1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
20 Severity : enhancement
22 Description: Enhancement for lfs(1) command to use numeric uid/gid.
26 Description: Improved cleanup procedure for conf-sanity test_46a.
30 Description: Do not start bulk transfer, if client tried to reconnect.
31 Details : Do not start bulk transfer, if client tried to reconnect,
32 but failed due to active rpcs. Check for exp_abort_active_req
37 Description: Prevent inconsistences between linux and lustre mount structures.
38 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
39 Make the sleep interruptible.
41 Severity : enhancement
43 Description: provide server to client comms path
44 Details : server to client communications path via new LDLM_SET_INFO rpc
48 Description: truncate starts GFP_FS allocation under transaction causing deadlock
49 Details : ldiskfs_truncate calls grab_cache_page which may start page
50 allocation under an open transaction. This may lead to
51 calling prune_icache with consequent lustre reentrance.
55 Description: send correctly lsm on open replay
56 Details : MDS is trust to LSM size on replay open, but client can set wrong size
61 Description: enable client interface failover
62 Details : When a child reconnects from another NID, properly update export
63 nid hash position and ldlm reverse import.
65 Severity : enhancement
67 Description: Communicate OST degraded/readonly state via statfs to MDS
68 Details : Flags in the statfs returned from OSTs indicate whether the
69 OST is in a degraded RAID state, or if the filesystem has
70 turned read-only after a filesystem error is detected.
75 Description: don't panic if EPROTO was hit when reading symlink
76 Details : correctly handling request reference in error cases.
78 Severity : enhancement
80 Description: Add LustreNetLink, a kernel-userspace communcation path.
82 Severity : enhancement
84 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
89 Description: access to llog context before init.
90 Details : move handling CATALOGS file at osc layer and forbid access to llog
94 Frequency : rare, only when client is writing to OST on local node
96 Description: Avoid deadlock for local client writes
97 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
98 memory freeing context. This allows OST threads to set the
99 PF_MEMALLOC flag on task structures in order to allocate memory
100 from reserved pools and complete IO.
101 Use GFP_HIGHUSER for OST allocations for non-local client writes,
102 so that the OST threads generate memory pressure and allow
103 inactive pages to be reclaimed.
105 Severity : enhancement
107 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
112 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
113 Details : move ldlm namespace creation in setup phase to avoid grab
114 _lprocfs_lock with cli_sem held.
118 Description: Temporarily disable grant shrink.
119 Details : Disable the feature for debugging.
122 Frequency : only when formatting test filesystems
124 Description: Unable to run several mkfs.lustre on loop devices at the same time
125 Details : mkfs.lustre returns error 256 on the concurrent loop devices
126 formatting. The solution is to proper handle the error.
128 Severity : enhancement
130 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
132 Severity : enhancement
134 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
139 Description: resolve race between obd_disconnect and class_disconnect_exports
140 Details : if obd_disconnect will be called to already disconnected export he
141 forget release one reference and osc module can't unloaded.
143 Severity : enhancement
145 Description: Allow tuning service thread via /proc
146 Details : For each service a new
147 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
148 created that can be used to set min/max thread counts, and get the
149 current number of running threads.
154 Descriptoin: don't return error if have partially created objects for file.
155 Details : lov_update_create_set uses set->set_success as index for created
156 objects, so if some requests will be failed, they will have hole
157 at end of array and qos_shrink_lsm for allocate correct lsm.
159 Severity : enhancement
161 Description: Update OFED support to 1.4.1
166 Description: Reduce small size read RPC
167 Details : Set read-ahead limit for every file and only do read-ahead when
168 available read-ahead pages are bigger than 1M to avoid small size
171 Severity : enhancement
173 Description: move AT tunable parameters for more consistent usage
174 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
176 Severity : enhancement
178 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
179 Details : allow skip disconnected ost for send statfs request and hide error
185 Description: fix racy locking of mballoc block bitmaps causing BUG
186 Details : The locking of the mballoc buddy bitmap and the in-memory
187 block bitmap was using two different spin locks in some
188 cases. This made it possible to incorrectly access the
189 mballoc bitmap while another process was modifying it,
190 causing a sanity assertion to fail. While no on-disk corruption
191 was reported, there was some risk of this happening.
194 Frequency : rare, on llog test 6
196 Descriptoin: don't allow connect to already connected import
197 Details : allowing connect to already connected import is hide connecting problem.
202 Descriptoin: allow kill process which wait statahead result
203 Details : for some reasons 'ls' can stick in waiting result from statahead,
204 in this case need way for kill this process.
206 Severity : enhancement
208 Description: Add state history info file, enhance import info file
209 Details : Track import connection state changes in a new osc/mdc proc file;
210 add overview-type data to the osc/mdc import proc file.
212 Severity : enhancement
214 Description: MDS create should not wait for statfs RPC while holding DLM lock.
216 Severity : enhancement
218 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
223 Descriptoin: OST grant too much space to client even there are not enough space.
224 Details : Client will shrink its grant cache to OST if there are no write
225 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
226 this grant cache if there are already not enough avaible space
227 (left_space < total_clients * 32M).
230 Frequency : start MDS on uncleanly shutdowned MDS device
232 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
233 Details : stay in waiting mds<>ost recovery finished produce random bugs
234 due race between two ll_sync thread for one lov target. send
235 ACTIVATE event only if connect realy finished and import have
239 Frequency : rare, connect and disconnect target at same time
241 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
242 Details : don't call obd_disconnect under lov_lock. this long time
243 operation and can block ptlrpcd which answer to connect request.
248 Descriptoin: don't lose wakeup for imp_recovery_waitq
249 Details : recover_import_no_retry or invalidate_import and import_close can
250 both sleep on imp_recovery_waitq, but we was send only one wakeup
254 Frequency : always with long access acl
256 Descriptoin: mds can't pack reply with long acl.
257 Details : mds don't control size of acl but they limited by reint/getattr
260 Severity : enhancement
262 Description: Update to SLES10 kernel-2.6.16.60-0.33.
264 Severity : enhancement
266 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
269 Frequency : start MDS on uncleanly shutdowned MDS device
271 Descriptoin: aborting recovery hang on MDS
272 Details : don't throttle destroy RPCs for the MDT.
275 Frequency : on remount
277 Description: external journal device not working after the remount
278 Details : clear dev_rdonly flag for external journal devices in
284 Description: shutdown vs evict race
285 Details : client_disconnect_export vs connect request race.
286 if client will evicted at this time - we start invalidate
287 thread without referece to import and import can be freed
291 Frequency : rare, need acl's on inode.
293 Description: client can't handle ost additional correctly
294 Details : if ost was added after client connected to mds client can have
295 hit lnet_try_match_md ... to big messages to wide striped files.
296 in this case need teach client to handle config events about add
297 lov target and update client max ea size at that event.
299 Severity : enhancement
301 Description: Changelogs
302 Details : Changelogs are a lightweight mechanism to track filesystem
303 metadata and namespace changes. The changelog is recorded
304 permanently on the MDTs, and is periodically "consumed" / purged
305 when records are no longer needed.
307 Severity : enhancement
309 Description: compact fld format with extents
310 Details : Store range of seq rather than every seq in FLD. Seq
311 controller update FLD rather than clients. In Case of CMD, mdt0
312 has FLD, all other metadata server act as non persistent proxy
313 for FLD queries and cache fld entries in fld cache.
318 Description: don't skip ost target if they assigned to file
319 Details : Drop slow OSCs if we can, but not for requested start idx.
320 This means "if OSC is slow and it is not the requested
321 start OST, then it can be skipped, otherwise skip it only
322 if it is inactive/recovering/out-of-space.
326 Description: more cleanup in mds_lov
327 Details : not send LOV EA under replay, we can't know about they size at this
328 time. Don't allow client connect to mds before any ost connected,
329 for avoid problems with LOV EA size and returning EIO to client.
331 Severity : enhancement
333 Description: Interoperability at server side (Disk interoperability)
335 Severity : enhancement
337 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
339 Severity : enhancement
341 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
343 Severity : enhancement
345 Description: New client IO stack (CLIO).
347 Severity : enhancement
349 Description: Commit on sharing. Eliminate inter-client dependencies between
350 uncommitted transactions by doing transaction commits.
351 Thereby clients may recovery independently.
354 Frequency : Create a symlink file with a very long name
356 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
357 Details : If there is no extra space in the request for early cancels,
358 ldlm_req_handles_avail() returns 0 instead of a negative value.
360 Severity : enhancement
362 Description: Add /proc entry for import status
363 Details : The mdc, osc, and mgc import directories now have
364 an import directory that contains useful import data for debugging
367 Severity : enhancement
369 Description: Re-disable certain /proc logging
370 Details : Enable and disable client's offset_stats, extents_stats and
371 extents_stats_per_process stats logging on the fly.
374 Frequency : Only on FC kernels 2.6.22+
376 Description: oops in statahead
377 Details : Do not drop reference count for the dentry from VFS when lookup,
378 VFS will do that by itself.
380 Severity : enhancement
382 Description: Generic /proc file permissions
383 Details : Set /Proc file permissions in a more generic way to enable non-
384 root users operate on some /proc files.
388 Description: Hitting mdc_commit_close() ASSERTION
389 Details : Properly handle request reference release in
390 ll_release_openhandle().
394 Description: quota recovery deadlock during mds failover
395 Details : This patch includes att18982, att18236, att18237 in bz14840.
397 1. fix osts hang when mds does failover with quotaon
398 2. prevent watchdog storm when osts threads wait for the
403 Frequency : only patchless client
404 Description: add workaround for race between add/remove dentry from hash
406 Severity : enhancement
408 Description: Allow OST glimpses to return PW locks
412 Description: LBUG when llog conf file is full
413 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
417 Description: Prevent import from entering FULL state when server in recovery
421 Description: service mount cannot take device name with ":"
422 Details : Only when device name contains ":/" will mount treat it as
428 Description: replace ptlrpcd with the statahead thread to interpret the async
429 statahead RPC callback
433 Frequency : on recovery
434 Description: I/O failures after umount during fail back
435 Details : if client reconnected to restarted server we need join to recovery
436 instead of find server handler is changed and process self
437 eviction with cancel all locks.
439 Severity : enhancement
441 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
443 Severity : enhancement
445 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
447 Severity : enhancement
449 Description: Upcall on Lustre log has been dumped
450 Details : Allow for a user mode script to be called once a Lustre log has
451 been dumped. It passes the filename of the dumped log to the
452 script, the location of the script can be specified via
453 /proc/sys/lnet/debug_log_upcall.
458 Description: avoid idr_remove called for id which is not allocated.
459 Details : Move assigment s_dev for clustered nfs to end of initialization,
460 to avoid problem with error handling.
465 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
466 Details : When connection is reused this not moved from CONN_UNUSED_HASH
467 into CONN_USED_HASH and this prodice warning when put connection
468 again in unused hash.
471 Severity : enhancement
473 Description: File striping can now be set to use an arbitrary pool of OSTs.
475 Severity : enhancement
477 Description: Export bytes_read/bytes_write count on OSC/OST.
481 Description: Early reply size mismatch, MGC loses connection
482 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
483 the connect flags are properly negotiated.
