage.
3229 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3231 Description: writes to a missing object would leak memory on the OST
3232 Details : If there is an inconsistency between the MDS and OST filesystems,
3233 such that the MDS references an object that doesn't exist, writes
3234 to that object will leak memory due to incorrect cleanup in the
3235 error handling path, eventually running out of memory on the OST.
3240 Description: Creating too long symlink causes lustre errors
3241 Details : Check symlink and name lengths before sending requests to MDS.
3244 Frequency : only if flock is enabled (not on by default)
3246 Description: posix locks not released on fd closure on 2.6.9+
3247 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3248 kernels, this caused such locks not to be released on fd close and
3249 then assertions on fs unmount about still used locks.
3252 Frequency : MDS failover only, very rarely
3254 Description: clients may get ASSERTION(granted_lock != NULL)
3255 Details : When request was taking a long time, and a client was resending
3256 a getattr by name lock request. The were multiple lock requests
3257 with the same client lock handle and
3258 mds_getattr_name->fixup_handle_for_resent_request found one of the
3259 lock handles but later failed with ASSERTION(granted_lock != NULL).
3264 Description: handle->h_buffer_credits > 0, assertion failure
3265 Details : h_buffer_credits is zero after truncate, causing assertion
3266 failure. This patch extends the transaction or creates a new
3270 Frequency : NFS re-export or patchless client
3271 Bugzilla : 11179, 10796
3272 Description: Crash on NFS re-export node (__d_move)
3273 Details : We do not want to hash the dentry if we don't have a lock.
3274 But if this dentry is later used in d_move, we'd hit uninitialised
3275 list head d_hash, so we just do this to init d_hash field but
3276 leave dentry unhashed.
3279 Frequency : NFS re-export or patchless client
3281 Description: NFS exports has problem with symbolic link
3282 Details : lustre client didn't properly install dentry when re-exported
3283 to NFS or running patchless client.
3286 Frequency : NFS re-export or patchless client
3288 Description: Various nfs/patchless fixes.
3289 Details : fixes reuse disconected alias for lookup process - this fixes
3290 warning "find_exported_dentry: npd != pd",
3291 fix permission error with open files at nfs.
3292 fix apply umask when do revalidate.
3295 Frequency : occasional
3297 Description: Crash on NFS re-export node
3298 Details : calling clear_page() on the wrong pointer triggered oops in
3299 generic_mapping_read().
3302 Frequency : rarely, using O_DIRECT IO
3304 Description: unaligned directio crashes client with LASSERT
3305 Details : check for unaligned buffers before trying any requests.
3308 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3310 Description: stale data returned from RAID cache
3311 Details : If only a small amount of IO is done to the RAID device before
3312 reading it again it is possible to get stale data from the RAID
3313 cache instead of reading it from disk.
3316 Frequency : always for sles10 kernel
3318 Description: sles10 support
3319 Details : ll_follow_link: compile fixes and using of nd_set_link
3320 under newer kernels.
3323 Frequency : depends on arch, kernel and compiler version, always on sles10
3326 Description: recursive or deep enough symlinks cause stack overflow
3327 Details : getting rid of large stack-allocated variable in
3331 Frequency : depends on hardware
3333 Description: lustre write performance loss in the SLES10 kernel
3334 Details : the performance loss is caused by using of write barriers in the
3335 ext3 code. The SLES10 kernel turns barrier support on by
3336 default. The fix is to undo that change for ldiskfs.
3338 ------------------------------------------------------------------------------
3340 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3342 * Support for kernels:
3343 2.6.9-42.0.3EL (RHEL 4)
3344 2.6.5-7.276 (SLES 9)
3345 2.4.21-47.0.1.EL (RHEL 3)
3346 2.6.12.6 vanilla (kernel.org)
3350 Frequency : quota enabled and large files being deleted
3352 Description: releasing more than 4GB of quota at once hangs OST
3353 Details : If a user deletes more than 4GB of files on a single OST it
3354 will cause the OST to spin in an infinite loop. Release
3355 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3360 Description: statfs data retrieved from /proc may be stale or zero
3361 Details : When reading per-device statfs data from /proc, in the
3362 {kbytes,files}_{total,free,avail} files, it may appear
3363 as zero or be out of date.
3366 Frequency : systems with MD RAID1 external journal devices
3368 Description: lconf's call to blkid is confused by RAID1 journal devices
3369 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3370 of returning all block devices that match the journal UUID.
3373 Frequency : always, for aggregate stripe size over 4GB
3375 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3376 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3377 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3380 Frequency : always if "lfs find" used on a local file/directory
3382 Description: "lfs find" segfaults if used on a local file/directory
3383 Details : The case where a directory component was not specified wasn't
3384 handled correctly. Handle this properly.
3387 Frequency : always on ppc64
3389 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3390 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3391 perform endianness conversion.
3394 Frequency : rarely (truncate to non-zero file size after write under load)
3395 Bugzilla : 10730, 10687
3396 Description: Files padded with zeros to next 4K multiple
3397 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3398 it is possible that files that are truncated to a non-zero size
3399 immediately after being written are filled with zero bytes beyond
3400 the truncated size. No file data is lost.
3402 Severity : enhancement
3404 Description: Allow recovery/failover for liblustre clients.
3405 Details : liblustre clients were unaware of failover configurations until
3408 Severity : enhancement
3410 Description: user file locks should fail when not mounting with flock option
3411 Details : Set up an error-returning stub in ll_file_operations.lock field
3412 to prevent incorrect behaviour when client is mounted without
3413 flock option. Also, set up properly f_op->flock field for
3417 Frequency : always on ia64
3419 Description: "lfs df" loops on printing out MDS statfs information
3420 Details : The obd_ioctl_data was not initialized and in some systems
3421 this caused a failure during the ioctl that did not return
3422 an error. Initialize the struct and return an error on failure.
3425 Frequency : SLES 9 only
3427 Description: Error of copying files with lustre special EAs as root
3428 Details : Client side always return success for setxattr call for lustre
3429 special xattr (currently only "trusted.lov").
3432 Frequency : rarely on clusters with both ia64+i386 clients
3434 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3435 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3436 clients are doing concurrent writes to the same file it is
3437 possible that the ia64 clients may LASSERT because the OST
3438 extent locks are not PAGE_SIZE aligned. Ensure that grown
3439 locks are always aligned on the request boundary.
3442 Frequency : specific use, occasional
3444 Description: Overwriting in use executable truncates on-disk binary image
3445 Details : If one node attempts to overwrite an executable in use by
3446 another node, we now correctly return ETXTBSY instead of
3447 truncating the file.
3449 Severity : enhancement
3451 Description: Async OSC create to avoid the blocking unnecessarily.
3452 Details : If a OST has no remain object, system will block on the creating
3453 when need to create a new object on this OST. Now, ways use
3454 pre-created objects when available, instead of blocking on an
3455 empty osc while others are not empty. If we must block, we block
3456 for the shortest possible period of time.
3461 Description: chmod on Lustre root is propagated to other clients
3462 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3463 invalid by the follow_mount time.
3466 Frequency : liblustre clients only
3468 Description: Race in 'instant cancel' lock handling could lead to such locks
3469 never to be granted in case of SMP MDS
3470 Details : Do not destroy not yet granted but cbpending locks in
3474 Frequency : replay/resend of open
3476 Description: non null lock assetion failure in mds_intent_policy
3477 Details : Trying to replay/resend lockless open requests resulted in
3478 mds_open() returning 0 with no lock. Now it sets a flag if
3479 a lock is going to be returned.
3481 Severity : enhancement
3483 Description: Checksum enhancements
3484 Details : New checksum enhancements allow for resending RPCs that failed
3487 Severity : enhancement
3489 Description: Tunables on number of dirty pages in cacche
3490 Details : Allow to set limit on number of dirty pages cached.
3495 Description: client crash on unmount - lock still has references
3496 Details : In some error handling cases it was possible to leak a lock
3497 reference on a client while accessing a file. This was not
3498 harmful to the client during operation, but would cause the
3499 client to crash when the filesystem is unmounted.
3502 Frequency : specific case, rare
3504 Description: ETXTBSY on mds though file not in use by client
3505 Details : ETXTBSY is no longer incorrectly returned when attempting to
3506 chmod or chown a directory that the user previously tried to
3507 execute or a currently-executing binary.
3510 Frequency : extremely rare except on liblustre-based clients
3512 Description: Lustre space not freed when files are deleted
3513 Details : Clean up open-unlinked files after client eviction. Previously
3514 the unlink was skipped and the files remained as orphans.
3519 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3520 Details : In some cases it was possible to send a blocking callback to
3521 a client doing a glimpse, even though that client didn't get
3522 a lock granted. When the glimpse lock is cancelled on the OST
3523 the freed lock is left on the waiting list and corrupted the list.
3526 Frequency : all core dumps
3528 Description: Broke core dumps to lustre
3529 Details : Negative dentry may be unhashed if parent does not have UPDATE
3530 lock, but some callers, e.g. do_coredump, expect dentry to be
3531 hashed after successful create, hash it in ll_create_it.
3533 ------------------------------------------------------------------------------
3535 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3537 * Support for kernels:
3538 2.6.9-42.0.2.EL (RHEL 4)
3539 2.6.5-7.276 (SLES 9)
3540 2.4.21-40.EL (RHEL 3)
3541 2.6.12.6 vanilla (kernel.org)
3545 Frequency : always on RHEL 3
3547 Description: Number of open files grows over time
3548 Details : The number of open files grows over time, whether or not
3549 Lustre is started. This was due to a filp leak introduced
3550 by one of our kernel patches.
3552 ------------------------------------------------------------------------------
3554 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3556 * Support for kernels:
3557 2.6.9-42.EL (RHEL 4)
3558 2.6.5-7.276 (SLES 9)
3559 2.4.21-40.EL (RHEL 3)
3560 2.6.12.6 vanilla (kernel.org)
3565 Bugzilla : 5719, 9635, 9792, 9684
3566 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3567 Details : If a server is under heavy load and cannot reply to new
3568 connection requests before the client resends the (re)connect,
3569 the connection handling code can behave badly if two service
3570 threads are concurrently handing separate (re)connections from
3571 the same client. Add better locking to the connection handling
3572 code, and ensure that only a single connection will be processed
3573 for a given client UUID, even if the lock is dropped.
3575 Severity : enhancement
3577 Description: add TCP zero-copy support to kernel
3578 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3579 sends if the hardware supports scatter-gather and checksumming.
3580 This allows socklnd to do client-write and server-read more
3581 efficiently and reduce CPU utilization from skbuf copying.
3584 Frequency : only if NFS exporting from client
3586 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3587 Details : If setting attributes on a file created under NFS that had
3588 never been opened it would be possible to oops the client
3589 if the file had no objects.
3592 Frequency : always for liblustre
3594 Description: liblustre client does MDS+OSTs setattr RPC for each write
3595 Details : When doing a write from a liblustre client, the client
3596 incorrectly issued an RPC to the MDS and each OST the file was
3597 striped over in order to update the timestamps. When writing
3598 with small chunks and many clients this could overwhelm the MDS
3599 with RPCs. In all cases it would slow down the write because
3600 these RPCs are unnecessary.
3602 Severity : enhancement
3604 Description: allow number of MDS service threads to be changed at module load
3605 Details : It is now possible to change the number of MDS service threads
3606 running. Adding "options mds mds_num_threads={N}" to the MDS's
3607 /etc/modprobe.conf will set the number of threads for the next
3608 time Lustre is restarted (assuming the "mds" module is also
3609 reloaded at that time). The default number of threads will
3610 stay the same, 32 for most systems.
3615 Description: OST crash if filesystem is unformatted or corrupt
3616 Details : If an OST is started on a device that has never been formatted
3617 or if the filesystem is corrupt and cannot even mount then the
3618 error handling cleanup routines would dereference a NULL pointer.
3623 Description: NULL pointer deref in llap_from_page.
3624 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3625 incorrect) mapping if the page was truncated/reclaimed while it was
3626 searched for. Check for this condition and skip such pages when
3627 doing readahead. Introduce extra check to llap_from_page() to
3628 verify page->mapping->host is non-NULL (so page is not anonymous).