488 Description: quotas are not honored with O_DIRECT
489 Details : all writes with the flag O_DIRECT will use grants which leads to
490 this problem. Now using OBD_BRW_SYNC to guard this.
494 Description: add quota statistics
495 Details : 1. sort out quota proc entries and proc code.
496 2. add quota statistics
498 Severity : enhancement
500 Description: enable quota support for HEAD.
504 Description: Properly propagate oinfo flags from lov to osc for statfs
505 Details : restore missing copy oi_flags to lov requests.
507 Severity : enhancement
509 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
510 : l_getgroups(8), lst(8), routerstat(8)
511 Details : included man pages for llobdstat(8), llstat(8),
512 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
514 Severity : enhancement
516 Description: configure's --enable-quota should check the
517 : kernel .config for CONFIG_QUOTA
518 Details : configure is terminated if --enable-quota is passed but
519 : no quota support is in kernel
523 Description: Remove portals compatibility
524 Details : Remove portals compatibility, not interoperable with releases
529 Description: Resolve device initialization race
530 Details : Prevent proc handler from accessing devices added to the
531 obd_devs array but yet be intialized.
533 Severity : enhancement
535 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
537 Severity : enhancement
539 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
544 Description: Using wrong pointer in osc_brw_prep_request
545 Details : Access to array[-1] can produce panic if kernel compiled with
546 CONFIG_PAGE_ALLOC enabled
550 Description: Client runs out of low memory
551 Details : Consider only lowmem when counting initial number of llap pages
555 Description: *optional* service tags registration
556 Details : if the "service tags" package is installed on a Lustre node
557 When the filesystem is mounted, a local-node service tag will
558 be created. See http://inventory.sun.com/ for more information
559 about the Service Tags asset management system.
563 Description: Kernel BUG tries to release flock
564 Details : Lustre does not destroy flock lock before last reference goes
565 away. So always drop flock locks when client is evicted and
566 perform unlock regardless of successfulness of speaking to MDS.
570 Description: add refcount protection for osc callbacks, avoid panic on shutdown
574 Description: sanity test 65a fails if stripecount of -1 is set
575 Details : handle -1 striping on filesystem in ll_dirstripe_verify
580 Description: ASSERTION(CheckWriteback(page,cmd)) failed
581 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
584 Severity : enhancement
586 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
590 Description: do not process already freed flock
591 Details : flock can possibly be freed by another thread before it reaches
592 to ldlm_flock_completion_ast.
596 Description: LBUG during stress test
597 Details : Need properly lock accesses the flock deadlock detection list.
601 Description: oops in page fault handler
602 Details : kernel page fault handler can return two special 'pages' in error
603 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
607 Description: timeout with invalidate import.
608 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
609 should be handled by ptlrpcd. This produce long age waiting and
610 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
612 Severity : enhancement
614 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
618 Description: enable MGS and MDT services start separately
619 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
620 a co-located MGS without starting the MGS, which is a complement
621 to 'nosvc' mount option.
625 Frequency : after recovery
626 Description: precreate to many object's after del orphan.
627 Details : del orphan st in oscc last_id == next_id and this triger growing
628 count of precreated objects. Set flag LOW to skip increase count
629 of precreated objects.
633 Frequency : rare, on clear nid stats
634 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
635 Details : when clean nid stats sometimes try destroy live entry,
636 and this produce panic in free.
640 Description: Stack overflow during MDS log replay
641 ease stack pressure by using a thread dealing llog_process.
645 Description: wait until IO finished before start new when do lock cancel.
646 Details : VM protocol want old IO finished before start new, in this case
647 need wait until PG_writeback is cleared until check dirty flag and
648 call writepages in lock cancel callback.
650 Severity : enhancement
652 Description: using special macro for print time and cleanup in includes.
656 Description: mds_mfd_close() ASSERTION(rc == 0)
657 Details : In mds_mfd_close(), we need protect inode's writecount change
658 within its orphan write semaphore to prevent possible races.
662 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
666 Frequency : rare, on shutdown ost
667 Description: don't hit live lock with umount ost.
668 Details : shrink_dcache_parent can be in long loop with destroy dentries,
669 use shrink_dcache_sb instead.
673 Description: don't panic with use echo client
674 Details : echo client pass NULL as client nid pointer and this produce null
679 Description: fix build on ppc32
680 Details : compile code with -m64 flag produce wrong object file for ppc32.
684 Description: add message levels for liblustreapi
688 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
689 Details : Move the 'good_osts' check before the 'total_bavail' check. This
690 will result in an -EAGAIN and in the exit call path we call
691 alloc_rr() which will with increasing aggressiveness attempt to
692 aquire precreated objects on the minimum number of required OSCs.
696 Description: Use old size assignment to avoid deadlock
697 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
698 the scheduling while holding a spinlock. We do not need locking
699 for size in ll_update_inode() because size is only updated from
700 the MDS for directories or files without objects, so there is no
701 other place to do the update, and concurrent access to such inodes
702 are protected by the inode lock.
706 Description: resolve "_IOWR redefined" build error on SLES10
710 Description: dump the memory debugging after all modules are unloaded to
711 suppress false negative in conf_sanity test 39
713 Severity : enhancement
715 Description: build kernel-ib packages for OFED 1.3 in our release cycle
720 Description: fix SLES kernel versioning
721 Details : the kernel version for our SLES 10 kernel did not include a "-"
722 before the "smp" at the end. while this was not a problem in
723 general, it did mean that software trying to use the kernel
724 version to try to detect a vendor specific kernel would fail.
725 this was most evident by the OFED build scripts.
729 Description: Don't update lov_desc members until making sure they are valid
730 Details : When updating lov_desc members via proc fs, need fix their
731 validities before doing the real update.
735 Description: don't put request into delay list while invalidate in flight.
736 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
737 import in flight. this produce timeout for invalidate and sometimes
738 can cause stale data.
742 Frequency : on ppc only
743 Description: not convert ost objects for directory because it's not exist.
744 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
748 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
750 Severity : enhancement
752 Description: Update to support 2.6.22.14 vanilla kernel.
756 Frequency : rare, on recovery
757 Description: read procfs can produce deadlock in some situation
758 Details : Holding lprocfs lock which send rpc can produce block for destroy
759 obd objects and this also block reconnect with -EALREADY. This
760 isn't fix all lprocfs bugs - but make it rare.
762 Severity : enhancement
764 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
767 Frequency : frequent on X2 node
769 Description: mdc_set_open_replay_data LBUG
770 Details : Set replay data for requests that are eligible for replay.
774 Description: lustre_mgs: operation 101 on unconnected MGS
775 Details : When MGC is disconnected from MGS long enough, MGS will evict the
776 MGC, and late on MGC cannot successfully connect to MGS and a lot
777 of the error messages complaining that MGS is not connected.
781 Frequency : on network error
782 Description: panic with double free request if network error
783 Details : mdc_finish_enqueue is finish request if any network error ocuring,
784 but it's true only for synchronus enqueue, for async enqueue
785 (via ptlrpcd) this incorrect and ptlrpcd want finish request
788 Severity : enhancement
790 Description: client-side metadata stat-ahead during readdir(directory readahead)
791 Details : perform client-side metadata stat-ahead when the client detects
792 readdir and sequential stat of dir entries therein
795 Frequency : on start mds
797 Description: Implement get_info(last_id) in obdfilter.
800 Frequency : occasional
802 Description: Correctly check stale fid, not start epoch if ost not support SOM
803 Details : open with flag O_CREATE need set old fid in op_fid3 because
804 op_fid2 was overwritten with new generated fid, but mds can answer
805 with one of these two fids and both is not stale. Setattr
806 incorrectly started an epoch and assume will be called
807 done_writing, but without SOM done_writing ever being called.
810 Frequency : rare, depends on device drivers and load
812 Description: MDS or OSS nodes crash due to stack overflow
813 Details : Code changes in 1.8.0 increased the stack usage of some functions.
814 In some cases, in conjunction with device drivers that use a lot
815 of stack the MDS (or possibly OSS) service threads could overflow
816 the stack. One change which was identified to consume additional
817 stack has been reworked to avoid the extra stack usage.
820 Frequency : occasional
822 Description: Do not fail import if osc_interpret_create gets -EAGAIN
823 Details : If osc_interpret_create got -EAGAIN it immediately exits and
824 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
825 oscc_has_objects and see OSC has no objests and call
826 oscc_internal_create to resend create request.
828 Severity : enhancement
830 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
832 Severity : enhancement
834 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
837 Frequency : very rare
839 Description: Fix replay if there is an un-replied request and open
840 Details : In some cases, older replay request will revert the
841 mcd->mcd_last_xid on MDS which is used to record the client's
844 Severity : enhancement
846 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
848 Severity : enhancement
850 Description: Add rhel5 support to HEAD.
852 Serverity : enhancement
854 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
859 Description: Don't allow skipping OSTs if index has been specified.
860 Details : Don't allow skipping OSTs if index has been specified, make
861 locking in internal create lots better.
865 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
866 Details : access to bitfield in structure is always rounded to long
867 and this produce problem with not atomic change any bit.
871 Description: Lustre make rpms failed.
872 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
873 builds Lustre rpms from tarball.
876 Frequency : rare on shutdown ost
878 Description: If llog cancel was not send before clean_exports phase, this can
879 produce deadlock in llog code.
880 Details : If llog thread has last reference to obd and call class_import_put
881 this produce deadlock because llog_cleanup_commit_master wait when
882 last llog_commit_thread exited, but this never success because was
883 called from llog_commit_thread.
887 Description: allow userland application know is lost one of stripes.
888 Details : fill lvb_blocks with error code on ost and return it to
889 application if error flag found.
893 Description: NULL lov_tgts causing MDS oops
894 Details : more safe checks for NULL lov_tgts for avoid oops.
896 Severity : enhancement
898 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
902 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
903 Details : If an OST is down the MDS will hang indefinitely in
904 obd_statfs_rqset() waiting for the statfs data. While for
905 MDS QOS usage of statfs, it should not stuck in waiting.
907 Severity : enhancement
909 Description: remote_acl support
910 Details : Support ACL-based permission check for remote user.
911 Support setfacl/getfacl for remote user with the utils
912 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
913 the system "{s,g}etfacl" utils.
915 Severity : enhancement
917 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
919 Severity : enhancement
921 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
925 Description: make mgs_setparam() handle fsname containing dash
926 Details : fsname containing a dash does not work with lctl conf_param
928 Severity : enhancement
930 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
932 Severity : enhancement
934 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
938 Frequency : rare, at shutdown
939 Description: access already free / zero obd_namespace.
940 Details : if client_disconnect_export was called without force flag set,
941 and exist connect request in flight, this can produce access to
942 NULL pointer (or already free pointer) when connect_interpret
943 store ocd flags in obd_namespace.
947 Frequency : only at startup
948 Description: not alloc memory with spinlock held.
949 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
950 if any spinlock held.
952 Severity : enhancement
954 Description: make lustre randomly fail allocating memory
955 Details : Make lustre randomly failed allocating memory for testing purpose.
957 Severity : enhancement
959 Description: lost problems with lov objid file
960 Details : Fixes some scability and access to not inited memory problems
961 in work with lov objdid file.
966 Description: lfs find does not continue on file error
967 Details : Continue other files processing when a file/dir is absent.
971 Description: Inconsistent usage of lustre_pack_reply()
972 Details : Standardize the usage of lustre_pack_reply() such that it
973 always generate a CERROR on failure.
976 Frequency : occasional
978 Description: MDS hang or stay in waiting lock
979 Details : If client receive lock with CBPENDING flag ldlm need send lock
980 cancel as separate rpc, to avoid situation when cancel request
981 can't processed due all i/o threads stay in wait lock.
985 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
989 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
993 Description: Update to SLES9 kernel-2.6.5-7.287.3.
997 Description: Update to RHEL4 latest kernel.
999 Severity : enhancement
1001 Description: Build SLES10 patchless client fails
1002 Details : The configure was broken by run ./configure with
1003 --with-linux-obj=.... argument for patchless client. When the
1004 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1005 can't search header adequately. Use absolute path such as
1006 -I($LINUX)/include instead.
1010 Description: interrupt oig_wait produce painc on resend.
1011 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1012 private set, and this produce situation when rq_ptlrpcd_data not
1013 copyed to new allocated request and triggered LBUG on assert
1014 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1015 ptlrpcd set if request is changed and can be safety changed to use
1020 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1021 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1022 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1023 these macros and replace them with their difinitions instead.
1027 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1028 Details : Update which_patch & target file for SLES10 latest kernel.
1031 Description: Remove client patches from SLES 10 kernel.
1032 Details : This causes SLES 10 clients to behave as patchless clients
1033 even on a Lustre-patched (server) kernel.
1035 Severity : enhancement
1037 Description: self-adjustable client's lru lists
1038 Details : use adaptive algorithm for managing client cached locks lru
1039 lists according to current server load, other client's work
1040 pattern, memory activities, etc. Both, server and client
1041 side namespaces provide number of proc tunables for controlling
1044 Severity : enhancement
1046 Description: light-weight GSS support
1047 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1048 and reduce performance overhead.
1050 Severity : enhancement
1052 Description: Linux keyring support
1053 Details : Support using service of Linux keyring for Lustre GSS internal
1054 context refresh/cache mechanism.
1058 Description: Fix errors in lfs documentation
1059 Details : Fixes man pages
1063 Description: don't use GFP_* in generic Lustre code.
1064 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1068 Description: obdclass is limited by single OBD_ALLOC(idarray)
1069 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1073 Description: warnings with build patchless client with vanila 2.6.19 and up
1074 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1075 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1079 Frequency : Only for SLES
1080 Description: sanity_quota fail test_1
1081 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1082 file, which makes TSTID[2] inunique.
1086 Frequency : Only for SLES9
1087 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1088 Details : Update target/ChangeLog/which_patch .
1092 Frequency : Only for test_10 in sanity-quota.sh
1093 Description: a bug in quota test code
1094 Details : checking a test flag in a wrong place
1098 Frequency : Only for test_13 in sanity-quota.sh
1099 Description: a bug in quota test script
1100 Details : shouldn't check overfull allocated quota
1104 Description: jbd statistics
1105 Details : Port older jbd statistics patch for sles10
1107 Severity : enhancement
1109 Description: Add printing inode info into message about error in writepage.
1114 Description: reply_lock_interpret crash due to race with it and lock cancel.
1115 Details : Do not replay locks that are being cancelled. Do not reference
1116 locks by their address during replay, just by their handle.
1120 Frequency : When flocks are used.
1121 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1122 Details : Flock locks might destroy just granted lock if it could be merged
1123 with another existing flock, this is done in completion handler,
1124 so teach ldlm_cli_enquque_fini that this is a valid case for
1130 Description: Oops in read and write path when failing to allocate lock.
1131 Details : Check if lock allocation failed and return error back.
1135 Description: lstripe command fails for valid OST index
1136 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1137 instead of lov->desc.ld_active_tgt_count.
1141 Description: sanity.sh failed test 103
1142 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1147 Description: df doesn't work properly if diskfs blocksize != 4K
1148 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1152 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1153 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1155 Severity : enhancement
1157 Description: Adaptive timeouts
1158 Details : RPC timeouts adapt to changing server load and network
1159 conditions to reduce resend attempts and improve recovery time.
1163 Description: llapi_file_create() does not allow some changes
1164 Details : add llapi_file_open() that allows specifying the mode and
1165 open flags, and also returns an open file handle.
1169 Description: merge and cleanup kernel patches.
1170 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1174 Description: Add journal checksum support.(Kernel part)
1175 Details : The journal checksum feature adds two new flags i.e
1176 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1177 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1178 indicates that the commit block contains the checksum for
1179 the blocks described by the descriptor blocks. Now commit
1180 record can be sent to disk without waiting for descriptor
1181 blocks to be written to disk. This behavior is controlled
1182 using JBD2_FEATURE_ASYNC_COMMIT flag.
1186 Description: OSS needs mutliple precreate threads
1187 Details : Add ability to start more than one create thread per OSS.
1191 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1192 Details : execute lfs setstripe on client
1196 Description: mds_obd_create error creating tmp object
1197 Details : When the user sets quota on root, llog will be affected and can't
1198 create files and write files.
1201 Frequency : Always on ia64 patchless client, and possibly others.
1203 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1204 Details : This allows the patchless client to be loaded on architectures
1205 without this export.
1209 Description: RedHat Update kernel for RHEL5
1210 Details : Add the kernel config file for RHEL5.
1214 Description: RedHat Update kernel for RHEL5
1215 Details : Modify the kernel config file more closer RHEL5.
1219 Description: Build failure against Centos5 (RHEL5)
1220 Details : Define PAGE_SIZE when it isn't present.
1224 Description: test_7 on sanity-quota.sh can't be used on separate machines
1225 Details : add facet to handle it
1229 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1230 Details : with NFS, the anon dentry's parent was set to itself in
1231 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1232 corresponding dentry other than use rec->ur_name.
1234 Severity : enhancement
1236 Description: lfs setstripe enhancement
1237 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1241 Description: enable data checksumming by default
1242 Details : enable checksum by default, allow --disable-checksum
1243 configure option and "-o nochecksum" mount option. Checksums
1244 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1248 Description: lustre support for RHEL5
1249 Details : Add support for RHEL5.
1253 Description: Client eviction due to failover config
1254 Details : after a connection loss, the lustre client should attempt to
1255 reconnect to the last active server first before trying the
1256 other potential connections.
1260 Description: when mds and osts use different quota unit(32bit and 64bit),
1261 quota will be released repeatly.
1262 Details : void sending multiple quota reqs to mds, which will keep the
1263 status between the reqs.
1267 Description: osts not allocated evenly to files
1268 Details : change the condition to increase offset_idx
1271 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1273 Description: Data corruption for OSTs that are formatted larger than 2TB
1275 Details : When generating the bio request for lustre file writes the
1276 sector number would overflow a temporary variable before being
1277 used for the IO. The data reads correctly from Lustre (which
1278 will overflow in a similar manner) but other file data or
1279 filesystem metadata may be corrupted in some cases.
1283 Description: Tune the kernel for good SCSI performance.
1284 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1285 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1290 Description: rewrite ext2-derived code in obdclass/uuid.c
1291 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1292 from scratch preserving functionality.
1296 Description: "lfs find -obd UUID" prints directories
1297 Details : "lfs find -obd UUID" will return all directory names instead
1298 of just file names. It is incorrect because the directories
1299 do not reside on the OSTs.
1303 Description: use i_size_read and i_size_write in 2.6 port
1304 Details : replace inode->i_size access with i_size_read/write()
1307 Frequency : when removing large files
1309 Description: scheduling issue during removal of large Lustre files
1310 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1311 It causes scheduling issues when removing large files (17TB in the
1315 Frequency : only with liblustre clients on XT3
1317 Description: evictions taking too long
1318 Details : allow llrd to evict clients directly on OSTs
1321 Frequency : only on ppc
1323 Description: /proc/fs/lustre/devices broken on ppc
1324 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1325 We should make sure the type of 'index' is loff_t and then cast
1326 explicitly as needed below. Do not assign an explicitly cast
1330 Frequency : only for rhel5
1332 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1333 Details : Modify the target file & which_kernel.
1335 Severity : enhancement
1337 Description: omit set fsid for export NFS
1338 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1339 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1343 Frequency : Always, for kernels after 2.6.16
1344 Description: Fix warning idr_remove called for id=.. which is not allocated.
1345 Details : Last kernels save old s_dev before kill super and not allow
1346 to restore from callback - restore it before call kill_anon_super.
1350 Description: buffer overruns could theoretically occur
1351 Details : llapi_semantic_traverse() modifies the "path" argument by
1352 appending values to the end of the origin string, and a
1353 overrun may occur. Adding buffer overrun check in liblustreapi.
1357 Description: Fix error on 'ls .' at the top of the Lustre mount.
1358 Details : Don't revalidate dentry if it is a root dentry.
1362 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1363 Details : Modify vm-tunables-rhel4.patch.
1367 Description: Kernel config for 2.6.18-vanilla.
1368 Details : Modify targets/2.6-vanilla.target.in.
1369 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1370 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1371 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1372 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1376 Description: improve handling recoverable errors
1377 Details : if request processig with error which can be recoverable on server
1378 request should be resend, otherwise page released from cache and
1383 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1384 Details : Modify target file & which_patch.
1385 A flaw was found in the IA32 system call emulation provided
1386 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1387 value could be stored in the %RAX register, which could trigger an
1388 out-of-bounds system call table access. An untrusted local user
1389 could exploit this flaw to run code in the kernel
1390 (ie a root privilege escalation). (CVE-2007-4573).
1394 Description: O_DIRECT bypasses client statistics.
1395 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1396 read_ahead_stats were updated. copied the stats section from
1397 osc_send_oap_rpc() into async_internal().
1401 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1405 Description: change order of libsysio includes
1406 Details : '#include sysio.h' should always come before '#include xtio.h'
1410 Description: handle "serious error: objid * already exists" more gracefully
1411 Details : If LAST_ID value on disk is smaller than the objects existing in
1412 the O/0/d* directories, it indicates disk corruption and causes an
1413 LBUG(). If the object is 0-length, then we should use the existing
1414 object. This will help to avoid a full fsck in most cases.
1416 Severity : enhancement
1418 Description: adapt the lustre_config script to support the upgrade case
1419 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1420 server targets to 1.6 in parallel.
1424 Description: To avoid grant space > avaible space when the disk is almost
1425 full. Without this patch you might see the error "grant XXXX >
1426 available" or some LBUG about grant, when the disk is almost
1428 Details : In filter_check_grant, for non_grant cache write, we should
1429 check the left space by if (*left > ungranted + bytes), instead
1430 of (*left > ungranted), because only we are sure the left space
1431 is enough for another "bytes", then the ungrant space should be
1432 increase. In client, we should update cl_avail_grant only there
1433 is OBD_MD_FLGRANT in the reply.