3631 Frequency : Sometimes when using sys_sendfile
3633 Description: "page not covered by a lock" warnings from ll_readpage
3634 Details : sendfile called ll_readpage without right page locks present.
3635 Now we introduced ll_file_sendfile that does necessary locking
3636 around call to generic_file_sendfile() much like we do in
3640 Frequency : with certain MDS communication failures at client mount time
3642 Description: NULL pointer deref after failed client mount
3643 Details : a client connection request may delayed by the network layer
3644 and not be sent until after the PTLRPC layer has timed out the
3645 request. If the client fails the mount immediately it will try
3646 to clean up before the network times out the request. Add a
3647 reference from the request import to the obd device and delay
3648 the cleanup until the network drops the request.
3651 Frequency : occasionally during client (re)connect
3653 Description: assertion failure during client (re)connect
3654 Details : processing a client connection request may be delayed by the
3655 client or server longer than the client connect timeout. This
3656 causes the client to resend the connection request. If the
3657 original connection request is replied in this interval, the
3658 client may trip an assertion failure in ptlrpc_connect_interpret()
3659 which thought it would be the only running connect process.
3662 Frequency : only with obd_echo servers and clients that are rebooted
3664 Description: kernel BUG accessing uninitialized data structure
3665 Details : When running an obd_echo server it did not start the ping_evictor
3666 thread, and when a client was evicted an uninitialized data
3667 structure was accessed. Start the ping_evictor in the RPC
3668 service startup instead of the OBD startup.
3670 Severity : enhancement
3671 Bugzilla : 10193 (patchless)
3672 Description: Remove dependency on various unexported kernel interfaces.
3673 Details : No longer need reparent_to_init, exit_mm, exit_files,
3674 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3677 Frequency : rare (only users of deprecated and unsupported LDAP config)
3679 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3680 Details : LDAP apparently contains 'lustreName' attributes instead of
3681 'name'. A simple remapping of the name is sufficient.
3684 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3686 Description: waiting_locks_callback trips kernel BUG if client is evicted
3687 Details : Running with the dump_on_timeout debug flag turned on makes
3688 it possible that the waiting_locks_callback() can try to dump
3689 the Lustre kernel debug logs from an interrupt handler. Defer
3690 this log dumping to the expired_lock_main() thread.
3692 Severity : enhancement
3694 Description: Support NFS exporting on 2.6 kernels.
3695 Details : Implement non-rawops metadata methods for NFS server to use without
3696 changing NFS server code.
3699 Frequency : very rare (synthetic metadata workload only)
3701 Description: two racing renames might cause an MDS thread to deadlock
3702 Details : Running the "racer" program may cause one MDS thread to rename
3703 a file from being the source of a rename to being the target of
3704 a rename at exactly the same time that another thread is doing
3705 so, and the second thread has already enqueued these locks after
3706 doing a lookup of the target and is trying to relock them in
3707 order. Ensure that we don't try to re-lock the same resource.
3710 Frequency : only very large systems with liblustre clients
3712 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3713 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3714 all OSTs in parallel. This allows the eviction of stale liblustre
3715 clients to proceed much faster than if they were done in series,
3716 and also offers similar improvements for other set_info RPCs.
3721 Description: excessive CPU usage during initial read phase on client
3722 Details : During the initial read phase on a client, it would agressively
3723 retry readahead on the file, consuming too much CPU and impacting
3724 performance (since 1.4.5.8). Improve the readahead algorithm
3725 to avoid this, and also improve some other common cases (read
3726 of small files in particular, where "small" is files smaller than
3727 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3732 Description: MDS crash when receiving packet with unknown intent.
3733 Details : Do not LBUG in unknown intent case, just return -EFAULT
3735 Severity : enhancement
3736 Bugzilla : 9293, 9385
3737 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3738 Details : Do not serialize getattr (non-intent version) and statfs.
3741 Frequency : occasional, when OST network is overloaded/intermittent
3743 Description: client evicted by OST after bulk IO timeout
3744 Details : If a client sends a bulk IO request (read or write) the OST
3745 may evict the client if it is unresposive to its data GET/PUT
3746 request. This is incorrect if the network is overloaded (takes
3747 too long to transfer the RPC data) or dropped the OST GET/PUT
3748 request. There is no need to evict the client at all, since
3749 the pinger and/or lock callbacks will handle this, and the
3750 client can restart the bulk request.
3753 Frequency : Always when mmapping file with no objects
3755 Description: client crashes when mmapping file with no objects
3756 Details : Check that we actually have objects in a file before doing any
3757 operations on objects in ll_vm_open, ll_vm_close and
3763 Description: Request leak when working with deleted CWD
3764 Details : Introduce advanced request refcount tracking for requests
3765 referenced from lustre intent.
3767 Severity : Enhancement
3769 Description: Cache open file handles on client.
3770 Details : MDS now will return special lock along with openhandle, if
3771 requested and client is allowed to hold openhandle, even if unused,
3772 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3773 closing files for every read/write openration.
3775 Severity : Enhancement
3777 Description: Cache open negative dentries on client when possible.
3778 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3779 negative dentries on lock revocation.
3784 Description: Remounting a client read-only wasn't possible with a zconf mount
3785 Details : It wasn't possible to remount a client read-only with llmount.
3787 Severity : enhancement
3788 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3789 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3790 MPICH is included to add Lustre-specific ADIO interfaces.
3791 This is based closely on the UFS ADIO layer and only differs
3792 in file creation, in order to allow the OST striping to be set.
3793 This is user-contributed code and not supported by CFS.
3798 Description: extended inode attributes (immutable, append-only) work improperly
3799 when 2.4 and 2.6 kernels are used on client/server or vice versa
3800 Details : Introduce kernel-independent values for these flags.
3802 Severity : enhancement
3805 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3806 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3807 granularity. Now, Fractional MB granularity is be supported,
3808 this is very useful for low memory system.
3810 Severity : enhancement
3812 Description: Getattr by fid
3813 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3814 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3815 have locks and avoids getattr rpc requests for it on every stat.
3818 Frequency : Always, for filesystems larger than 2TB
3820 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3821 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3822 filesystems larger than 2TB to be mounted with mballoc enabled.
3825 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3827 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3828 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3829 filesystems larger than 2TB is mounted then the kernel will
3830 silently corrupt the start of the filesystem. CONFIG_LBD is
3831 enabled for all CFS-supported kernels, but the possibility of
3832 this happening with a modified kernel config exists.
3834 Severity : enhancement
3836 Description: add client O_DIRECT support for 2.6 kernels
3837 Details : It is now possible to do O_DIRECT reads and writes to files
3838 in the Lustre client mountpoint on 2.6 kernel clients.
3840 Severity : enhancement
3842 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3843 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3844 parallel, not waiting for response from every OST before sending
3845 a rpc to the next OST.
3850 Description: setattr vs write race when updating file timestamps
3851 Details : Client processes that update a file timestamp into the past
3852 right after writing to the file (e.g. tar) it is possible that
3853 the updated file modification time can be reset to the current
3854 time due to a race between processing the setattr and write RPC.
3856 Severity : enhancement
3858 Description: Bring 'lfs find' closer in line with regular Linux find.
3859 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3860 -print0 options and obtains all the needed info through the lustre
3863 Severity : enhancement
3865 Description: support up to 1024 configured devices on one node
3866 Details : change obd_dev array from statically allocated to dynamically
3867 allocated structs as they are first used to reduce memory usage
3872 Description: Flush dirty partially truncated pages during truncate
3873 Details : Immediatelly flush partially truncated pages in filter_setattr,
3874 this way we completely avoid having any pages in page cache on OST
3875 and can retire ugly workarounds during writes to flush such pages.
3880 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3881 Details : protect inode from truncation within vfs_unlink() context
3882 just take a reference before calling vfs_unlink() and release it
3883 when parent's i_sem is free.
3888 Description: last_id value checked outside lock on OST caused LASSERT failure
3889 Details : If there were multiple MDS->OST object precreate requests in
3890 flight, it was possible that the OST's last object id was checked
3891 outside a lock and incorrectly tripped an assertion. Move checks
3892 inside locks, and discard old precreate requests.
3895 Frequency : always, if extents are used on OSTs
3897 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3898 Details : OSTs using the extents format would not zero the high 16 bits of
3899 the index physical block number. This is not a problem for any
3900 OST filesystems smaller than 16TB, and no kernels support ext3
3901 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3902 new/modified files) and can be fixed for existing filesystems
3903 with e2fsprogs-1.39-cfs1.
3908 Description: import connection selection may be incorrect if timer wraps
3909 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3910 import connections to be ignored if the 32-bit jiffies counter
3911 wraps. Use a 64-bit jiffies counter.
3914 Frequency : very large clusters immediately after boot
3916 Description: LNET request buffers exhausted under heavy short-term load
3917 Details : If a large number of client requests are generated on a service
3918 that has previously never seen so many requests it is possible
3919 that the request buffer growth cannot keep up with the spike in
3920 demand. Instead of dropping incoming requests, they are held in
3921 the LND until the RPC service can accept more requests.
3924 Frequency : Sometimes during replay
3926 Description: Assertion failure in ll_local_open after replay.
3927 Details : If replay happened on an open request reply before we were able
3928 to set replay handler, reply will become not swabbed tripping the
3929 assertion in ll_local_open. Now we set the handler right after
3930 recognising of open request
3933 Frequency : very rare
3935 Description: kernel reports "badness in vsnprintf"
3936 Details : Reading from the "recovery_status" /proc file in small chunks
3937 may cause a negative length in lprocfs_obd_rd_recovery_status()
3938 call to vsnprintf() (which is otherwise harmless). Exit early
3939 if there is no more space in the output buffer.
3941 Severity : enhancement
3943 Description: clear OBD RPC statistics by writing to them
3944 Details : It is now possible to clear the OBD RPC statistics by writing
3945 to the "stats" file.
3950 Description: Client mtime is not the same on different clients after utimes
3951 Details : In some cases, the client was using the utimes() syscall on
3952 a file cached on another node. The clients now validate the
3953 ctime from the MDS + OSTs to determine which one is right.
3958 Description: Inability to activate failout mode
3959 Details : lconf script incorrectly assumed that in pythong string's numeric
3960 value is used in comparisons.
3963 Frequency : always with multiple stripes per file
3965 Description: Inefficient object allocation for mutli-stripe files
3966 Details : When selecting which OSTs to stripe files over, for files with
3967 a stripe count that divides evenly into the number of OSTs,
3968 the MDS is always picking the same starting OST for each file.
3969 Return the OST selection heuristic to the original design.
3974 Description: mount failures may take full timeout to return an error
3975 Details : Under some heavy load conditions it is possible that a
3976 failed mount can wait for the full obd_timeout interval,
3977 possibly several minutes, before reporting an error.
3978 Instead return an error as soon as the status is known.
3980 Frequency : quota enabled and large files being deleted
3982 Description: releasing more than 4GB of quota at once hangs OST
3983 Details : If a user deletes more than 4GB of files on a single OST it
3984 will cause the OST to spin in an infinite loop. Release
3985 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3990 Description: statfs data retrieved from /proc may be stale or zero
3991 Details : When reading per-device statfs data from /proc, in the
3992 {kbytes,files}_{total,free,avail} files, it may appear
3993 as zero or be out of date.
3996 Frequency : systems with MD RAID1 external journal devices
3998 Description: lconf's call to blkid is confused by RAID1 journal devices
3999 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4000 of returning all block devices that match the journal UUID.
4003 Frequency : always, for aggregate stripe size over 4GB
4005 Description: assertion fails when trying to use 4GB stripe size
4006 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4007 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4010 Frequency : always on ppc64
4012 Description: the first write on an ext3 filesystem with mballoc got stuck
4013 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4014 perform endianness conversion.
4016 ------------------------------------------------------------------------------
4018 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4020 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4021 INTEROPERATE with older versions automatically. Please read the
4022 user documentation before upgrading any part of a live system.
4023 * WARNING: Lustre networking configuration changes are required with
4024 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4027 * Support for kernels:
4028 2.6.9-22.0.2.EL (RHEL 4)
4029 2.6.5-7.244 (SLES 9)
4030 2.6.12.6 vanilla (kernel.org)
4033 Severity : enhancement
4034 Bugzilla : 7981/8208
4035 Description: Introduced Lustre Networking (LNET)
4036 Details : LNET is new networking infrastructure for Lustre, it includes
4037 a reorganized network configuration mode (see the user
4038 documentation for full details) as well as support for routing
4039 between different network fabrics. Lustre Networking Devices
4040 (LNDs) for the supported network fabrics have also been
4041 created for this new infrastructure.