1437 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1438 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1439 described in CVE-2007-4573. This problem could allow untrusted
1440 local users to gain root access.
1443 Frequency : when using O_DIRECT and quotas
1445 Description: Incorrect file ownership on O_DIRECT output files
1446 Details : block usage reported by 'lfs quota' does not take into account
1447 files that have been written with O_DIRECT.
1452 Description: touch file failed when fs is not full
1453 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1454 otherwise we can get ENOSP while fs is not full.
1458 Description: parallel lock callbacks
1459 Details : Instead of sending blocking and completion callbacks as separated
1460 requests, adding them to a set and sending in parallel.
1463 Frequency : only for Cray XT3
1464 Bugzilla : 12829/13455
1465 Description: Changing primary group doesn't change the group lustre assigns to
1467 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1468 overridden with the primary group provided by the group upcall,
1469 whereas the supplied fsgid can be trusted if it is in the list of
1470 supplementary groups returned by the group upcall.
1472 Severity : enhancement
1474 Description: Allow masking D_WARNING, D_ERROR messages from console
1475 Details : Console messages can now be disabled via lnet.printk.
1479 Description: User code with malformed file open parameter crashes client node
1480 Details : Before packing join_file req, all the related reference should be
1481 checked carefully in case some malformed flags cause fake
1482 join_file req on client.
1486 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1487 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1488 mode is equal to the granted mode. However, it should also check
1489 that LDLM_FL_CP_REQD is not set.
1493 Description: Heavy nfs access might result in deadlocks
1494 Details : After ELC code landed, it is now improper to enqueue any mds
1495 locks under och_sem, because enqueue might want to decide to
1496 cancel open locks for same inode we are holding och_sem for.
1500 Description: Client eviction while running blogbench
1501 Details : A lot of unlink operations with concurrent I/O can lead to a
1502 deadlock causing evictions. To address the problem, the number of
1503 oustanding OST_DESTROY requests is now throttled to
1504 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1505 callbacks are processed in priority.
1509 Description: enable ACLs on MDS by default
1510 Details : ACLs must be enabled on MDS by default.
1513 Frequency : PPC/PPC64 only
1515 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1516 Details : fix duplicated definitions between asm-ppc64/types.h and
1517 lustre_types.h on PPC.
1520 Frequency : PPC/PPC64 only
1522 Description: asm-ppc/segment.h does not exist
1523 Details : fix compile issue on PPC.
1527 Description: better handle error messages in extents code
1530 Frequency : RHEL4 only
1532 Description: mkfs is very slow on IA64/RHEL4
1533 Details : A performance regression has been discovered in the MPT Fusion
1534 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1535 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1536 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1538 Severity : enhancement
1540 Description: SNMP support enhancement
1541 Details : Adding total number of sampled request for an MDS node in snmp
1544 Severity : enhancement
1546 Description: Optimize ldlm waiting list processing for PR extent locks
1547 Details : When processing waiting list for read extent lock and meeting read
1548 lock that is same or wider to it that is not contended, skip
1549 processing rest of the list and immediatelly return current
1550 status of conflictness, since we are guaranteed there are no
1551 conflicting locks in the rest of the list.
1555 Description: Time out and refuse to reconnect
1556 Details : When the failover node is the primary node, it is possible
1557 to have two identical connections in imp_conn_list. We must
1558 compare not conn's pointers but NIDs, otherwise we can defeat
1559 connection throttling.
1563 Description: port llog fixes from b1_6 into HEAD
1564 Details : Port llog reference couting and some llog cleanups from b1_6
1565 (bug 10800) into HEAD, for protect from panic and access to already
1566 free llog structures.
1570 Description: Detect stride IO mode in read-ahead
1571 Details : When a client does stride read, read-ahead should detect that and
1572 read-ahead pages according to the detected stride pattern.
1576 Description: data checksumming impacts single node performance
1577 Details : add support for several checksum algorithm. Currently, only CRC32
1578 and Adler-32 are supported. The checksum type can be changed on
1579 the fly via /proc/fs/lustre/osc/*/checksum_type.
1583 Description: use adler32 for page checksums
1584 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1589 Description: build for x2 fails
1590 Details : fix compile issue on Cray systems.
1594 Description: Properly match for duplicate locks
1595 Details : Due to different lock order from skiplists code, we need to
1596 traverse entire list for now
1599 Frequency : only on PPC/SLES10
1601 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1602 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1603 be defined. Add a hack in mkfs_lustre.c to work around this compile
1608 Description: LASSERT on MDS when client holding flock lock dies
1609 Details : ldlm pool logic depends on number of granted locks equal to
1610 number of released locks which is not true for flock locks, so
1611 just exclude such locks from consideration.
1615 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1616 Details : Use fsfilt_sync() for both the whole filesystem sync and
1617 individual file sync to eliminate dangerous inode locking
1618 with I_LOCK that can lead to a deadlock.
1622 Description: performance in 1.6.3
1623 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1624 sure that 1MB requests are not fragmented by the block layer.
1626 Severity : enhancement
1628 Description: organize the server-side client stats on per-nid basis
1629 Details : Change the structure of stats under obdfilter and mds to
1637 The "uuid"s file would list the uuids of _active_ exports.
1638 And the clear entry is to clear all stats and stale nids.
1640 Severity : enhancement
1642 Description: eliminate client locks in face of contention
1643 Details : file contention detection and lockless i/o implementation
1644 for contended files.
1648 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1651 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1653 Description: MDS may lose file striping (and hence file data) in some cases
1654 Details : If there are additional extended attributes stored on the MDS,
1655 in particular ACLs, SELinux, or user attributes (if user_xattr
1656 is specified for the client mount options) then there is a risk
1657 of attribute loss. Additionally, the Lustre file striping
1658 needs to be larger than default (e.g. striped over all OSTs),
1659 and an additional attribute must be stored initially in the
1660 inode and then increase in size enough to be moved to the
1661 external attribute block (e.g. ACL growing in size) for file
1666 Description: skiplist implementation simplification
1667 Details : skiplists are used to group compatible locks on granted list
1668 that was implemented as tracking first and last lock of each
1669 lock group the patch changes that to using doubly linked lists
1673 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1674 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1675 because LASSERT on both the data supplied by a client, and the
1676 data on disk is dangerous and incorrect.
1678 Severity : enhancement
1680 Description: Slow truncate/writes to huge files at high offsets.
1681 Details : Directly associate cached pages to lock that protect those pages,
1682 this allows us to quickly find what pages to write and remove
1683 once lock callback is received.
1687 Description: more ldlm soft lockups
1688 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1689 starve other threads from the resource lock for a long time in
1690 case of long waiting queue, so change the debug level from
1691 D_OTHER to the less frequently used D_INFO.
1693 Severity : enhancement
1695 Description: add -gid, -group, -uid, -user options to lfs find
1699 Description: Hung threads in invalidate_inode_pages2_range
1700 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1701 once one is completed. As a result, some RPCs are stuck in the
1702 queue and are never sent.
1706 Description: filter threads hungs on waiting journal commit
1707 Details : Cleanup filter group llog code, then only filter group llog will
1708 be only created in the MDS/OST syncing process.
1712 Description: Procfs and llog threads access destoryed import sometimes.
1713 Details : Sync the import destoryed process with procfs and llog threads by
1714 the import refcount and semaphore.
1716 Severity : enhancement
1718 Description: openlock cache of b1_6 port to HEAD
1723 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1724 Details : If insertion of an extent fails, then discard the inode
1725 preallocation and free data blocks else it can lead to duplicate
1730 Description: don't always update ctime in ext3_xattr_set_handle()
1731 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1732 In some cases the ctime should not be updated, for example for
1733 2.0->1.8 compatibility it is necessary to delete an xattr and it
1734 should not update the ctime.
1738 Bugzilla : 15713/16362
1739 Description: Assertion in iopen_connect_dentry in 1.6.3
1740 Details : looking up an inode via iopen with the wrong generation number can
1741 populate the dcache with a disconneced dentry while the inode
1742 number is in the process of being reallocated. This causes an
1743 assertion failure in iopen since the inode's dentry list contains
1744 both a connected and disconnected dentry.
1748 Description: assertion failure in ldlm_handle2lock()
1749 Details : fix a race between class_handle_unhash() and class_handle2object()
1750 introduced in lustre 1.6.5 by bug 13622.
1755 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1756 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1761 Description: oops in sd_iostats_seq_show()
1762 Details : unloading/reloading the scsi low level driver triggers a kernel
1763 bug when trying to access the sd iostat file.
1768 Description: Kernel panics during QLogic driver reload
1769 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1770 patch, causing memory corruption.
1775 Description: journal_dev option does not work in b1_6
1776 Details : pass mount option during pre-mount.
1778 Severity : enhancement
1780 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1781 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1782 extent information of a file. It can be used to map logical blocks
1783 in a file to physical blocks in the block device.
1787 Description: LDLM soft lockups - improvement
1788 Details : It is be possible to send the lock handle along with each read
1789 or write request because the client is already doing a lock match
1790 itself so there isn't any reason the OST should have to re-do that
1794 Frequency : only on Cray X2
1796 Description: X2 build failures
1797 Details : fix build failures on Cray X2.
1801 Description: xid & resent requests
1802 Details : Initialize RPC XID from clock at startup (randomly if clock is
1805 Severity : enhancement
1807 Description: Add lustre_start utility to start or stop multiple Lustre servers
1812 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1813 Details : In case of memory pressure, list_del() can be called twice on
1814 req->rq_history_list, causing a kernel oops.
1818 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1819 Details : fix stack overflow in the distributed lock manager by defering
1820 export eviction after a failed AST to the elt thread instead of
1821 handling it in the dlm interpret routine.
1825 Description: Convert some comments to new format.
1826 Details : Update documenting comments to match doxygen conventions.
1830 Description: Grammar fixes.
1831 Details : A couple of trivial spelling fixes.
1835 Description: OSD_COUNTERS-mandatory
1836 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1837 to simplify the code.
1841 Description: simplify cmm_device freeing logic.
1842 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1846 Description: Add lockdep support to dt_object_operations locking interface.
1847 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1848 parameter indicating lock ordering. Update mdd code to use new
1853 Description: Introduce failloc constants for lockless IO tests.
1854 Details : Add two new failloc constants to test lockless IO. Only one of
1855 them in implemented---another is checked in yet to be landed
1860 Description: Add lockdep support for inode mutex.
1861 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1862 in the situations where Lustre has to lock more than one inode
1867 Description: Add optional invariants checking support.
1868 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1869 configure switch. This macro is to be used for consistency and
1870 sanity checks that are too expensive to be left in `production'
1875 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1876 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1881 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1882 Details : Use spin_lock_nested() in (the only) situation where more than
1883 one ldlm_lock is locked simultaneously. Also, fix possible
1884 dead-lock in ldlm_lock_change_resource() by enforcing particular
1889 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1890 Details : Instead of specifying each ldlm_lock call-back through separate
1891 parameter, wrap them into struct ldlm_callback_suite.
1895 Description: Kill join_lru obd method and its callers.
1896 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1897 in memory---a requirement implemented through ->o_join_lru() obd
1898 method in HEAD. Remove this method and its users.
1902 Description: Add asynchronous ldlm ENQUEUE completion handler.
1903 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1904 case that stock ldlm_completion_ast() cannot handle as it waits
1905 until lock is granted. Introduce new ldlm_completion_ast_async()
1906 for this. Also comment ldlm_completion_ast().
1910 Description: ldlm_error <-> errno conversion.
1911 Details : Add functions to map (rather arbitrary) between LDLM error codes
1912 and standard errno values. CLIO needs this to prevent LDLM specific
1913 constants from escaping ldlm and osc.
1917 Description: Kill unused ldlm_handle2lock_ns() function.
1918 Details : Kill unused ldlm_handle2lock_ns() function.
1922 Description: Add lu_ref support to ldlm_lock
1923 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1924 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1925 and ldlm_resource. LDLM interface has to be changed, because code that
1926 releases a reference on a lock, has to "know" what reference this is.
1927 In the most frequent case
1929 lock = ldlm_handle2lock(handle);
1931 LDLM_LOCK_PUT(lock);
1933 no changes are required. When any other reference (received _not_ from
1934 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1935 instead of LDLM_LOCK_PUT().
1937 Arguably, changes are pervasive, and interface requires some discipline
1938 for proper use. On the other hand, it was very instrumental in finding
1939 a few leaked lock references.
1943 Description: Add ldlm_lock_addref_try().
1944 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1945 attempts to addref a lock that might be being canceled
1950 Description: Add ldlm_weigh_callback().
1951 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1952 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1953 hard-coded `number of pages' logic.
1957 Description: Add lockdep annotations to llog code.
1958 Details : Use appropriately tagged _nested() locking calls in the places
1959 where llog takes more than one ->lgh_lock lock.
1963 Description: Add loi_kms_set().
1964 Details : Wrap kms updates into a helper function.
1968 Description: Constify instances of struct lsm_operations.
1969 Details : Constify instances of struct lsm_operations.
1973 Description: lu_conf support.
1974 Details : On a server, a file system object is uniquely identified
1975 by a fid, which is sufficient to locate and load all object
1976 state (inode). On a client, on the other hand, more data are
1977 necessary instantiate an object. Change lu_object_find() and
1978 friends to take additional `lu_conf' argument describing object.
1979 Typically this includes layout information.
1983 Description: lu_context fixes.
1984 Details : Introduce new lu_context functions that are needed on the client
1985 side, where some system threads (ptlrpcd) are shared by multiple
1986 modules, and so cannot be stopped during module shutdown.
1990 Description: Add start and stop methods to lu_device_type_operations.
1991 Details : Introduce two new methods in lu_device_type_operations, that are
1992 invoked when first instance of a given type is created and last one
1993 is destroyed respectively. This is need by CLIO.
1997 Description: Add lu_ref support to struct lu_device.
1998 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1999 track leaked references.
2003 Description: Introduce lu_kmem_descr.
2004 Details : lu_kmem_descr and its companion interface allow to create
2005 and destroy a number of kmem caches at once.
2009 Description: Fix lu_object finalization race.
2010 Details : Fix a race between lu_object_find() finding an object and its
2011 concurrent finalization. This race is (most likely) not possible
2012 on the server, but might happen on the client.
2016 Description: Introduce lu_ref interface.
2017 Details : lu_ref is a debugging module allowing to track references to
2018 a given object. It is quite cpu expensive, and has to be
2019 explicitly enabled with --enable-lu_ref. See usage description
2024 Description: Factor lu_site procfs stats into a separate function.
2025 Details : Separate lu_site stats printing code into a separate function
2026 that can be reused on a client.
2030 Description: Constify instances of struct {lu,dt,md}_device_operations.
2031 Details : Constify instances of struct {lu,dt,md}_device_operations.
2035 Description: Introduce struct md_site and move meta-data specific parts of
2036 struct lu_site here.
2037 Details : Move md-specific fields out of struct lu_site into special struct
2038 md_site, so that lu_site can be used on a client.
2042 Description: Kill mdd_lov_destroy().
2043 Details : Remove unused mdd code.
2047 Description: Add st_block checking to multistat.c.
2048 Details : Add st_block checking to multistat.c.
2052 Description: Add lu_ref support to struct obd_device.
2053 Details : Add lu_ref tracking to obd_device.
2057 Description: Kill obd_set_fail_loc().
2058 Details : Remove unused code.
2062 Description: Add special type for ptlrpc_request interpret functions.
2063 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2064 there. Actual usage will be in CLIO.
2068 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2069 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2070 doesn't work with lockdep.
2074 Description: Add rwv.c test program.
2075 Details : New testing program exercising readv(2) and writev(2) (Qian).
2079 Description: Add sendfile.c test program.
2080 Details : New testing program exercising sendfile(2) (Jay).
2084 Description: Ratelimit a message that can be very frequent.
2085 Details : Ratelimit a memory allocation failure message that can
2090 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2091 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2095 Description: Liblustre build fixes.
2096 Details : Liblustre build fixes.
2100 Description: libcfs: add cfs_{need,cond}_resched() interface.
2101 Details : libcfs: add cfs_{need,cond}_resched() definition and
2102 implementations for Linux, NT, and liblustre.
2104 Severity : enhancement
2106 Description: More exported tunables for mballoc
2107 Details : Add support for tunable preallocation window and new tunables for
2108 large/small requests
2112 Description: Detect corruption of block bitmap and checking for preallocations
2113 Details : Checks validity of on-disk block bitmap. Also it does better
2114 checking of number of applied preallocations. When corruption is
2115 found, it turns filesystem readonly to prevent further corruptions.
2119 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2120 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2121 window to 0. If it does hit cache miss, and read-pattern is still
2122 stride-io mode, does not reset the stride window, but also does not
2123 increase the stride window length in this case.
2127 Frequency : only for big-endian servers
2128 Description: Check if system is big-endian while mounting fs with extents feature
2129 Details : Mounting a filesystem with extents feature will fail on big-endian
2130 systems since ext3-based ldiskfs is not supported on big-endian
2131 systems. This can be over-riden with "bigendian_extents" mount option.
2133 Severity : enhancement
2135 Description: The root squash functionality
2136 Details : A security feature, which is to prevent users from being able
2137 to mount lustre on their desktop, run as root, and delete
2138 all of the files in the filesystem. The goal is accomplished by
2139 remapping user id (UID) and group id (GID) of the root user to
2140 a UID and GID specified by the system administartor via Lustre
2141 configuration management server (MGS). The functionality also
2142 allows to specify sets of clients for which the remapping does
2147 Description: Excessive recovery window
2148 Details : With AT enabled, the recovery window can be excessively long (6000+
2149 seconds). To address this problem, we no longer use
2150 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2151 timeout no longer depends on the service time, it is set to
2152 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2153 time via pb_service_time.
2157 Description: Watchdog triggered on MDS failover
2158 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2159 the OSTs know that the MDS "UUID" can be reused for the same export
2160 from a different NID, so we do not need to wait for the export to be
2164 Frequency : rare, only if using MMP with Linux RAID
2166 Description: MMP doesn't work with Linux RAID
2167 Details : While using HA for Lustre servers with Linux RAID, it is possible
2168 that MMP will not detect multiple mounts. To make this work we
2169 need to unplug the device queue in RAID when the MMP block is being
2170 written. Also while reading the MMP block, we should read it from
2171 disk and not the cached one.
2173 Severity : enhancement
2175 Description: open file using fid
2176 Details : A file can be opened using just its fid, like
2177 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2180 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2183 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2184 Details : add proper locking for creation of jbd_4k slab cache
2188 Description: MMP check in ext3_remount() fails without displaying any error
2189 Details : When multiple mount protection fails during remount, proper error
2192 Severity : enhancement
2194 Description: Allow stripe size to be up to 4G-64k
2195 Details : Fix math logic to allow large stripe sizes.
2199 Description: add check for >8TB ldiskfs filesystems
2200 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2201 Don't allow >8TB ldiskfs filesystems to be mounted without
2202 force_over_8tb mount option
2207 Descriptoin: OSS complains "too many blocks" and crashes
2208 Details : In filter_commitrw_write():
2209 1. lquota_chkquota();
2210 2. obdo_from_inode();
2211 3. lquota_pending_commit();
2212 "2." has changed uid and gid in this case, then caused
2213 lquota_chkquota() and lquota_pending_commit() to use
2214 different uid and gid. That is the root of the bug.
2218 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2219 Details : Detect and print out the 'stripe count = -1' case properly.
2221 --------------------------------------------------------------------------------
2223 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2225 * Support for kernels:
2226 2.6.5-7.283 (SLES 9),
2227 2.6.9-55.EL (RHEL 4),
2228 2.6.16.46-0.14 (SLES 10),
2229 2.6.18.8 vanilla (kernel.org)
2230 * Client support for unpatched kernels:
2231 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2232 2.6.16 - 2.6.22 vanilla (kernel.org)
2233 * Due to recently discovered recovery problems, we do not recommend
2234 using patchless RHEL 4 clients with this or any earlier release.
2235 * Recommended e2fsprogs version: 1.39.cfs8
2236 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2237 * Starting with this release, the ldiskfs backing filesystem required
2238 by Lustre is now in its own package, lustre-ldiskfs. This package
2239 should be installed. It is versioned separately from Lustre and
2240 may be released separately in future.
2244 Description: block reactivating mgc import until all deactivates complete
2245 Details : Fix race when failing back MDT/MGS to itself (testing)
2248 Frequency : at statup only
2250 Description: mds_lov_synchronize race leads to various problems
2251 Details : simultaneous MDT->OST connections at startup can cause the
2252 sync to abort, leaving the OSC in a bad state.
2254 Severity : enhancement
2256 Description: add optional extra BUILD_VERSION info
2257 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2258 to override the lustre version.
2261 Frequency : 2.6.18 servers only
2263 Description: ll_kern_mount() doesn't release the module reference
2264 Details : The ldiskfs module reference count never drops down to 0
2265 because ll_kern_mount() doesn't release the module reference.
2270 Description: server LBUG when using old ost_num_threads parameter
2271 Details : Accept the old ost_num_threads parameter but warn that it
2272 is deprecated, and fix an off-by-one error that caused an LBUG.
2277 Description: Transient SCSI error results in persistent IO issue
2278 Details : iobuf->dr_error is not reinitialized to 0 between two
2282 Frequency : sometimes when underlying device returns I/O errors
2284 Description: OSTs not going read-only during write failures
2285 Details : OSTs are not remounted read-only when the journal commit threads
2286 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2287 instead of the ext3 wrappers.
2291 Description: poor connect scaling with increasing client count
2292 Details : Don't run filter_grant_sanity_check for more than 100 exports
2293 to improve scaling for large numbers of clients.
2296 Frequency : SLES10 only
2298 Description: sanity-quota.sh quotacheck failed: rc = -22
2299 Details : Quotas cannot be enabled on SLES10.
2302 Frequency : liblustre clients only
2304 Description: getdirentries does not give error when run on compute nodes
2305 Details : getdirentries does not fail when the size specified as an argument
2306 is too small to contain at least one entry
2308 Severity : enhancement
2310 Description: Add LNET router traceability for debug purposes
2311 Details : If a checksum failure occurs with a router as part of the
2312 IO path, the NID of the last router that forwarded the bulk data
2313 is printed so it can be identified.