4043 Severity : enhancement
4044 Description: Introduced Access control lists
4045 Details : clients can set ACLs on files and directories in order to have
4046 more fine-grained permissions than the standard Unix UGO+RWX.
4047 The MDS must be started with the "-o acl" mount option.
4049 Severity : enhancement
4050 Description: Introduced filesystem quotas
4051 Details : Administrators may now establish per-user quotas on the
4054 Severity : enhancement
4056 Description: Configuration change for the XT3
4057 The PTLLND is now used to run Lustre over Portals on the XT3
4058 The configure option(s) --with-cray-portals are no longer used.
4059 Rather --with-portals=<path-to-portals-includes> is used to
4060 enable building on the XT3. In addition to enable XT3 specific
4061 features the option --enable-cray-xt3 must be used.
4066 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4067 Details : A race between allocating a new llcd and re-getting the llcd_lock
4068 allowed another thread to grab newly-allocated llcd.
4070 Severity : enhancement
4072 Description: 2.6 OST async journal commit and locking fix to improve performance
4073 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4074 2.6 kernels are now async as they already were in 2.4 kernels so
4075 that they can commit concurrently with the network bulk transfer.
4076 For block-allocated files the filter allocation semaphore is held
4077 to avoid filesystem fragmentation during allocation. BKL lock
4078 removed for 2.6 xattr operations where it is no longer needed.
4083 Description: lconf incorrectly determined whether two IP networks could talk
4084 Details : In some more complicated routing and multiple-network
4085 configurations, lconf will avoid trying to make a network
4086 connection to a disjoint part of the IP space. It was doing the
4087 math incorrectly for one set of cases.
4092 Description: Fix for potential infinite loop processing records in an llog.
4093 Details : If an llog record is corrupted/zeroed, it is possible to loop
4094 forever in llog_process(). Validate the llog record length
4095 and skip the remainder of the block on error.
4098 Frequency : occasional (liblustre only)
4100 Description: liblustre could not open files whose last component is a symlink
4101 Details : sysio_path_walk() would incorrectly pass the open intent to
4102 intermediate path components.
4105 Frequency : rare (liblustre only with non-standard tuning)
4106 Bugzilla : 7201 (7350)
4107 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4108 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4109 after it has been released, only do this for referenced locks
4111 Severity : enhancement
4113 Description: specifying an (invalid) directory default stripe_size of -1
4114 would reset the directory default striping
4115 Details : stripe_size -1 was used internally to signal directory stripe
4116 removal, now use "all default" to signal dir stripe removal
4117 as a directory striping of "all default" is not useful
4120 Frequency : common for large clusters running liblustre clients
4122 Description: doing an ls when liblustre clients are running is slow
4123 Details : sending a glimpse AST to a liblustre client waits for every AST
4124 to time out, as liblustre clients will not respond. Since they
4125 cannot cache data we refresh the OST lock LVB from disk instead.
4127 Severity : enhancement
4129 Description: doing an ls at the same time as file IO can be slow
4130 Details : enqueue and other "small" requests can be blocked behind many
4131 large IO requests. Create a new OST IO portal for non-IO
4132 requests so they can be processed faster.
4135 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4137 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4138 created file with mode 000
4139 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4140 allow an ftruncate() as a non-root user to a file with mode 000.
4141 Lustre now respects this flag to disable mode checks when
4142 truncating a file owned by the user
4145 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4148 Description: Revoking locks from clients that went dead or catatonic might take
4150 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4151 cancellation of such locks instant on servers without waiting for
4152 any reply from clients. Clients drops these locks when cancel
4153 notification from server is received without replying.
4156 Frequency : liblustre-only, when liblustre client dies or becomes busy
4158 Description: Doing ls on Linux clients can take a long time with active
4160 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4161 granting such locks to it in the first place if possible. Locks
4162 are taken by proxy on the OST during the read or write and
4163 dropped immediately afterward. Add connect flags handling, do
4164 not grant locks to liblustre clients for glimpse ASTs.
4166 Severity : enhancement
4168 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4169 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4170 sequential reads and behaves badly with applications doing random
4171 reads. Improve it by reading ahead at least read region, and
4172 avoiding excessive large RPC for small reads.
4174 Severity : enhancement
4176 Description: Creating more than 1000 files for a single job may cause a load
4177 imbalance on the OSTs if there are also a large number of OSTs.
4178 Details : qos_prep_create() uses an OST index reseed value that is an
4179 even multiple of the number of available OSTs so that if the
4180 reseed happens in the middle of the object allocation it will
4181 still utilize the OSTs as uniformly as possible.
4186 Description: OST or MDS may oops in ping_evictor_main()
4187 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4188 but doesn't restart at beginning of obd_exports_timed list
4191 Severity : enhancement
4193 Description: improve by-nid export eviction on the MDS and OST
4194 Details : allow multiple exports with the same NID to be evicted at one
4195 time without re-searching the exports list.
4198 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4200 Description: MDS may oops in groups_free()
4201 Details : in rare race conditions a newly allocated group_info struct is
4202 freed again, and this can be NULL. The 2.4 compatibility code
4203 for groups_free() checked for a NULL pointer, but 2.6 did not.
4206 Frequency : common for liblustre clients doing little filesystem IO
4207 Bugzilla : 9352, 7313
4208 Description: server may evict liblustre clients accessing contended locks
4209 Details : if a client is granted a lock or receives a completion AST
4210 with a blocking AST already set it would not reply to the AST
4211 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4214 Frequency : lfs setstripe, only systems with more than 160 OSTs
4216 Description: unable to set striping with a starting offset beyond OST 160
4217 Details : llapi_create_file() incorrectly limited the starting stripe
4218 index to the maximum single-file stripe count.
4221 Frequency : LDAP users only
4223 Description: lconf did not handle in-kernel recovery with LDAP properly
4224 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4226 Severity : enhancement
4228 Description: bind OST threads to NUMA nodes to improve performance
4229 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4230 node and do their allocations there to localize memory access
4232 Severity : enhancement
4234 Description: llmount can determine client NID directly from Myrinet (GM)
4235 Details : the client NID code from gmnalnid was moved directly into
4236 llmount, removing the need to use this or specifying the
4237 client NID explicitly when mounting GM clients with zeroconf
4240 Frequency : if client is started with down MDS
4242 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4243 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4244 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4245 ptlrpc_import_delayed_req() because we didn't check if the
4246 request was interrupted, and we also didn't break out of the
4247 event loop if there was no timeout
4252 Description: data loss during non-page-aligned writes to a single file from
4253 both multiple nodes and multiple threads on one node at same time
4254 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4255 inconsistency led to false short-reads, that were cached and later
4256 used by ->prepare_write() to fill in partially written page,
4257 leading to data loss.
4260 Frequency : always, if lconf --abort_recovery used
4262 Description: lconf --abort_recovery fails with 'Operation not supported'
4263 Details : lconf was attempting to abort recovery on the MDT device and not
4266 Severity : enhancement
4268 Description: remove cleanup logs
4269 Details : replace lconf-generated cleanup logs with lustre internal
4270 cleanup routines. Eliminates the need for client-cleanup and
4273 Severity : enhancement
4275 Description: add support for EAs (user and system) on lustre filesystems
4276 Details : it is now possible to store extended attributes in the Lustre
4277 client filesystem, and with the user_xattr mount option it
4278 is possible to allow users to store EAs on their files also
4280 Severity : enhancement
4282 Description: Add possibility (config option) to show minimal available OST free
4284 Details : When compiled with --enable-mindf configure option, statfs(2)
4285 (and so, df) will return least minimal free space available from
4286 all OSTs as amount of free space on FS, instead of summary of
4287 free spaces of all OSTs.
4289 Severity : enhancement
4291 Description: do not expand extent locks acquired on OST-side
4292 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4293 by server: they are not cached anyway.
4296 Frequency : when mmap is used/binaries executed from Lustre
4298 Description: Unmmap pages before throwing them away from read cache.
4299 Details : llap_shrink cache now attempts to unmap pages before discarding
4300 them (if unmapping failed - do not discard). SLES9 kernel has
4301 extra checks that trigger if this unmapping is not done first.
4306 Description: lconf didn't resolve symlinks before checking to see whether a
4307 given mountpoint was already in use
4310 Frequency : when migrating failover services
4311 Bugzilla : 6395, 9514
4312 Description: When migrating a subset of services from a node (e.g. failback
4313 from a failover service node) the remaining services would
4314 time out and evict clients.
4315 Details : lconf --force (implied by --failover) sets the global obd_timeout
4316 to 5 seconds in order to quickly disconnect, but this caused
4317 other RPCs to time out too quickly. Do not change the global
4318 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4320 Severity : enhancement
4321 Frequency : if MDS is started with down OST
4322 Bugzilla : 9439,5706
4323 Description: Allow startup/shutdown of an MDS without depending on the
4324 availability of the OSTs.
4325 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4326 Add appropriate locking and lov-osc refcounts for safe
4327 cleaning. Add osc abort_inflight calls in case the
4328 synchronize never started.
4331 Frequency : occasional (Cray XT3 only)
4333 Description: root not authorized to access files in CRAY_PORTALS environment
4334 Details : The client process capabilities were not honoured on the MDS in
4335 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4336 been accessed by an authorized user then root was able to access
4337 the file on the local client also. The root user capabilities
4338 are now allowed on the MDS, as this environment has secure UID.
4341 Frequency : occasional
4343 Description: ldiskfs "too long searching" message happens too often
4344 Details : A debugging message (otherwise harmless) prints too often on
4345 the OST console. This has been reduced to only happen when
4346 there are fragmentation problems on the filesystem.
4351 Description: Division by zero in statfs when all OSCs are inactive
4352 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4353 lov_statfs divides by value returned by lov_get_stripecnt().
4357 Bugzilla : 9489, 3273
4358 Description: First write from each client to each OST was only 4kB in size,
4359 to initialize client writeback cache, which caused sub-optimal
4360 RPCs and poor layout on disk for the first writen file.
4361 Details : Clients now request an initial cache grant at (re)connect time
4362 and so that they can start streaming writes to the cache right
4363 away and always do full-sized RPCs if there is enough data.
4364 If the OST is rebooted the client also re-establishes its grant
4365 so that client cached writes will be honoured under the grant.
4370 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4371 Details : Now I/O RPCs go to different portal number and (presumably) fast
4372 lock requests (and glimses) and other RPCs get their own service
4373 threads pool that should be able to service those RPCs
4376 Severity : enhancement
4378 Description: Ability to exchange lustre version between client and servers and
4379 issue warnings at client side if client is too old. Also for
4380 liblustre clients there is ability to refuse connection of too old
4382 Details : New 'version' field is added to connect data structure that is
4383 filled with version info. That info is later checked by server and
4387 Frequency : rare, liblustre only.
4388 Bugzilla : 9296, 9581
4389 Description: Two simultaneous writes from liblustre at offset within same page
4390 might proceed at the same time overwriting eachother with stale
4392 Details : I/O lock withing llu_file_prwv was released too early, before data
4393 actually was hitting the wire. Extended lock-holding time until
4394 server acknowledges receiving data.
4397 Frequency : extremely rare. Never observed in practice.
4399 Description: avoid generating lustre_handle cookie of 0.
4400 Details : class_handle_hash() generates handle cookies by incrementing
4401 global counter, and can hit 0 occasionaly (this is unlikely, but
4402 not impossible, because initial value of cookie counter is
4403 selected randonly). Value of 0 is used as a sentinel meaning
4404 "unassigned handle" --- avoid it. Also coalesce two critical
4405 sections in this function into one.
4407 Severity : enhancement
4409 Description: allow liblustre clients to delegate truncate locking to OST
4410 Details : To avoid overhead of locking, liblustre client instructs OST to
4411 take extent lock in ost_punch() on client's behalf. New connection
4412 flag is added to handle backward compatibility.