2318 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2319 Details : Due to a race condition, liblustre clients were occasionally
2320 evicted incorrectly.
2322 Severity : enhancement
2324 Description: lfs setstripe use optional parameters instead of postional
2327 Severity : enhancement
2329 Description: Nanosecond timestamp support for ldiskfs
2330 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2331 resolution timestamps. There is not yet support for this at
2332 the Lustre filesystem level.
2335 Frequency : during server recovery
2337 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2338 Details : request with rq_no_resend flag not awake l_wait_event if they get
2342 Frequency : nfs export on patchless client
2344 Description: connectathon hang when test nfs export over patchless client
2345 Details : Disconnected dentry cannot be found with lookup, so we do not need
2346 to unhash it or make it invalid
2349 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2350 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2351 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2356 Description: open req refcounting wrong on reconnect
2357 Details : If reconnect happened between getting open reply from server and
2358 call to mdc_set_replay_data in ll_file_open, we will schedule
2359 replay for unreferenced request that we are about to free.
2360 Subsequent close will crash in variety of ways.
2361 Check that request is still eligible for replay in
2362 mdc_set_replay_data().
2367 Description: disable writes to filesystem when reading health_check file
2368 Details : the default for reading the health_check proc file has changed
2369 to NOT do a journal transaction and write to disk, because this
2370 can cause reads of the /proc file to hang and block HA state
2371 checking on a healthy but otherwise heavily loaded system. It
2372 is possible to return to the previous behaviour during configure
2373 with --enable-health-write.
2375 Severity : enhancement
2377 Description: 64-bit inode version
2378 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2379 to the inode. This will be required for version-based recovery.
2384 Description: MDS fails to start if a duplicate client export is detected
2385 Details : in some rare cases it was possible for a client to connect to
2386 an MDS multiple times. Upon recovery the MDS would detect this
2387 and fail during startup. Handle this more gracefully.
2389 Severity : enhancement
2391 Description: Add -o localflock option to simulate old noflock
2393 Details : This will achieve local-only flock/fcntl locks
2399 Description: log_commit_thread vs filter_destroy race leads to crash
2400 Details : Take import reference before releasing llog record semaphore
2405 Description: Wrong request locking in request set processing
2406 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2407 imp_delayed_list, in this place should be used imp_lock.
2410 Frequency : when reconnection
2412 Description: Grant Leak when osc reconnect to OST
2413 Details : When osc reconnect ost, OST(filter) should check whether it
2414 should grant more space to client by comparing fed_grant and
2415 cl_avail_grant, and return the granted space to client instead
2416 of "new granted" space, because client will call osc_init_grant
2417 to update the client grant space info.
2420 Frequency : when client reconnect to OST
2422 Description: Grant Leak when osc do resend and replay bulk write
2423 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2424 bulk write request, because the grant info while be sync between
2425 OSC and OST when reconnect, and we should ignore the grant info
2426 these of resend/replay write req.
2431 Description: Grant space more than avaiable left space sometimes.
2432 Details : When then OST is about to be full, if two bulk writing from
2433 different clients came to OST. Accord the avaliable space of the
2434 OST, the first req should be permitted, and the second one
2435 should be denied by ENOSPC. But if the seconde arrived before
2436 the first one is commited. The OST might wrongly permit second
2437 writing, which will cause grant space > avaiable space.
2440 Frequency : when client is evicted
2442 Description: Grant might be wrongly erased when osc is evicted by OST
2443 Details : when the import is evicted by server, it will fork another
2444 thread ptlrpc_invalidate_import_thread to invalidate the
2445 import, where the grant will be set to 0. While the original
2446 thread will update the grant it got when connecting. So if
2447 the former happened latter, the grant will be wrongly errased
2448 because of this race.
2453 Description: Checking Stale with correct fid
2454 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2455 to check whether it is stale, because sometimes, we want the
2456 enqueue happened anyway, and op_data.fid2 will not be initialized.
2458 Severity : enhancement
2460 Description: update patchless client
2461 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2464 Frequency : only with 2.4 kernel
2466 Description: random memory corruption
2467 Details : size of struct ll_inode_info is to big for union inode.u and this
2468 can be cause of random memory corruption.
2473 Description: Memory leak in recovery
2474 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2475 It should also check obd_fail before fsfilt_start, otherwise if
2476 fsfilt_start return -EROFS,(failover mds during mds recovery).
2477 then the req will return with repmsg->transno = 0 and rc = EROFS.
2478 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2479 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2480 be freed no matter whether fsfilt_commit success or not.
2483 Frequency : only with huge count clients
2485 Description: Prevents from taking the superblock lock in llap_from_page for
2487 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2488 for llap_from_page call in ll_removepage prevents from taking the
2489 superblock lock for a soon died page.
2494 Description: Not check open intent error before release open handle
2495 Details : in some rare cases, the open intent error is not checked before
2496 release open handle, which may cause
2497 ASSERTION(open_req->rq_transno != 0), because it tries to release
2498 the failed open handle.
2503 Description: Set cat log bitmap only after create log success.
2504 Details : in some rare cases, the cat log bitmap is set too early. and it
2505 should be set only after create log success.
2509 Description: Accessing a block bevice can re-enable I/O when Lustre is
2510 tearing down a device.
2511 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2516 Description: service threads may hog cpus when there are a lot of requests
2517 Details : Insert cond_resched to give other threads a chance to use some CPU
2522 Description: the cat log was not initialized in recovery
2523 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2524 unlink log on mds will not be initialized until mds post
2525 recovery. And also in mds post recovery, the unlink log will
2526 initialization will be done asynchronausly, so there will be race
2527 between add unlink log and unlink log initialization.
2531 Description: brw_stats were being printed incorrectly
2532 Details : brw_stats were being printed as log2 but all of them were not
2533 recorded as log2. Also remove some code duplication arising from
2534 filter_tally_{read,write}.
2538 Frequency : rare, only in recovery.
2539 Description: ASSERTION(req->rq_type != LI_POISON) failed
2540 Details : imp_lock should be held while iterating over imp_sending_list for
2541 prevent destroy request after get timeout in ptlrpc_queue_wait.
2545 Description: replay-single.sh test 52 fails
2546 Details : A lock's skiplist need to be cleanup when it being unlinked
2547 from its resource list.
2551 Description: Short directio read returns full requested size rather than
2553 Details : Direct I/O operations should return actual amount of bytes
2554 transferred rather than requested size.
2556 Severity : enhancement
2558 Description: metadata RPC reduction (e.g. for rm performance)
2559 Details : decrease the amount of synchronous RPC between clients and servers
2560 by canceling conflicing lock before the operation on the client
2561 and packing thier handles into the main operation RPC to server.
2563 Severity : enhancement
2565 Description: Async OSC create to avoid the blocking unnecessarily.
2566 Details : If a OST has no remain object, system will block on the creating
2567 when need to create a new object on this OST. Now, ways use
2568 pre-created objects when available, instead of blocking on an
2569 empty osc while others are not empty. If we must block, we block
2570 for the shortest possible period of time.
2574 Description: improve handling recoverable errors
2575 Details : if request processig with error which can be recoverable on server
2576 request should be resend, otherwise page released from cache and
2579 Severity : enhancement
2581 Description: refine locking for avoid write wrong info into lov_objid file
2582 Details : fix possible races with add new target and write/update data in
2585 --------------------------------------------------------------------------------
2587 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2592 Frequency : on some architectures
2594 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2595 Details : Uninitialized flags sometimes cause configuration commands to
2599 Frequency : patchless clients only
2601 Description: missing __iget() symbol export
2602 Details : The __iget() symbol export is missing. To avoid the need for
2603 this on patchless clients the deathrow inode reaper is turned
2604 off, and we depend on the VM to clean up old inodes. This
2605 dependency was during via the fix for bug 12181.
2607 --------------------------------------------------------------------------------
2609 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2611 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2612 INTEROPERATE with older versions automatically. In many cases a
2613 special upgrade step is needed. Please read the
2614 user documentation before upgrading any part of a live system.
2615 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2616 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2617 * WARNING: Lustre configuration and startup changes are required with
2618 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2622 Severity : enhancement
2624 Description: MountConf
2625 Details : Lustre configuration is now managed via mkfs and mount
2626 commands instead of lmc and lconf. New obd types (MGS, MGC)
2627 are added for dynamic configuration management. See
2628 https://mail.clusterfs.com/wikis/lustre/MountConf for
2631 Severity : enhancement
2633 Description: dynamic OST addition
2634 Details : OSTs can now be added to a live filesystem
2636 Severity : enhancement
2638 Description: startup order invariance
2639 Details : MDTs and OSTs can be started in any order. Clients only
2640 require the MDT to complete startup.
2642 Severity : enhancement
2644 Description: parallel, asynchronous orphan cleanup
2645 Details : orphan cleanup is now performed in separate threads for each
2646 OST, allowing parallel non-blocking operation.
2648 Severity : enhancement
2650 Description: optimized stripe assignment
2651 Details : stripe assignments are now made based on ost space available,
2652 ost previous usage, and OSS previous usage, in order to try
2653 to optimize storage space and networking resources.
2655 Severity : enhancement
2657 Description: Permanently set tunables
2658 Details : All writable /proc/fs/lustre tunables can now be permanently
2659 set on a per-server basis, at mkfs time or on a live system.
2661 Severity : enhancement
2663 Description: Lustre message v2
2664 Details : Add lustre message format v2.
2666 Severity : enhancement
2668 Description: client OST exclusion list
2669 Details : Clients can be started with a list of OSTs that should be
2670 declared "inactive" for known non-responsive OSTs.
2674 Description: SPEC SFS validation failure on NFS v2 over lustre.
2675 Details : Changes the blocksize for regular files to be 2x RPC size,
2676 and not depend on stripe size.
2678 Severity : enhancement
2680 Description: Multiple MD RPCs in flight.
2681 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2682 To avoid overly-overloading MDS, introduce a limit on number of
2683 MDS RPCs in flight for a single client and add /proc controls
2684 to adjust this limit.
2686 Severity : enhancement
2688 Description: client read/write statistics
2689 Details : Add client read/write call usage stats for performance
2690 analysis of user processes.
2691 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2692 file access. extents_stats shows chunk size distribution.
2693 extents_stats_per_process show chunk size distribution per
2696 Severity : enhancement
2698 Description: mds statistics
2699 Details : Add detailed mds operations statistics in
2700 /proc/fs/lustre/mds/*/stats.
2704 Description: Failure of copying files with lustre special EAs.
2705 Details : Client side always return success for setxattr call for lustre
2706 special xattr (currently only "trusted.lov").
2711 Description: Refcount LNET uuids
2712 Details : The global LNET uuid list grew linearly with every startup;
2713 refcount repeated list entries instead of always adding to
2716 Severity : enhancement
2718 Description: Dynamic service threads
2719 Details : Within a small range, start extra service threads
2720 automatically when the request queue builds up.