4414 Severity : enhancement
4415 Bugzilla : 4928, 7341, 9758
4416 Description: allow number of OST service threads to be specified
4417 Details : a module parameter allows the number of OST service threads
4418 to be specified via "options ost ost_num_threads={N}" in the
4419 OSS's /etc/modules.conf or /etc/modprobe.conf.
4423 Bugzilla : 6146, 9635, 9895
4424 Description: servers crash with bad pointer in target_handle_connect()
4425 Details : In rare cases when a client is reconnecting it was possible that
4426 the connection request was the last reference for that export.
4427 We would temporarily drop the export reference and get a new
4428 one, but this may have been the last reference and the export
4429 was just destroyed. Get new reference before dropping old one.
4431 Severity : enhancement
4432 Frequency : if client is started with failover MDS
4434 Description: Allow multiple MDS hostnames in the mount command
4435 Details : Try to read the configuration from all specified MDS
4436 hostnames during a client mount in case the "primary"
4439 Severity : enhancement
4441 Description: Stop sending data to evicted clients as soon as possible.
4442 Details : Check if the client we are about to send or are sending data to
4443 was evicted already. (Check is done every second of waiting,
4444 for which l_wait_event interface was extended to allow checking
4445 of exit condition at specified intervals).
4448 Frequency : rare, normally only when NFS exporting is done from client
4450 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4452 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4455 Frequency : always, if srand() is called before liblustre initialization
4457 Description: Liblustre uses system PRNG disturbing its usage by user application
4458 Details : Introduce internal to lustre fast and high-quality PRNG for
4459 lustre usage and make liblustre and some other places in generic
4460 lustre code to use it.
4462 Severity : enhancement
4463 Bugzilla : 9477, 9557, 9870
4464 Description: Verify that the MDS configuration logs are updated when xml is
4465 Details : Check if the .xml configuration logs are newer than the config
4466 logs stored on the MDS and report an error if this is the case.
4467 Request --write-conf, or allow starting with --old_conf.
4469 Severity : enhancement
4471 Description: Handle symlinks in the path when checking if Lustre is mounted.
4472 Details : Resolve intermediate symlinks when checking if a client has
4473 mounted a filesystem to avoid duplicate client mounts.
4478 Description: lconf can hit an error exception but still return success.
4479 Details : The lconf command catches the Command error exception at the top
4480 level script context and will exit with the associated exit
4481 status, but doesn't ensure that this exit status is non-zero.
4486 Description: failure of ptlrpc thread startup can cause oops
4487 Details : Starting a ptlrpc service thread can fail if there are a large
4488 number of threads or the server memory is very fragmented.
4489 Handle this without oopsing.
4492 Frequency : always, only if liblustre and non-default acceptor port was used
4494 Description: liblustre cannot connect to servers with non-default acceptor port
4495 Details : tcpnal_set_default_params() was not called and was therefore
4496 ignoring the environment varaible TCPNAL_PORT, as well as other
4497 TCPNAL_ environment variables
4502 Description: two objects could be created on the same OST for a single file
4503 Details : If an OST is down, in some cases it was possible to create two
4504 objects on a single OST for a single file. No problems other
4505 than potential performance impact and spurious error messages.
4509 Bugzilla : 5681, 9562
4510 Description: Client may oops in ll_unhash_aliases
4511 Details : Client dcache may become inconsistent in race condition.
4512 In some cases "getcwd" can fail if the current directory is
4518 Description: Inode refcounting problems in NFS export code
4519 Details : link_raw functions used to call d_instantiate without obtaining
4520 extra inode reference first.
4524 Bugzilla : 9942, 9903
4525 Description: Referencing freed requests leading to crash, memleaks with NFS.
4526 Details : We used to require that call to ll_revalidate_it was always
4527 followed by ll_lookup_it. Also with revalidate_special() it is
4528 possible to call ll_revalidate_it() twice for the same dentry
4529 even if first occurence returned success. This fix changes semantic
4530 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4531 reference on a request referred from the intent.
4532 ll_intent_release() then releases such a request.
4535 Frequency : rare, normally benchmark loads only
4537 Description: unlinked inodes were kept in memory on the client
4538 Details : If a client is repeatedly creating and unlinking files it
4539 can accumulate a lot of stale inodes in the inode slab cache.
4540 If there is no other client load running this can cause the
4541 client node to run out of memory. Instead flush old inodes
4542 from client cache that have the same inode number as a new inode.
4545 Frequency : SLES9 2.6.5 kernel and long filenames only
4546 Bugzilla : 9969, 10379
4547 Description: utime reports stale NFS file handle
4548 Details : SLES9 uses out-of-dentry names in some cases, which confused
4549 the lustre dentry revalidation. Change it to always use the
4553 Frequency : rare, unless heavy write-truncate concurrency is continuous
4554 Bugzilla : 4180, 6984, 7171, 9963, 9331
4555 Description: OST becomes very slow and/or deadlocked during object unlink
4556 Details : filter_destroy() was holding onto the parent directory lock
4557 while truncating+unlinking objects. For very large objects this
4558 may block other threads for a long time and slow overall OST
4559 responsiveness. It may also be possible to get a lock ordering
4560 deadlock in this case, or run out of journal credits because of
4561 the combined truncate+unlink. Solution is to do object truncate
4562 first in one transaction without parent lock, and then do the
4563 final unlink in a new transaction with the parent lock. This
4564 reduces the lock hold time dramatically.
4567 Frequency : rare, 2.4 kernels only
4569 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4570 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4571 before unmounting their filesystem in order to avoid deadlock.
4572 The kernel_locked() function in 2.4 kernels only checks whether
4573 the kernel lock is held, not whether it is this process that is
4574 holding it as 2.6 kernels do.
4579 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4580 Details : The client ptlrpc code may be trying to reconnect to a down
4581 server before a previous connection attempt has timed out.
4582 Increase the reconnect interval to be longer than the connection
4583 timeout interval to avoid sending duplicate connections to
4587 Frequency : echo_client brw_test command
4589 Description: fix echo_client to work with OST preallocated code
4590 Details : OST preallocation code (5137) didn't take echo_client IO path
4591 into account: echo_client calls filter methods outside of any
4592 OST thread and, hence, there is no per-thread preallocated
4593 pages and buffers to use. Solution: hijack pga pages for IO. As
4594 a byproduct, this avoids unnecessary data copying.
4598 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4599 Description: Client can oops in mdc_commit_close() after open replay
4600 Details : It was possible for the MDS to return an open request with no
4601 transaction number in mds_finish_transno() if the client was
4602 evicted, but without actually returning an error. Clients
4603 would later try to replay that open and may trip an assertion
4604 Simplify the client close codepath, and always return an error
4605 from the MDS in case the open is not successful.
4608 Frequency : rare, 2.6 OSTs only
4610 Description: OST may deadlock under high load on fragmented files
4611 Details : If there was a heavy load and highly-fragmented OST filesystems
4612 it was possible to have all the OST threads deadlock waiting on
4613 allocation of biovecs, because the biovecs were not released
4614 until the entire RPC IO was completed. Instead, release biovecs
4615 as soon as they are complete to ensure forward IO progress.
4617 Severity : enhancement
4619 Description: Support for specifying external journal device at mount
4620 Details : If an OST or MDS device is formatted with an external journal
4621 device, this device major/minor is stored in the ext3 superblock
4622 and may not be valid for failover. Allow detecting and
4623 specifying the external journal at mount time.
4628 Description: Mounting an MDS with pending unlinked files may cause oops
4629 Details : target_finish_recovery() calls mds_postrecov() which returned
4630 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4631 considers this an error and immediately begins cleaning up the
4632 lov, just after starting the mds_lov process
4634 Severity : enhancement
4636 Description: Implement 'lfs df' to report actual free space on per-OST basis
4637 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4638 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4639 usage of objects; '-h' to report in human readable format.
4641 ------------------------------------------------------------------------------
4643 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4650 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4651 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4652 at startup during orphan unlinking. Instead, skip buddy update
4653 before setup, as it will be regenerated after recovery is complete.
4658 Description: If an OST is inactive, its locks might reference stale inodes.
4659 Details : lov_change_cbdata() must iterate over all namespaces, even if
4660 they are inactive to clear inode references from the lock.
4662 Severity : enhancement
4663 Frequency : occasional, if non-standard max_dirty_mb used
4665 Description: Client will block write RPCs if not enough grant
4666 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4667 then the client will block writes while waiting for another RPC
4668 to complete instead of consuming its dirty limit. With change
4669 we get improved performance when max_dirty_mb is small.
4671 Severity : enhancement
4672 Bugzilla : 3389, 6253
4673 Description: Add support for supplementary groups on the MDS.
4674 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4675 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4676 lookups for user supplementary groups into a cache.
4680 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4681 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4682 for liblustre O_EXCL is enforced by MDS.
4686 Description: readdir never returns NULL in liblustre.
4687 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4688 dentry in struct dirent.
4692 Frequency : liblustre only, depends on application IO pattern
4693 Description: liblustre clients evicted if not contacting servers
4694 Details : Don't put liblustre clients into the ping_evictor list, so
4695 they will not be evicted by the pinger ever.
4697 Severity : enhancement
4699 Description: Add ability to evict clients by NID from MDS.
4700 Details : By echoing "nid:$NID" string into
4701 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4702 $NID would be instantly evicted from this MDS and from all active
4703 OSTs connected to it.
4707 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4708 Details : lookup_it_finish() used to query file size from OSTs that was not
4713 Description: service threads change working directory to that of init
4714 Details : Starting lustre service threads may pin the working directory
4715 of the parent thread, making that filesystem busy. Threads
4716 now change to the working directory of init to avoid this.
4720 Frequency : during shutdown only
4721 Description: shutdown with a failed MDS or OST can cause unmount to hang
4722 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4727 Frequency : 2.6 only
4728 Description: chmod/chown may include an extra supplementary group
4729 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4730 supplementary group and if none is specified this is used.
4733 Bugzilla : 5479 (6816)
4735 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4736 Details : If lookup is in progress on a file that is unlinked we might try
4737 to revalidate the inode and fail in revalidate after lookup is
4738 complete and ll_file_open() enqueues the open again but
4739 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4742 Frequency : always, if lconf --abort_recovery used
4744 Description: lconf --abort_recovery fails with 'Operation not supported'
4745 Details : lconf was attempting to abort recovery on the MDT device and not
4748 ------------------------------------------------------------------------------
4750 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4755 Frequency : rare (only unsupported configurations with a node running as an
4757 Bugzilla : 6514, 5137
4758 Description: Mounting a Lustre file system on a node running as an OST could
4760 Details : OSTs now preallocates memory needed to write out data at
4761 startup, instead of when needed, to avoid having to
4762 allocate memory in possibly low memory situations.
4763 Specifically, if the file system is mounted on on OST,
4764 memory pressure could force it to try to write out data,
4765 which it needed to allocate memory to do. Due to the low
4766 memory, it would be unable to do so and the node would
4767 become unresponsive.
4769 Severity : enhancement
4771 Description: Addition of lconf --service command line option
4772 Details : lconf now accepts a '--service <arg>' option, which is
4773 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4775 Severity : enhancement
4777 Description: Failover mode is now the default for OSTs.
4778 Details : By default, OSTs will now run in failover mode. To return to
4779 the old behaviour, add '--failout' to the lmc line for OSTs.
4781 Severity : enhancement
4783 Description: Health checks are now provided for MDS and OSTs
4784 Details : Additional detailed health check information on MSD and OSTs
4785 is now provided through the procfs health_check value.
4788 Frequency : occasional, depends on IO load
4790 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4791 after numerous create/delete cycles
4792 Details : The ext3 inode allocation policy would not allocate new inodes
4793 very well on the OSTs because there are no new directories
4794 being created. Instead we look for groups with free space if
4795 the parent directories are nearly full.
4800 Description: Network or server problems during mount may cause partially
4801 mounted clients instead of returning an error.
4802 Details : The config llog parsing code may overwrite the error return
4803 code during mount error handling, returning success instead
4809 Description: MDS can fail to allocate large reply buffers
4810 Details : After long uptimes the MDS can fail to allocate large reply
4811 buffers (e.g. zconf client mount config records) due to memory
4812 fragmentation or consumption by the buffer cache. Preallocate
4813 some large reply buffers so that these replies can be sent even
4814 under memory pressure.
4818 Frequency : rare (liblustre)
4819 Description: fsx running with liblustre complained that using truncate() to
4820 extend the file doesn't work. This patch corrects that issue.
4821 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4822 ATTR_SIZE and lsm use in llu_setattr_raw.
4826 Frequency : rare, only 2.6 kernels
4827 Description: Unusual file access patterns on the MDS may result in inode
4828 data being lost in very rare circumstances.
4829 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4830 code in the 2.6 kernel caused the inode and/or EA data not to
4831 be read from disk, causing single-file corruption.
4835 Frequency : rare, only 2.6 filesystems using extents
4836 Description: Heavy concurrent write and delete load may cause data corruption.
4837 Details : It was possible under high-load situations to have an extent
4838 metadata block in the block device cache from a just-unlinked
4839 file overwrite a newly-allocated data block. We now unmap any
4840 metadata buffers that alias just-allocated data blocks.
4844 Frequency : filesystems with default stripe_count larger than 77
4845 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4846 Details : lconf specifies an inode size of 4096 bytes when the default
4847 stripe_count is larger than 77. This conflicts with the default
4848 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4849 this case to avoid pinning too much memory for large EAs.
4851 ------------------------------------------------------------------------------
4853 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4858 Frequency : rare (extremely heavy IO load with hundreds of clients)
4860 Description: Client is evicted, gets IO error writing to file
4861 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4862 caused clients to be evicted for AST timeouts. The fixes in
4863 bug 5192 mean we no longer need to have such short AST timeouts
4864 so ldlm_timeout has been increased.
4867 Frequency : occasional during --force or --failover shutdown under load
4868 Bugzilla : 5949, 4834
4869 Description: Server oops/LBUG if stopped with --force or --failover under load
4870 Details : a collection of import/export refcount and cleanup ordering
4871 issues fixed for safer force cleanup
4874 Frequency : only filesystems larger than 120 OSTs
4875 Bugzilla : 5990, 6223
4876 Description: lfs getstripe would oops on a very large filesystem
4877 Details : lov_getconfig used kfree on vmalloc'd memory
4880 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4881 Bugzilla : 6242, 6243
4882 Description: reading from files that had been truncated to a non-zero size
4883 but never opened returned no data
4884 Details : ll_file_read() reads zeros from no-object files to EOF
4889 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4890 Details : There's little harm in aborting MDS/OSS recovery and letting it
4891 try again, so I removed the LASSERT and return an error instead.
4893 Severity : enhancement
4895 Description: New debugging infrastructure for tracking down data corruption
4896 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4897 (b) cover more of the client-side code path, and (c) try to narrow
4898 down where problems occurred
4902 Bugzilla : 3819, 4364, 4397, 6313
4903 Description: Racing close and eviction MDS could cause assertion in mds_close
4904 Details : It was possible to get multiple mfd references during close and
4905 client eviction, leading to one thread referencing a freed mfd.
4907 Severity: : enhancement
4908 Bugzilla : 3262, 6359
4909 Description: Attempts to reconnect to servers are now more aggressive.
4910 Details : This builds on the enhanced upcall-less recovery that was added
4911 in 1.4.2. When trying to reconnect to servers, clients will
4912 now try each server in the failover group every 10 seconds. By
4913 default, clients would previously try one server every 25 seconds.
4918 Description: After recovery, certain operations trigger a failed
4919 assertion on a client.
4920 Details : Failing over an mds, using lconf -d --failover, while a
4921 client was doing a readdir() call would cause the client to
4922 LBUG after recovery completed and the readdir() was resent.
4924 Severity : enhancement
4926 Description: Default groups are now added by lconf
4927 Details : You can now run lconf --group <servicename> without having to
4928 manually add groups with lmc.
4931 Frequency : occasional
4933 Description: Nodes with an elan id of 0 trigger a failed assertion
4936 Frequency : always when accessing e.g. tty/console device nodes
4938 Description: tty and some other devices nodes cannot be used on lustre
4939 Details : file's private_data field is used by device data and lustre
4940 values in there got lost. New field was added to struct file to
4941 store fs-specific private data.
4944 Frequency : when exporting Lustre via NFS
4946 Description: NFSD failed occasionally when looking up a path component
4947 Details : NFSD is looking up ".." which was broken in ext3 directories
4948 that had grown large enough to become hashed.
4951 Frequency : Clusters with multiple interfaces not on the same subnet
4953 Description: Nodes will repeatedly try to reconnect to an interface which it
4954 cannot reach and report an error to the log.
4955 Details : Extra peer list entries will be created by lconf with some peers
4956 unreachable. lconf now validates the peer before adding it.
4959 Frequency : Only if a default stripe is set on the filesystem root.
4961 Description: Setting a default stripe on the filesystem root prevented the
4962 filesystem from being remounted.
4963 Details : The client was sending extra request flags in the root getattr
4964 request and did not allocate a reply buffer for the dir EA.
4967 Frequency : occasional, higher if lots of files are accessed by one client
4968 Bugzilla : 6159, 6097
4969 Description: Client trips assertion regarding lsm mismatch/magic
4970 Details : While revalidating inodes the VFS looks up inodes with ifind()
4971 and in rare cases can find an inode that is being freed.
4972 The ll_test_inode() code will free the lsm during ifind()
4973 when it finds an existing inode and then the VFS later attaches
4974 this free lsm to a new inode.
4978 Bugzilla : 6422, 7030
4979 Description: MDS deadlock between mkdir and client eviction
4980 Details : Creating a new file via mkdir or mknod (starting a transaction
4981 and getting the ns lock) can deadlock with client eviction
4982 (gets ns lock and trying to finish a synchronous transaction).
4985 Frequency : occasional
4986 Description: While starting a server, the fsfilt_ext3 module could not be
4988 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4989 kernels. Previously, lconf would still use the ext3 name
4990 when trying to load modules. Now, it will correctly use
4991 ext3 on 2.4 and ldiskfs on 2.6.
4993 Severity : enhancement
4994 Description: The default stripe count has been changed to 1
4995 Details : The interpretation of the default stripe count (0, to lfs
4996 or lmc) has been changed to mean striping across a single
4997 OST, rather than all available. For general usage we have
4998 found a stripe count of 1 or 2 works best.
5000 Severity : enhancement
5001 Description: Add support for compiling against Cray portals.
5002 Details : Conditional compiling for some areas that are different
5006 Frequency : occasional
5007 Bugzilla : 6409, 6834
5008 Description: Creating files with an explicit stripe count may lead to
5009 a failed assertion on the MDS
5010 Details : If some OSTs are full or unavailable, creating files may
5011 trigger a failed assertion on the MDS. Now, Lustre will
5012 try to use other servers or return an error to the
5016 Frequency : occasional
5018 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5019 caused lock timeout during readahead (since 1.4.2).
5020 Details : Processes doing readahead might match a lock that hasn't been
5021 granted yet if there are overlapping and conflicting lock
5022 requests. The readahead process waits on ungranted lock
5023 (original lock is CBPENDING), while OST waits for that process
5024 to cancel CBPENDING read lock and eventually evicts client.
5026 Severity : enhancement
5028 Description: Initial enabling of flock support for clients
5029 Details : Implements fcntl advisory locking and file status functions.
5030 This feature is provided as an optional mount flag (default
5031 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5032 locking are implemented yet, and those that are are not guaranteed
5033 to be completely correct in production environments.
5034 mount -t lustre -o [flock|noflock] ...
5037 Frequency : occasional
5039 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5040 assertion in the ext3 JBD (journaling) layer.
5041 Details : The b_committed_data struct is protected by the big kernel lock
5042 in 2.4 kernels, serializing journal_commit_transaction() and
5043 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5044 access to this struct. In 2.6 kernels there is finer grained
5045 locking to improve SMP performance of the JBD layer.
5049 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5054 Description: merge and cleanup kernel patches.
5057 -----------------------------------------------------------------------------
5059 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5061 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5062 versions of Lustre. Please update all servers and clients to
5063 version 1.4.2 or later at the same time. You must also run
5064 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5065 to update the configuration logs.
5067 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5068 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5069 - swab llog records of type '0' so we get proper header size/idx (5861)
5070 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5071 - fix rename of one directory over another leaking an inode (5953)
5072 - avoid SetPageDirty on 2.6 (5981)
5073 - don't re-add just-being-destroyed locks to the waiting list (5653)
5074 - when creating new directories, inherit the parent's custom
5075 striping settings if present parent (3048)
5076 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5077 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5078 - don't hold client locks on temporary worklist from l_lru (5666)
5079 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5080 - automatically evict dead clients (5921)
5081 - Update file size properly in create+truncate+fstat case (6196)
5082 - Do not unhash mountpoint dentries, do not allow removal of
5084 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5085 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5086 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5087 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5088 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5089 - Correctly handle reads of files with no objects (6243)
5090 - lctl recover will also mark a device active if deactivate used (5933)
5092 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5093 - allow --write-conf on an MDS with different nettype than client (5619)
5094 - don't write config llogs to MDS for mounts not from that MDS (5617)
5095 - lconf should create multiple TCP connections from a client (5201)
5096 - init scripts are now turned off by default; run chkconfig --on
5097 lustre and chkconfig --on lustrefs to use them
5098 - upcalls are no longer needed for clients to recover to failover
5100 - add --abort-recovery option to lconf to abort recovery on device
5102 - add support for an arbitrary number of OSTs (3026)
5103 - Quota support protocol changes.