2723 Frequency : mixed-endian client/server environments
2725 Description: mixed-endian crashes
2726 Details : The new msg_v2 system had some failures in mixed-endian
2729 Severity : enhancement
2731 Description: Easy OST removal
2732 Details : OSTs can be permanently deactivated with e.g. 'lctl
2733 conf_param lustre-OST0001.osc.active=0'
2735 Severity : enhancement
2737 Description: MGS proc entries
2738 Details : Added basic proc entries for the MGS showing what filesystems
2741 Severity : enhancement
2743 Description: provide MGS failover
2744 Details : Added config lock reacquisition after MGS server failover.
2746 Severity : enhancement
2748 Description: add Linux 2.4 support
2749 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2753 Description: a large application tries to do I/O to the same resource and dies
2754 in the middle of it.
2755 Details : Check the req->rq_arrival time after the call to
2756 ost_brw_lock_get(), but before we do anything about
2757 processing it & sending the BULK transfer request. This
2758 should help move old stale pending locks off the queue as
2759 quickly as obd_timeout.
2762 Frequency : when an incorrect nid is specified during startup
2764 Description: ptlrpc connect to non-existant node causes kernel crash
2765 Details : LNET can't be re-entered from an event callback, which
2766 happened when we expire a message after the export has been
2767 cleaned up. Instead, hand the zombie cleanup off to another
2770 Severity : enhancement
2772 Description: plain/inodebits lock performance improvement
2773 Details : Grouping plain/inodebits in granted list by their request modes
2774 and bits policy, thus improving the performance of search through
2778 Frequency : only if OST filesystem is corrupted
2780 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2781 Details : for a short time RPCs with bulk IO are in the replay list,
2782 but replay of bulk IOs is unimplemented. If the OST filesystem
2783 is corrupted due to disk cache incoherency and then replay is
2784 started it is possible to trip an assertion. Avoid putting
2785 committed RPCs into the replay list at all to avoid this issue.
2788 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2790 Description: System hang on startup
2791 Details : This bug allowed the liblustre (e.g. catamount) client to
2792 return to the app before handling all startup RPCs. This
2793 could leave the node unresponsive to lustre network traffic
2794 and manifested as a server ptllnd timeout.
2796 Severity : enhancement
2798 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2799 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2800 boolean to control whether to print peer debug info when a
2801 client's RPC times out.
2804 Frequency : only for kernels with patches from Lustre below 1.4.3
2806 Description: Remove old rdonly API
2807 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2810 Frequency : only for devices with external journals
2812 Description: Set external device read-only also
2813 Details : During a commanded failover stop, we set the disk device
2814 read-only while the server shuts down. We now also set any
2815 external journal device read-only at the same time.
2818 Frequency : when upgrading from 1.4 while trying to change parameters
2820 Description: The wrong (new) MDC name was used when setting parameters for
2821 upgraded MDT's. Also allows changing of OSC (and MDC)
2822 parameters if --writeconf is specified at tunefs upgrade time.
2825 Frequency : when setting specific ost indicies
2827 Description: QOS code breaks on skipped indicies
2828 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2829 created with --index need not be sequential.
2831 Severity : enhancement
2833 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2834 Details : The uninit_groups feature works in conjunction with the kernel
2835 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2836 up the pass1 processing of e2fsck. This is a read-only feature
2837 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2838 mount filesystems that have had this feature enabled.
2840 Severity : enhancement
2842 Description: Improve multi-block allocation algorithm to avoid fragmentation
2843 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2844 allocation locality and avoid filesystem fragmentation.
2846 ------------------------------------------------------------------------------
2848 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2850 * Support for kernels:
2851 2.4.21-47.0.1.EL (RHEL 3)
2852 2.6.5-7.283 (SLES 9)
2853 2.6.9-42.0.10.EL (RHEL 4)
2854 2.6.12.6 vanilla (kernel.org)
2855 2.6.16.27-0.9 (SLES 10)
2856 * Recommended e2fsprogs version: 1.39.cfs5
2858 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2862 Frequency : occasional, depends on client load and configuration
2863 Bugzilla : 12181, 12203
2864 Description: data loss for recently-modified files
2866 Details : In some cases it is possible that recently written or created
2867 files may not be written to disk in a timely manner (this should
2868 normally be within 30s unless client IO load is very high).
2869 The problem appears as zero-length files or files that are a
2870 multiple of 1MB in size after a client crash or client eviction
2871 that are missing data at the end of the file.
2873 This problem is more likely to be hit on clients where files are
2874 repeatedly created and unlinked in the same directory, clients
2875 have a large amount of RAM, have many CPUs, the filesystem has
2876 many OSTs, the clients are rebooted frequently, and/or the files
2877 are not accessed by other nodes after being written.
2879 The presence of the problem can be detected by looking at
2880 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2881 smaller than the second (nr_unused) then dirty files will not
2882 be flushed automatically to disk. "sync; sleep 10" should be
2883 run several times on the node before unmounting it to update
2884 Lustre (this is also safe to run on nodes without this problem).
2886 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2887 that can cause this same problem, so customers using that kernel
2888 also need to update the kernel in addition to Lustre. In order
2889 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2891 It is normal that files written just before a client crash (less
2892 than 30s) may not yet have been flushed to disk, even for local
2896 Frequency : frequent on thin XT3 nodes
2898 Description: UUID collision on thin XT3 Linux nodes
2899 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2900 randomly, since we relied on an insufficiently-seeded PRNG.
2905 Description: OSS hangs after "All ost request buffers busy"
2906 Details : A deadlock between quota and journal operations caused OSS
2907 hangs after printing "All ost request buffers busy."
2910 Frequency : always on liblustre builds
2912 Description: Cleanup compiler warnings on liblustre
2915 Frequency : always on liblustre builds on XT3
2917 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2921 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2922 > 32000 subdirectories
2923 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2924 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2925 subdirectory count crosses 32000. This will aid e2fsck to
2926 correctly handle more than 32000 subdirectories.
2929 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2931 Description: System hang on startup
2932 Details : This bug allowed the liblustre (e.g. catamount) client to
2933 return to the app before handling all startup RPCs. This
2934 could leave the node unresponsive to lustre network traffic
2935 and manifested as a server ptllnd timeout.
2937 Severity : enhancement
2939 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2940 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2941 boolean to control whether to print peer debug info when a
2942 client's RPC times out.
2947 Description: make O_SYNC working on 2.6 kernels
2948 Details : 2.6 kernels use different method for mark pages for write,
2949 so need add a code to lustre for O_SYNC work.
2954 Description: Failure to close file and release space on NFS
2955 Details : Put inode details into lock acquired in ll_intent_file_open.
2956 Use mdc_intent_lock in ll_intent_open to properly
2957 detect all kind of errors unhandled by mdc_enqueue.
2962 Description: proc file read during shutdown sometimes raced obd removal,
2964 Details : Add lock to prevent obd access after proc file removal.
2967 Frequency : Only for files larger than 4GB on 32-bit clients.
2969 Description: improperly doing page alignment of locks
2970 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2971 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2976 Description: RPCs being resent when they shouldn't be
2977 Details : Some RPCs that should not be resent are being resent. This
2978 can cause inconsistencies in the RPC state machine. Do not
2979 resend such requests.
2982 Frequency : rare, only with NFS export
2984 Description: Crash on NFS re-export node
2985 Details : under very unusual load conditions an assertion is hit in
2986 ll_intent_file_open()
2989 Frequency : only if OST filesystem is corrupted
2991 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2992 Details : for a short time RPCs with bulk IO are in the replay list,
2993 but replay of bulk IOs is unimplemented. If the OST filesystem
2994 is corrupted due to disk cache incoherency and then replay is
2995 started it is possible to trip an assertion. Avoid putting
2996 committed RPCs into the replay list at all to avoid this issue.
3001 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3002 Details : Large single O_DIRECT read and write calls can fail to allocate
3003 a sufficiently large buffer to process the request. In case of
3004 allocation failure the allocation is retried with a smaller
3005 buffer and broken into smaller requests.
3007 Severity : enhancement
3009 Description: Add -o localflock option to simulate old noflock behaviour.
3010 Details : This will achieve local-only flock/fcntl locks coherentness.
3015 Description: versioning check is incomplete
3016 Details : Checking the version difference of client vs. server, report
3017 error if the gap is too big.
3022 Description: add support PG_writeback bit
3023 Details : add support for PG_writeback bit for Lustre, for more carefull
3024 work with page cache in 2.6 kernel. This also fix some deadlocks
3025 and remove hack for work O_SYNC with 2.6 kernel.
3027 Severity : enhancement
3029 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3030 Details : The uninit_groups feature works in conjunction with the kernel
3031 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3032 up the pass1 processing of e2fsck. This is a read-only feature
3033 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3034 mount filesystems that have had this feature enabled.
3036 Severity : enhancement
3038 Description: Improve multi-block allocation algorithm to avoid fragmentation
3039 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3040 allocation locality and avoid filesystem fragmentation.
3042 ------------------------------------------------------------------------------
3044 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3046 * Support for kernels:
3047 2.6.9-42.0.3.EL (RHEL 4)
3048 2.6.5-7.276 (SLES 9)
3049 2.4.21-47.0.1.EL (RHEL 3)
3050 2.6.12.6 vanilla (kernel.org)
3051 2.6.16.21-0.8 (SLES10)
3052 * Recommended e2fsprogs version: 1.39.cfs2-0
3054 * The backwards-compatible /proc/sys/portals symlink has been removed
3055 in this release. Before upgrading, please ensure that you change
3056 any configuration scripts or /etc/sysctl.conf files that access
3057 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3058 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3059 in advance of the upgrade on any system running Lustre 1.4.6 or
3060 newer, since /proc/sys/lnet was added in that version.
3061 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3065 Frequency : only when quota is used
3067 Description: avoid scanning export list for quota master
3068 Details : Change the algorithms to avoid scanning export list in order
3069 to improve the efficiency.
3072 Frequency : MDS failover only, very rarely
3074 Description: "went back in time" messages on mds failover
3075 Details : The greatest transno may be lost when the current operation
3076 finishes with an error (transno==0) and the client's last_rcvd
3077 record is over-written. Save the greatest transno in the
3078 mds_last_transno for this case.
3081 Frequency : always for specific kernels and striping counts
3083 Description: client may get "Matching packet too big" without ACL support
3084 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3085 when trying to access files in certain configurations. The
3086 clients should in fact be denied when mounting because they do
3087 not understand ACLs.
3090 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3092 Description: many clients connecting with IO in progress causes connect timeouts
3093 Details : Avoid synchronous journal commits to avoid delays caused by many
3094 clients connecting/disconnecting when bulk IO is in progress.
3095 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3096 OST_IO_PORTAL to avoid delays behind potentially many pending
3100 Frequency : occasionally with multiple writers to a single file
3102 Description: shared writes to file may result in wrong size reported by stat()
3103 Details : Allow growing of kms when extent lock is cancelled
3106 Frequency : always with random mmap IO to multi-striped file
3108 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3109 Details : Only if the hole is at the end of OST object so that kms is too
3110 small. Fix is to increase kms accordingly in ll_nopage.