5104 - forward compatibility changes to wire structs (6007)
5105 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5106 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5107 - improved "lustre status" script
5108 - initialize blocksize for non-regular files (6062)
5109 - added --disable-server and --disable-client configure options (5782)
5110 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5111 - Vanilla 2.4.29 support
5112 - increase maximum number of obd devices to 520 (6242)
5113 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5114 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5115 SLES 9 2.6.5 kernels
5116 - limit stripes per file to 160 (the maximum EA size) (6093)
5118 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5121 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5122 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5123 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5124 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5125 - small ext3 extents cleanups and fixes (5733)
5126 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5127 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5128 - avoid needless client->OST connect, fix handle mismatch (5317)
5129 - fix DLM error path that led to out-of-sync client, long delays (5779)
5130 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5131 - fix several locking issues related to i_size (5492,5624,5654,5672)
5132 - don't move pending lock onto export if it is already evicted (5683)
5133 - fix kernel oops when creating .foo in unlinked directory (5548)
5134 - fix deadlock in obdfilter statistics vs. object create (5811)
5135 - use time_{before,after} to avoid timer jiffies wrap (5882)
5136 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5137 - Do not leak request if server was not able to process it (5154)
5138 - If mds_open unable to find parent dir, make that negative lookup(5154)
5139 - don't create new directories with extent-mapping (5909, 5936)
5141 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5142 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5143 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5144 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5145 - support for RHEL4 kernel on i686 (5773)
5146 - provide error messages when incompatible logs are encountered (5898)
5148 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5149 * version 1.4.0.10 (1.4.1 release candidate 1)
5151 - don't keep a lock reference when lock is not granted (4238)
5152 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5153 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5154 - drop import inflight refcount on signal_completed_replay error (5255)
5155 - unlock page after async write error during send (3677)
5156 - handle missing objects in filter_preprw_read properly (5265)
5157 - no transno return for symlink open, don't save no-trasno open (3440)
5158 - don't try to complete elan receive that already failed (4012)
5159 - free RPC server reply state on error (5406)
5160 - clean up thread from ptlrpc_start_thread() on error (5160)
5161 - readahead could read extra page into cache that wasn't ejected (5388)
5162 - prevent races in class_attach/setup/cleanup/detach (5260)
5163 - don't dereference de->d_inode after l_dput of de (5458)
5164 - use "int" for stripe value returned from lock_to_stripe (5544)
5165 - mballoc allocation and error-checking fixes in 2.6 (5504)
5166 - block device patches to fix I/O request sizes in 2.6 (5482)
5167 - look up hostnames for IB nals (5602)
5168 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5169 - don't start multiple acceptors for the same port (5277)
5170 - fix incorrect LASSERT in mds_getattr_name (5635)
5171 - export a proc file for general "ping" checking (5628)
5172 - fix "lfs check" to not block when the MDS is down (5628)
5174 - service request history (4965)
5175 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5176 - create an "evict_client" /proc entry on OSTs, like the MDS has
5177 - fix mount usage message, return errors per mount(8) (5168)
5178 - change grep [] to grep "[]" in tests so they work in more UMLs
5179 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5180 - remove ancient LOV_MAGIC_V0 EA support (5047)
5181 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5182 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5183 - allow readahead allocations to fail when low on memory (5383)
5184 - mmap locking landed again, after considerable improvement (2828)
5185 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5187 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5190 - send OST transaction number in read/write reply to free req (4966)
5191 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5192 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5193 - the watchdog thread now runs as interruptible (5246)
5194 - flock/lockf fixes (but it's still disabled, pending 5135)
5195 - don't use EXT3 constants in llite code (5094)
5196 - memory shortage at startup could cause assertion (5176)
5198 - reorganization of lov code
5199 - single portals codebase
5201 - add extents/mballoc support (5025)
5202 - direct I/O reads in the obdfilter (4048)
5203 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5205 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5208 - send OST transaction number in read/write reply to free req (4966)
5209 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5210 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5211 - the watchdog thread now runs as interruptible (5246)
5212 - handle missing objects in filter_preprw_read properly (5265)
5213 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5214 - drop import inflight refcount on signal_completed_replay error (5255)
5215 - unlock page after async write error during send (3677)
5216 - return original error code on reconstructed replies (3761)
5217 - no transno return for symlink open, don't save no-trasno open (3440)
5219 - add pid to ldlm debugging output (4922)
5220 - bump the watchdog timeouts -- we can't handle 30sec yet
5221 - extra debugging for orphan dentry/inode bug (5259)
5223 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5226 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5227 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5228 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5229 - fix NULL dereference in /proc/sys/portals/routes (4827)
5230 - allow failed mdc_close() operations to be interrupted (4561)
5231 - stop precreate on OST before MDS would time out on it (4778)
5232 - don't send partial-page writes before EOF from client (4410)
5233 - discard client grant for sub-page writes on large-page clients (4520)
5234 - don't free dentries not owned by NFS code, check generation (4806)
5235 - fix lsm leak if mds_create_objects() fails (4801)
5236 - limit debug_daemon file size, always print CERROR messages (4789)
5237 - use transno after validating reply (3892)
5238 - process timed out requests if import state changes (3754)
5239 - update mtime on OST during writes, return in glimpse (4829)
5240 - add mkfsoptions to LDAP (4679)
5241 - use ->max_readahead method instead of zapping global ra (5039)
5242 - don't interrupt __l_wait_event() during strace
5244 - add software watchdogs to catch hung threads quickly (4941)
5245 - make lustrefs init script start after nfs is mounted
5246 - fix CWARN/ERROR duplication (4930)
5247 - return async write errors to application if possible (2248)
5248 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5249 - print NAL number in %x format (4645)
5250 - update barely-supported suse-2.4.21-171 series (4842)
5251 - support for sles 9 %post scripts
5252 - support for building 2.6 kernel-source packages
5253 - support for sles km_* packages
5255 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5258 - ignore -ENOENT errors in osc_destroy (3639)
5259 - notify osc create thread that OSC is being cleaned up (4600)
5260 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5261 - reconstruct ost_handle() like mds_handle() (4657)
5262 - create a new thread to do import eviction to avoid deadlock (3969)
5263 - let lconf resolve symlinked-to devices (4629)
5264 - don't unlink "objects" from directory with default EA (4554)
5265 - hold socknal file ref over connect in case target is down (4394)
5266 - allow more than 32000 subdirectories in a single directory (3244)
5267 - fix blocks count for O_DIRECT writes (3751)
5268 - OST returns ENOSPC from object create when no space left (4539)
5269 - don't send truncate RPC if file size isn't changing (4410)
5270 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5271 - bind to privileged port in socknal and tcpnal (3689)
5273 - rate limit CERROR/CWARN console message to avoid overload (4519)
5274 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5275 - basic mmap support (3918)
5276 - kernel patch series update from b1_4 (4711)
5278 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5281 - avoid crash during MDS cleanup with OST shut down (2775)
5282 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5283 - don't use bad inodes on the MDS (3744)
5284 - dynamic object preallocation to improve recovery speed (4236)
5285 - don't hold spinlock over lock dumping or change debug flags (4401)
5286 - don't zero obd_dev when it is force cleaned (3651)
5287 - print grants to console if they go negative (4431)
5288 - "lctl deactivate" will stop automatic recovery attempts (3406)
5289 - look for existing locks in ldlm_handle_enqueue() (3764)
5290 - don't resolve lock handle twice in recovery avoiding race (4401)
5291 - revalidate should check working dir is a directory (4134)
5293 - don't always mark "slow" obdfilter messages as errors (4418)
5295 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5298 - don't close LustreDB during write_conf until it is done (3860)
5299 - fix typo in lconf for_each_profile (3821)
5300 - allow dumping logs from multiple threads at one time (3820)
5301 - don't allow multiple threads in OSC recovery (3812)
5302 - fix debug_size parameters (3864)
5303 - fix mds_postrecov to initialize import for llog ctxt (3121)
5304 - replace config semaphore with spinlock (3306)
5305 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5306 - don't allow enqueue to complete on a destroyed export (3822)
5307 - down write_lock before checking llog header bitmap (3825)
5308 - recover from lock replay timeout (3764)
5309 - up llog sem before sending rpc (3652)
5310 - reduce ns lock hold times when setting kms (3267)
5311 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5312 - fix NULL deref and obd_dev leak on setup error (3312)
5313 - replace some LBUG about llog ops with error handling (3841)
5314 - don't match INVALID dentries from d_lookup and spin (3784)
5315 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5316 - fix invalid assertion in ptlrpc_set_wait (3880)
5318 - add libwrap support for the TCP acceptor (3996)
5319 - add /proc/sys/portals/routes for non-root route listing (3994)
5320 - allow setting MDS UUID in .xml (2580)
5321 - print the stack of a process that LBUGs (4228)
5323 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5326 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5327 - make sure to unset replay flag from failed open requests (3440)
5328 - if default stripe count is 0, use OST count for inode size (3636)
5329 - update parent mtime/ctime on client for create/unlink (2611)
5330 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5331 - free recovery state on server during a forced cleanup (3571)
5332 - unregister_reply for resent reqs (3063)
5333 - loop back devices mounting and status check on 2.6 (3563)
5334 - fix resource-creation race that can provoke i_size == 0 (3513)
5335 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5336 - more debugging for page-accounting assertion (3746)
5337 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5338 - avoid deadlock after precreation failure (3758)
5339 - fix race and lock order deadlock in orphan handling (3450, 3750)
5340 - add validity checks when grabbing inodes from l_ast_data (3599)
5342 - add /proc/.../recovery_status to obdfilter (3428)
5343 - lightweight CDEBUG infrastructure, debug daemon (3668)
5344 - change default OSC RPC parameters to be better on small clusters
5345 - turn off OST read cache for files smaller than 32MB
5346 - install man pages and include them in rpms (3100)
5347 - add new init script for (un)mounting lustre filesystems (2593)
5348 - run chkconfig in %post for init scripts (3701)
5349 - drop scimac NAL (unmaintained)
5351 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5354 - clean kiobufs before and after use (3485)
5355 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5356 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5357 - change init script to fail more clearly if not run as root (1528)
5358 - allow clients to reconnect during replay (1742)
5359 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5360 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5361 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5362 - fix cond_resched (3554)
5363 - only evict unfinished clients after recovery (3515)
5364 - allow bulk resend, prevent data loss (3570)
5365 - dynamic ptlrpc request buffer allocation (2102)
5366 - don't allow unlinking open directory if it isn't empty (2904)
5367 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5368 - remove extraneous obd dereference causing LASSERT failure (3334)
5369 - don't use get_cycles() when creating temp. files on the mds (3156)
5370 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5371 - handle EEXIST for set-stripe, set proper directory name (3336)
5373 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5374 - fix ksocknal_fmb_callback() error messages (2918)
5376 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5379 - don't copy lvb into (possibly NULL) reply on error (2983)
5380 - don't deref dentry after dput, don't free lvb on error (2922)
5381 - use the kms to determine writeback rpc length (2947)
5382 - increment oti_logcookies when osc is inactive (2948)
5383 - update client's i_blocks count via lvb messages (2543)
5384 - handle intent open/close of special files properly (1557)
5385 - mount MDS with errors=remount-ro, like obdfilter (2009)
5386 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5387 - don't use cancelling-locks' kms values (2947)
5388 - use highest lock extent for kms, not last one (2925)
5389 - don't dereference ERR_PTR() dentry in error handling path (3107)
5390 - fix thread race in portals_debug_dumplog() (3122)
5391 - create lprocfs device entries at setup instead of at attach (1519)
5392 - common AST error handler, don't evict client on completion race (3145)
5393 - zero nameidata in detach_mnt in 2.6 (3118)
5394 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5395 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5396 - initialize RPC timeout timer earlier for 2.6 (3219)
5397 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5398 - print nal/nid for unknown nid (3258)
5399 - additional checks for oscc recovery before doing precreate (3284)
5400 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5401 - don't crash in mdc_close for bad permissions on open (3285)
5402 - zero i_rdev for non-device files (3147)
5403 - clear page->private before handing to FS, better assertion (3119)
5404 - tune the read pipeline (3236)
5405 - fix incorrect decref of invalidated dentry (2350)
5406 - provide read-ahead stats and refine rpc in flight stats (3328)
5407 - don't hold journal transaction open across create RPC (3313)
5408 - update atime on MDS at close time (3265)
5409 - close LDAP connection when recovering to avoid server load (3315)
5410 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5411 - don't leak open file on MDS after open resend (3325)
5412 - serialize filter_precreate and filter_destroy_precreated (3329)
5413 - loop device shouldn't call sync_dev() for nul device (3092)
5414 - clear page cache after eviction (2766)
5415 - resynchronize MDS->OST in background (2824)
5416 - refuse to mount the same filesystem twice on same mountpoint (3394)
5417 - allow llmount to create routes for mounting behind routers (3320)
5418 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5419 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5420 - fix rare problem with rename on htree directories (3417)
5422 - allow default OST striping configuration per directory (1414)
5423 - fix compilation for qswnal for 2.6 kernels (3125)
5424 - increase maximum number of MDS request buffers for large systems
5425 - change liblustreapi to be useful for external progs like lfsck (3098)
5426 - increase local configuration timeout for slow disks (3353)
5427 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5429 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5432 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5433 - don't overwrite extent policy data in reply if lock was blocked (2901)
5434 - drop filter export grants atomically with removal from device (2663)
5435 - del obd_self_export from work_list in class_disconnect_exports (2908)
5436 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5437 - swab reply message in mdc_close, other PPC fixes (2464)
5438 - fix destroying of named logs (2325)
5439 - overwrite old logs when running lconf --write_conf (2264)
5440 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5441 - fix race in target_handle_connect (2898)
5442 - mds_reint_create() should take same inode create lock (2926)
5443 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5444 - don't close files for self_export to avoid uninitialized obd (2936)