3113 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3115 Description: writes to a missing object would leak memory on the OST
3116 Details : If there is an inconsistency between the MDS and OST filesystems,
3117 such that the MDS references an object that doesn't exist, writes
3118 to that object will leak memory due to incorrect cleanup in the
3119 error handling path, eventually running out of memory on the OST.
3124 Description: Creating too long symlink causes lustre errors
3125 Details : Check symlink and name lengths before sending requests to MDS.
3128 Frequency : only if flock is enabled (not on by default)
3130 Description: posix locks not released on fd closure on 2.6.9+
3131 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3132 kernels, this caused such locks not to be released on fd close and
3133 then assertions on fs unmount about still used locks.
3136 Frequency : MDS failover only, very rarely
3138 Description: clients may get ASSERTION(granted_lock != NULL)
3139 Details : When request was taking a long time, and a client was resending
3140 a getattr by name lock request. The were multiple lock requests
3141 with the same client lock handle and
3142 mds_getattr_name->fixup_handle_for_resent_request found one of the
3143 lock handles but later failed with ASSERTION(granted_lock != NULL).
3148 Description: handle->h_buffer_credits > 0, assertion failure
3149 Details : h_buffer_credits is zero after truncate, causing assertion
3150 failure. This patch extends the transaction or creates a new
3154 Frequency : NFS re-export or patchless client
3155 Bugzilla : 11179, 10796
3156 Description: Crash on NFS re-export node (__d_move)
3157 Details : We do not want to hash the dentry if we don't have a lock.
3158 But if this dentry is later used in d_move, we'd hit uninitialised
3159 list head d_hash, so we just do this to init d_hash field but
3160 leave dentry unhashed.
3163 Frequency : NFS re-export or patchless client
3165 Description: NFS exports has problem with symbolic link
3166 Details : lustre client didn't properly install dentry when re-exported
3167 to NFS or running patchless client.
3170 Frequency : NFS re-export or patchless client
3172 Description: Various nfs/patchless fixes.
3173 Details : fixes reuse disconected alias for lookup process - this fixes
3174 warning "find_exported_dentry: npd != pd",
3175 fix permission error with open files at nfs.
3176 fix apply umask when do revalidate.
3179 Frequency : occasional
3181 Description: Crash on NFS re-export node
3182 Details : calling clear_page() on the wrong pointer triggered oops in
3183 generic_mapping_read().
3186 Frequency : rarely, using O_DIRECT IO
3188 Description: unaligned directio crashes client with LASSERT
3189 Details : check for unaligned buffers before trying any requests.
3192 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3194 Description: stale data returned from RAID cache
3195 Details : If only a small amount of IO is done to the RAID device before
3196 reading it again it is possible to get stale data from the RAID
3197 cache instead of reading it from disk.
3200 Frequency : always for sles10 kernel
3202 Description: sles10 support
3203 Details : ll_follow_link: compile fixes and using of nd_set_link
3204 under newer kernels.
3207 Frequency : depends on arch, kernel and compiler version, always on sles10
3210 Description: recursive or deep enough symlinks cause stack overflow
3211 Details : getting rid of large stack-allocated variable in
3215 Frequency : depends on hardware
3217 Description: lustre write performance loss in the SLES10 kernel
3218 Details : the performance loss is caused by using of write barriers in the
3219 ext3 code. The SLES10 kernel turns barrier support on by
3220 default. The fix is to undo that change for ldiskfs.
3222 ------------------------------------------------------------------------------
3224 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3226 * Support for kernels:
3227 2.6.9-42.0.3EL (RHEL 4)
3228 2.6.5-7.276 (SLES 9)
3229 2.4.21-47.0.1.EL (RHEL 3)
3230 2.6.12.6 vanilla (kernel.org)
3234 Frequency : quota enabled and large files being deleted
3236 Description: releasing more than 4GB of quota at once hangs OST
3237 Details : If a user deletes more than 4GB of files on a single OST it
3238 will cause the OST to spin in an infinite loop. Release
3239 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3244 Description: statfs data retrieved from /proc may be stale or zero
3245 Details : When reading per-device statfs data from /proc, in the
3246 {kbytes,files}_{total,free,avail} files, it may appear
3247 as zero or be out of date.
3250 Frequency : systems with MD RAID1 external journal devices
3252 Description: lconf's call to blkid is confused by RAID1 journal devices
3253 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3254 of returning all block devices that match the journal UUID.
3257 Frequency : always, for aggregate stripe size over 4GB
3259 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3260 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3261 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3264 Frequency : always if "lfs find" used on a local file/directory
3266 Description: "lfs find" segfaults if used on a local file/directory
3267 Details : The case where a directory component was not specified wasn't
3268 handled correctly. Handle this properly.
3271 Frequency : always on ppc64
3273 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3274 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3275 perform endianness conversion.
3278 Frequency : rarely (truncate to non-zero file size after write under load)
3279 Bugzilla : 10730, 10687
3280 Description: Files padded with zeros to next 4K multiple
3281 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3282 it is possible that files that are truncated to a non-zero size
3283 immediately after being written are filled with zero bytes beyond
3284 the truncated size. No file data is lost.
3286 Severity : enhancement
3288 Description: Allow recovery/failover for liblustre clients.
3289 Details : liblustre clients were unaware of failover configurations until
3292 Severity : enhancement
3294 Description: user file locks should fail when not mounting with flock option
3295 Details : Set up an error-returning stub in ll_file_operations.lock field
3296 to prevent incorrect behaviour when client is mounted without
3297 flock option. Also, set up properly f_op->flock field for
3301 Frequency : always on ia64
3303 Description: "lfs df" loops on printing out MDS statfs information
3304 Details : The obd_ioctl_data was not initialized and in some systems
3305 this caused a failure during the ioctl that did not return
3306 an error. Initialize the struct and return an error on failure.
3309 Frequency : SLES 9 only
3311 Description: Error of copying files with lustre special EAs as root
3312 Details : Client side always return success for setxattr call for lustre
3313 special xattr (currently only "trusted.lov").
3316 Frequency : rarely on clusters with both ia64+i386 clients
3318 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3319 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3320 clients are doing concurrent writes to the same file it is
3321 possible that the ia64 clients may LASSERT because the OST
3322 extent locks are not PAGE_SIZE aligned. Ensure that grown
3323 locks are always aligned on the request boundary.
3326 Frequency : specific use, occasional
3328 Description: Overwriting in use executable truncates on-disk binary image
3329 Details : If one node attempts to overwrite an executable in use by
3330 another node, we now correctly return ETXTBSY instead of
3331 truncating the file.
3333 Severity : enhancement
3335 Description: Async OSC create to avoid the blocking unnecessarily.
3336 Details : If a OST has no remain object, system will block on the creating
3337 when need to create a new object on this OST. Now, ways use
3338 pre-created objects when available, instead of blocking on an
3339 empty osc while others are not empty. If we must block, we block
3340 for the shortest possible period of time.
3345 Description: chmod on Lustre root is propagated to other clients
3346 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3347 invalid by the follow_mount time.
3350 Frequency : liblustre clients only
3352 Description: Race in 'instant cancel' lock handling could lead to such locks
3353 never to be granted in case of SMP MDS
3354 Details : Do not destroy not yet granted but cbpending locks in
3358 Frequency : replay/resend of open
3360 Description: non null lock assetion failure in mds_intent_policy
3361 Details : Trying to replay/resend lockless open requests resulted in
3362 mds_open() returning 0 with no lock. Now it sets a flag if
3363 a lock is going to be returned.
3365 Severity : enhancement
3367 Description: Checksum enhancements
3368 Details : New checksum enhancements allow for resending RPCs that failed
3371 Severity : enhancement
3373 Description: Tunables on number of dirty pages in cacche
3374 Details : Allow to set limit on number of dirty pages cached.
3379 Description: client crash on unmount - lock still has references
3380 Details : In some error handling cases it was possible to leak a lock
3381 reference on a client while accessing a file. This was not
3382 harmful to the client during operation, but would cause the
3383 client to crash when the filesystem is unmounted.
3386 Frequency : specific case, rare
3388 Description: ETXTBSY on mds though file not in use by client
3389 Details : ETXTBSY is no longer incorrectly returned when attempting to
3390 chmod or chown a directory that the user previously tried to
3391 execute or a currently-executing binary.
3394 Frequency : extremely rare except on liblustre-based clients
3396 Description: Lustre space not freed when files are deleted
3397 Details : Clean up open-unlinked files after client eviction. Previously
3398 the unlink was skipped and the files remained as orphans.
3403 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3404 Details : In some cases it was possible to send a blocking callback to
3405 a client doing a glimpse, even though that client didn't get
3406 a lock granted. When the glimpse lock is cancelled on the OST
3407 the freed lock is left on the waiting list and corrupted the list.
3410 Frequency : all core dumps
3412 Description: Broke core dumps to lustre
3413 Details : Negative dentry may be unhashed if parent does not have UPDATE
3414 lock, but some callers, e.g. do_coredump, expect dentry to be
3415 hashed after successful create, hash it in ll_create_it.
3417 ------------------------------------------------------------------------------
3419 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3421 * Support for kernels:
3422 2.6.9-42.0.2.EL (RHEL 4)
3423 2.6.5-7.276 (SLES 9)
3424 2.4.21-40.EL (RHEL 3)
3425 2.6.12.6 vanilla (kernel.org)
3429 Frequency : always on RHEL 3
3431 Description: Number of open files grows over time
3432 Details : The number of open files grows over time, whether or not
3433 Lustre is started. This was due to a filp leak introduced
3434 by one of our kernel patches.
3436 ------------------------------------------------------------------------------
3438 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3440 * Support for kernels:
3441 2.6.9-42.EL (RHEL 4)
3442 2.6.5-7.276 (SLES 9)
3443 2.4.21-40.EL (RHEL 3)
3444 2.6.12.6 vanilla (kernel.org)
3449 Bugzilla : 5719, 9635, 9792, 9684
3450 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3451 Details : If a server is under heavy load and cannot reply to new
3452 connection requests before the client resends the (re)connect,
3453 the connection handling code can behave badly if two service
3454 threads are concurrently handing separate (re)connections from
3455 the same client. Add better locking to the connection handling
3456 code, and ensure that only a single connection will be processed
3457 for a given client UUID, even if the lock is dropped.
3459 Severity : enhancement
3461 Description: add TCP zero-copy support to kernel
3462 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3463 sends if the hardware supports scatter-gather and checksumming.
3464 This allows socklnd to do client-write and server-read more
3465 efficiently and reduce CPU utilization from skbuf copying.
3468 Frequency : only if NFS exporting from client
3470 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3471 Details : If setting attributes on a file created under NFS that had
3472 never been opened it would be possible to oops the client
3473 if the file had no objects.
3476 Frequency : always for liblustre
3478 Description: liblustre client does MDS+OSTs setattr RPC for each write
3479 Details : When doing a write from a liblustre client, the client
3480 incorrectly issued an RPC to the MDS and each OST the file was
3481 striped over in order to update the timestamps. When writing
3482 with small chunks and many clients this could overwhelm the MDS
3483 with RPCs. In all