5445 - allow MDS with the same name as client node (2939)
5446 - hold dentry reference for closed log files for unlink (2325)
5447 - reserve space for all logs during transactions (2059)
5448 - don't evict page beyond end of stripe extent (2925)
5449 - don't oops on a deleted current working directory (2399)
5450 - handle hard links to targets without a parent properly (2517)
5451 - don't dereference NULL lock when racing during eviction (2867)
5452 - don't grow lock extents when lots of conflicting locks (2919)
5454 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5457 - account for cache space usage on clients to avoid data loss (974)
5458 - lfsck support in lustre kernel code (2349)
5459 - reduce journal credits needed for BRW writes (2370)
5460 - orphan handling to avoid losing space on client/server crashes
5461 - ptlrpcd can be blocked, stopping ALL progress (2477)
5462 - use lock value blocks to assist in proper KMS, faster stat (1021)
5463 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5464 - recovery for initial connections (2355)
5465 - fixes for mds_cleanup_orphans (1934)
5466 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5467 - block all file creations until orphan recovery completes (1901)
5468 - client remove rq_connection from request struct (2423)
5469 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5470 - recovery timer race (2670)
5471 - mdc_close recovey bug (2532)
5472 - ptlrpc cleanup bug (2710)
5473 - mds timeout on local locks (2588)
5474 - namespace lock held during RPCs (2431)
5475 - handle interrupted sync write properly (2503)
5476 - don't try to handle a message that hasn't been replied to (2699)
5477 - client assert failure during cleanup after abort recovery (2701)
5478 - leak mdc device after failed mount (2712)
5479 - ptlrpc_check_set allows timedout requests to complete (2714)
5480 - wait for inflight reqs when ptlrpcd finishes (2710)
5481 - make sure unregistered services are removed from the srv_list
5482 - reset bulk XID's when resending them (caught by 1138 test)
5483 - unregister_bulk after timeout
5484 - fix lconf error (2694)
5485 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5486 - readahead locks pages, leaves pending causing memory pressure (2673)
5487 - increase OST request buffers to 4096 on large machines (2729)
5488 - fix up permission of existing directories in simple_mkdir (2661)
5489 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5490 - don't assign transno to errored transactions (2742)
5491 - don't delete objects on OST if given a bogus objid from MDS (2751)
5492 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5493 - if rq_no_resend, then timeout request after recovery (2432)
5494 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5495 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5496 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5497 - MDS recovery shouldn't delete objects that it has given out (2730)
5498 - if enqueue arrives after completion, don't clobber LVB (2819)
5499 - don't unlock pages twice when trigger_group_io returns error (2814)
5500 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5501 - don't write pages to disk if there was an error (1450)
5502 - don't ping imports that have recovery disabled (2676)
5503 - take buffered bytes into account when balancing socknal conn (2817)
5504 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5505 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5506 - remove little-endian swabbing of llog records (1987)
5507 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5508 - retry reposting request buffers if they fail (1191)
5509 - grow extent at grant time to avoid granting a revoked lock (2809)
5510 - lock revoke doesn't evict page if covered by a second lock (2765)
5511 - disable VM readahead to avoid reading outside lock extents (2805)
5513 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5514 - updated kernel patches for hp-2.4.20 kernel (2681)
5516 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5519 - fix truncated write corruption (2366)
5520 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5522 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5523 - ensure len(uuid) < 37 in lmc (1171)
5524 - fix ia64 OOPS in llog_test (2255)
5525 - zero end of page at obdfilter for partial page writes (2648)
5526 - don't leave stale dentries around after renames (bug 2428)
5527 - fix timeouts when evicting a client with a single lock held (2642)
5528 - set deadline for the initial HELLO message to drain (2634)
5529 - print out dotted-quad IP addresses in the socknal (2302)
5531 - additional debugging for MDS client eviction problem (2443)
5532 - fix mkfsoptions support for osts (2603, 2604)
5534 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5537 - add series for the vanilla 2.6.0 kernel
5538 - add series for the vanilla 2.4.24 kernel
5539 - add series for a cray x86/64 UL kernel drop
5540 - fix xattr patches for the vanilla 2.4.19 series
5542 - generate true UUIDs in lmc (1171)
5543 - have portals stack dumping break in UML (2466)
5544 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5545 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5546 - wake callback waiting threads on client eviction (2460)
5547 - Add --ptldebug and --subsystem to lmc (1719)
5548 - update assertion to allow safe interrupt allocation
5549 - set rq_no_resend for cancel requests (2432)
5550 - recalculate ptlrpcd timeout after resend (2494)
5551 - call vfs_rmdir when removing pending directories (2368)
5552 - fix renaming a file to itself (2429)
5553 - lmc creates a default one-stripe lov (2454)
5554 - expand procfs space to handle large clusters (2326)
5555 - increase UML stack to avoid overflow
5556 - update lconf's list of debug and subsystem masks
5557 - fix lfs find --obd (2510)
5558 - /proc tunable for disabling filter read caching (2591)
5559 - stop rpm packages from altering slapd.conf (2301)
5560 - disable nagle in the socknal under 0conf (2578)
5561 - choose mds inode size based on stripe count (2572)
5562 - fix kernel-source rpm problems (2516)
5564 - add --disable-doc to avoid pdf generation (2421)
5565 - update documentation, tests, type-os, comments
5566 - avoid format warnings on ia64
5567 - remove the TOE NAL
5568 - tiny code cleanups by removing unused fields
5570 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5573 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5574 - remove the most blatant lies from BUILDING file (2371)
5575 - change default debug level to reasonable production setting
5576 - reduce client side cache size to reduce cache flush time
5577 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5578 - make TCP zerocopy and pinger support enabled by default (2476)
5579 - sync writes completed after process exits caused crashes (2319)
5580 - maintain correct mount count on the MDS (2356)
5581 - backout 1557, because 2316 wasn't really fixed
5582 - better file I/O statistics gathering in /proc
5583 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5584 - another kernel patch to fix zero-copy TCP function export
5585 - don't take duplicate lock when processing re-sent getattr (2420)
5586 - lctl uses obd_self_export instead of creating new conn (2353)
5587 - MDS/OST recovery case which requires object creation asserted (2425)
5588 - move lfs from /usr/sbin to /usr/bin in packages
5589 - fix race between mds_client_add and mds_client_free (2417)
5590 - use kmalloc instead of slabs in portals (2430)
5591 - don't create duplicate records when a failover MDS is present (2442)
5592 - remove unnecessary mount age check (2332)
5593 - don't remove directory inodes from locks prematurely (2451)
5594 - don't break if MDS service name is the same as hostname (2103)
5595 - fix races in client write RPC generation when cache full (2482)
5597 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5600 - remove now-unused request->rq_obd (278)
5601 - if an allocation fails, print out how much memory we've used (1933)
5602 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5603 - add the "configurable stack size" patch to most series files (1256)
5604 - ability to write large log records, for 100+ OST configs (2306)
5605 - fix NULL deref when filter_prep fails (2314)
5606 - fix operator precedence error in filter_sync
5607 - dynamic allocation of socknal TX descriptors (2315)
5608 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5609 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5610 - ensure that CWARN messages in Portals always get to the syslog
5611 - __init/__exit are not for prototype decls (ldlm_init/exit)
5612 - x86-64 compile warning fixes
5613 - fix gateway LMC keyword conflict (2318)
5614 - fix MDS lock inversions in getattr/reint paths (1844)
5615 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5616 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5617 - DLM race condition prevented some lock evictions (2328)
5618 - ENOMEM detection and retry on socknal sends (2230)
5619 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5620 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5621 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5622 - better error messages when a client is rejected during recovery (1505)
5623 - avoid cancelling locks which were never granted, after failure (2330)
5624 - fix i_sem/journal inversion in mds_client_add (2333)
5625 - fix truncate/getattr lock cycle deadlock (2334)
5626 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5627 - fix two rare exit paths which could leak an l_lock() ref (2321)
5628 - fencepost error in MDS/OST orphan recovery (2226)
5629 - make log record alignment 8 bytes (1988)
5630 - lstripe now fails when requested offset > ost_count (2237)
5631 - ensure that all kernel series have a complete list.h (1607)
5632 - fix crashes in special-file operations (2316)
5633 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5635 - allow configurable automake binary, for testing new versions
5636 - small update to the lfs documentation
5638 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5640 * fix negative export reference count in fsfilt_sync (2312)
5642 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5643 * release candidate 0.9.1
5645 - orphans are moved into the PENDING directory for possible recovery
5646 - replayed opens now open by fid for orphan/rename safety (1042)
5647 - last close of an orphan inode generates a transno (683)
5648 - chdir() and mount() now pin the directory entry (1020)
5649 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5650 - discard very old requests without processing them (1502)
5651 - remove some common, well-understood CERRORs (1505)
5652 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5653 - clear "grant" flags in OST replies until OST grant code lands (1644)
5654 - fix read performance by not clobbering i_blksize on client (1598)
5655 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5656 - make LRU size a /proc tunable, clears locks when reduced (707)
5657 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5658 - server locks take references on exports now (1558)
5659 - build fixes for 2.4.20-rh trees (1663)
5660 - return an error from lov_create if all OSCs are inactive (1751)
5661 - fix import levels when a reconnect happens without a timeout (1597)
5662 - exit early from mds_open if we get a lookup error (1749)
5663 - partial page read at EOF wouldn't wait for disk before sending (1642)
5664 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5665 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5666 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5667 - write/truncate lock inversion (1639)
5668 - Don't auto-load obdclass, portals modules during cleanup (1495)
5669 - fix timestamps from jumping to "now" (1763)
5670 - extra journal assertions (1648)
5671 - add an extra multiunlink test (1771)
5672 - fix read_record/write_record API (1776)
5673 - fix leak of offset_extent, possible incorrect i_size later (1772)
5674 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5675 - Debugging for the kqswnal_get_idle_tx problems (1820)
5676 - Allow recovery to be attempted multiple times (1536)
5677 - Write out MDS last_rcvd file after it is first created (1600)
5678 - Fix tx_descriptor leak in failed transmit situations (1827)
5679 - ext3 journaling fixes for assertion failure after IO error (1871)
5680 - class_export_put() on freed export after completion AST error (1896)
5681 - Fix revalidate looping in VFS (1322)
5682 - Don't access a freed export during MDS_REINT timeout (1521)
5683 - Add open-unlink recovery support on the MDS (1673,1764)
5684 - Return an error if no MDS data was read from last_rcvd (1946)
5685 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5686 - Don't setup a disk device that doesn't match exported UUID (317)
5687 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5688 - avoid committing NULL handle in force close
5689 - local.sh is now a one-stripe LOV configuration
5690 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5691 - fix ext3 htree duplicate directory entry corruption (1516)
5692 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5693 - update to new LOV EA format (2097)
5694 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5696 2003-06-15 Phil Schwan <phil@clusterfs.com>
5699 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5700 - per-import/export recovery handling (958, 931, 959)
5701 - multiple last-rcvd slots, for serving multiple FSes (949)
5702 - connections are again shared between multiple imp/exports (963, 964)
5703 - "umount -f" would hang if any requests needed to be sent (393, 978)
5704 - avoid pinning large req buffer by copying for queued messages (989)
5705 - add "uuid" to "lctl device" command to help upcalls (991)
5706 - "open" RPCs with transnos would confuse recovery counters (1037)
5707 - do proper endian conversion of all wire messages (288, 340, 891)
5708 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5709 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5710 - fix ptlrpc_connection leak in target_handle_connect (1174)
5711 - fix import refcounting bug in OST and MDS cleanup (1134)
5712 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5713 - fix very unlikely obd_types race condition (501)
5714 - remove osc_open hack for echo_client (1187)
5715 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5716 - a failure in read_inode2 leads to deadlock (1139)
5717 - cancel ack-locks as soon as transaction is committed (1072)
5718 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5719 - make sure to commitrw after any preprw to avoid deadlock (1162)
5720 - failing to execute a file in a lustre FS would lock inode (1203)
5721 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5722 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5723 - obd_uuid2tgt would walk off the end of the list (1255)
5724 - on IA64 the osc would give portals incorrect bulk size (1258)
5725 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5726 - fix lock inversion caused by new llite matching code (1282)
5727 - limit the number of dirty pages on a client to 10MB (1286)
5728 - timed out locks were not being corrected cancelled (1289)
5729 - fix O_DIRECT above 4GB on IA-32 (1292)
5730 * major user-visible changes
5731 - fail out/fail over policy now controlled by the upcall (993)
5733 - add OBD_PING to check server availability and failure (954)
5734 - lustre messages are now sent in sending host order (288, 340, 891)
5735 - add eadatalen to MDS getattr reply (340)
5736 - OST read replies may contain second buffer, with per-page status (593)
5738 2003-03-11 Phil Schwan <phil@clusterfs.com>
5741 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5742 - failed open()s could cause deadlock; fixed (867, 869)
5743 - stop cancelling OST locks when files are closed (481)
5744 - overlapping XID spaces caused network corruption (851, 853)
5745 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5746 - setattr_raw added, to do single-RPC, server-side setattrs
5747 - lmc/lconf syntax change for OST UUIDs
5748 - fix crashy race condition between ptlrpc_free_req and osc_close
5749 - don't use request in mdc_enqueue if we hit a timeout (889)
5750 - don't set the inode i_size for regular files from the MDS (896)
5751 - handle out of order completion AST (842)
5752 - don't LBUG if a lock request times out after receiving AST (913)
5753 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5754 - if a bad lock AST arrives, send an error instead of dropping entirely
5755 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5756 - fix leak in bulk IO when only partially completed (899, 900, 926)
5757 - fix O_DIRECT for ia64 (55)
5758 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5759 - C-z after timeout could hang a process forever; fixed (977)
5761 - client-side I/O cache (678, 924, 929, 941, 970)
5763 - READPAGE and SETATTRs which don't take server-side locks get
5766 2003-02-11 Phil Schwan <phil@clusterfs.com>
5769 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5770 - Fix fsfilt_extN_readpage() to read a full page of directory
5771 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5772 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5773 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5774 - Most intent operations were removed, in favour of a new RPC mode
5775 that does a single RPC to the server and bypasses most of the VFS
5776 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5777 - Aggressively cancel local locks on DLM servers
5778 - mds_reint_unlink sends EA to the client if it's the last nlink.
5779 client uses that EA to unlink OST objects.
5780 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5781 - recursive symlinks were fixed (439)
5782 - fixed NULL deref in DEBUG_REQ
5783 - filter_update_lastobjid no longer calls sync, which annoyed extN
5784 - fixed multi-client small-writes to a single file problem (445)
5785 - fixed mtime updates during file writes (607)
5786 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5787 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5788 - fixed Linux OST size reporting problem (444, 656)
5789 - OST now updates object mtime with writes or setattr (607, 619)
5790 - client verifies file size before zeroing page past EOF (445)
5791 - OST now writes last allocated objid to disk with allocation (108)
5792 - LOV on echo now works (409)
5794 - mds_reint_unlink sends a new buffer, with the EA included. this
5795 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5796 set if a regular file was being unlinked, and it was the last link
5797 - use PtlGet from the target for bulk writes (315)
5798 - OST now updates object mtime with writes or setattr (607, 619)
5799 - LDLM now has a grant-time callback to revalidate locked items, if
5801 - Many MDS operations were reorganized to combat race conditions
5803 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5804 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5806 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5809 - Fully reactivate OST imports after reconnection (512, others)
5810 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5811 - More graceful error handling for truncating on dead OST (515)
5812 - Don't error out unless we're actually accessing dead stripes (474)
5813 - Fix garbage sizes when stripes are missing (410)
5814 - LRU counters were broken, causing constant lock purge (433, 432)
5815 - garbage on read from stripes with failed OSTs (441)
5816 - mark OSCs as active before reconnecting during recovery (438)
5817 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5818 - lfind did not preserve OST order in output (443)
5819 - symlinks cause hung clients, incorrect data (439)
5820 - stop dereferencing request after dropping refcount (457)
5821 - don't LASSERT(spin_is_locked) on non-SMP (455)
5822 - fixes for many rename() bugs
5823 - fstat didn't correctly synchronize attributes (399)
5824 - server must handle lock cancellation during blocking AST prep (487)
5825 - bulk descriptors were free()d too soon (511)
5826 - fix paths in lconf, which would load incorrect modules (451, 507)
5827 - fix confusing lconf 'host not found' error message (386)
5828 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5829 - fix race condition in mdc_blocking_ast() for inode access (526)
5830 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5831 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5832 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5833 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5834 - fix lstripe in several ways (kernel side) (527)
5835 - fix request leak in ldlm_cli_enqueue (262)
5836 - incorrect OSC was marked inactive after OST failure
5837 - call mds_fs_cleanup before unmounting filesystem (524)
5838 - fix races between taking ns_lock and ldlm_lock_change_resource
5839 - fix races updating LOV export open file list
5840 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5841 - fix recovery NULL deref in ldlm_cli_cancel_unused
5842 - fix some DLM races by using new hash table for lock handles (419)
5843 - permit the client to specify desired inodes, at replay
5844 - duplicate requests when we queue them for replay reintegration
5845 - fix last_rcvd offset calculation
5846 - sync after each recovered transaction, so we always make progress
5847 - never, not always, ERESTART requests without transnos
5848 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5850 - skip replay if the MDS says that the client is already connected
5851 - don't check for a recovery-enabled export to match lctl's UUID
5852 - don't INC_USE_COUNT for phantom exports
5853 - don't crash when cleaning up phantom exports (567)
5854 - don't double-finish or set replay data for errored mdc_open requests
5855 - abort requests when they time out, so we don't get old replies
5856 - send/receive replies for AST messages again
5857 - if the client says that it doesn't have the lock, cancel it on the
5859 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5860 mark it as destroyed, it will all work out when decref is called
5861 - fix module use counts (22, 581)
5863 - ASTs now expect a reply (server cancels lock on error reply)
5865 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5868 - fix many simultaneous client startup (392)
5869 - fix dentry->d_it clobbering
5870 - credentials weren't being shipped for readdir/getattr operations
5871 - remove invalid assertions triggered during some concurrent MD
5873 - proper Lustre versions added (336, 389)
5874 - fix memory leak for create error case (398)
5875 - fix LOV locking bug that would get cli/srv out of sync
5876 - fix echo client over LOV (409)
5877 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5878 - fix double-O_EXCL intent crash (424)
5879 - avoid sending multiple lock CANCELs (352)
5881 - MDS can do multi-client recovery (modulo bugs in new code)
5883 - many updates, edits, cleanups
5885 2002-11-18 Phil Schwan <phil@clusterfs.com>
5888 - fix null d_it dereference (346)
5889 - fix full OST/dbench hang (333)
5890 - fix permission problem with file removal (286)
5891 - fix removal of OSCs from LOV when they fail
5892 - fix NULL deref during bulk timeout (214)
5893 - fix problems related to multiple filesystems on one MDS (241)
5894 - fixed serious subtle metadata locking bugs
5895 - free locks on clients when inodes are removed due to memory
5897 - fix inode pointer in lock data (285)
5898 - partial support for multiple MDS on a single host (241)
5899 - data locks weren't cancelled at clear_inode time (290, 311)
5900 - intent locks could lead to unbounded lock growth (205)
5901 - added a maximum lock count, an LRU list, and a flusher
5902 - fix multiple rename (365)
5903 - properly abstracted the echo client
5904 - OSC locked 1 byte too many; fixed
5905 - rewrote brw callback code:
5906 - fixed recovery bugs related to LOVs (306)
5907 - fixed too-many-pages-in-one-write crash (191)
5908 - fixed (again) crash in sync_io_timeout (214)
5909 - probably fixed callback-related race (385)
5911 - Add capability to MDS protocol
5912 - LDLM cancellations and callbacks on different portals
5914 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5917 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5918 - defer open object destruction to close time (601981, 138)
5919 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5920 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5921 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5922 - don't call obd_setattr in ll_file_release for destroyed objects
5923 * protocol change to lustre_msg: move |version| and add |flags|
5924 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5925 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5926 * added replay of create, unlink, link and rename operations during
5927 MDS failover; recovery should be much more robust now
5928 * remove failed OSCs from LOVs (only lov_create uses this so far)
5929 * the lustre-HOWTO was brought (more) up to date (582544)
5931 2002-10-23 Phil Schwan <phil@clusterfs.com>
5934 - in-use dentries weren't being reused properly (617851)
5935 - prevent multiple LDLM setup (599178)
5936 - fix LOV size calculations for truncate (617853)
5937 - fix client handling of MDS intent errors (POSIX)
5938 - fix permission bug in lovstripe.c test (624321)
5939 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5940 - truncate past end of file could corrupt data
5941 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5942 - a race in recovery could return ETIMEDOUT to apps (623947)
5943 - building outside the source directory was fixed
5944 * the lustre-HOWTO was brought (more) up to date (582544)
5945 * major progress was made on recovery functionality
5947 2002-10-10 Phil Schwan <phil@clusterfs.com>
5950 - recovery deadlock fix
5951 - rm -rf causes LBUG fix (617817)
5952 - file open by multiple tasks fix (618962)
5953 - directory permissions bugs (602707 and 620007)
5954 - journal_stop fixed with locking (611313)
5955 - O_APPEND failures resolved (618273, perhaps 614459)
5956 - lconf PATH fix (619770)
5957 - IA64 build fix (621450)
5958 - RPC buffer sizes scale with amount of memory
5960 2002-10-01 Phil Schwan <phil@clusterfs.com>
5963 - locks would be cancelled without throwing away data pages,
5964 resulting in inconsistent data (605627)
5965 - inode attributes were not always being refreshed (605627, 612449)
5966 - lconf now continues to cleanup after lctl reports an error
5967 - MDS now enforces user permissions (602707)
5968 - lprocfs cleanup fixed, but not yet enabled (614157)
5969 - fixed infinite server hang, should a client not respond to an AST
5970 - avoid going into recovery if user calls readlink() with a buffer
5971 that's too small (613941)
5972 - AST RPCs no longer require replies (614867) -- this may be changed
5973 - don't crash server if client sends an IOV that's too big (611336)
5974 - fixed lock conversion deadlock (611892)
5975 - fixed the following of symlinks (614622)
5976 * recovery: the server can remove locks from a client that dies, other
5977 clients can make progress
5978 * more extN patch fixes
5979 * compile-time configurable ptlrpc buffer allocations
5981 - collaborative read cache document
5982 - Lustre Lite Performance CDR document-in-progress
5984 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5987 - fix typo in patch-2.4.18
5989 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5992 - clear ptlrpc request each time in handle_incoming_request()
5993 - unlink of files now destroys the object on the OST
5995 2002-09-19 Peter Braam <braam@clusterfs.com>
5997 * add hard link support
5998 * change obdfile creation method
5999 * kernel patch changed
6001 2002-09-19 Peter Braam <braam@clusterfs.com>
6004 - stack overflow bug in extN fixed
6006 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6008 * documentation updates
6009 - add man pages for config tools
6010 - update tests/README to describe testing with new config tools
6011 - finish metadata API descriptions
6012 * bug fixes and cleanups
6013 - statfs workaround for 16TB limit
6014 - LOV stripe allocation improved, can stripe on subset of OSTs
6015 - LOV file size/IO offset was wrong for files > 4GB in size
6016 - object EA data was being dropped, caused files to be unreadable
6017 - memory overflow with non-LOV OST caused memory corruption
6018 - fixed regression tests to work with new config tools, obdfilter
6019 - fixed bug when directory size became larger than 1 block
6020 - fixed bug (for single client case) when PWD was deleted
6021 - invalidate local directory pages when doing intent-based ops
6022 - avoid LDLM oops when lock callback contained bad data
6024 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6026 * documentation updates
6027 * bug fixes and cleanups
6028 - configuration tools
6031 - 64-bit compile warnings
6032 - 64-bit internal statfs data
6034 * test_brw on persistent OST devices
6036 * lprocfs (disabled)
6038 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6040 * documentation updates
6041 * bug fixes and cleanups
6042 * configuration tools
6044 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6047 * many small fixes to 0_5_4
6048 * io/network handling
6049 * thinkos in MDS operations
6051 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6054 * crucial basic fixes to 0.5.3
6055 * IOR, Iozone work over Elan
6058 2002-08-07 Phil Schwan <phil@clusterfs.com>
6059 * version 0_5_3, our first alpha
6060 * we use the new Portals iovs
6061 * documentation updates
6062 * bug fixes and cleanups
6063 * small changes in the DLM wire protocol
6065 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6066 * version 0_5_1 with some initial stability,
6067 * locking on MD and file I/O.
6068 * documentation updates
6069 * several bug fixes since 0.5.0
6070 * small changes in wire protocol
6072 2002-07-18 Phil Schwan <phil@clusterfs.com>
6074 * delivered as Lustre Light Alpha
6075 * fixed a crash after handling invalid MDS requests
6076 * fixed directory pages for architectures with non-4k pages sizes
6078 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6079 * release version v0_4_4
6080 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6082 * Fixes a number of interruption problems with OST operations.
6083 * Update documentation for portals header changes
6084 * Move all wire protocol structs/defines to lustre_idl.h
6085 * Fixes symlink length bug.
6086 * Add tcpdump to repository.
6088 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6089 * release version v0_4_3
6090 * Fixes statfs for inodes on extN.
6091 * Fixes bug in runtests which would delete /etc/hosts.
6092 * Use 64-bit object IDs wherever possible (not into VFS though)
6093 Remove ost_get_info, which is unused by lustre, and out of date.
6095 2002-07-03 Peter Braam <braam@clusterfs.com>
6096 * release version v0_4_2 Fixes a lookup error (type not passed)
6097 * move forward to head of Portals
6098 * move forward to latest Lustre kernel
6100 2002-06-25 Peter Braam <braam@clusterfs.com>
6101 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob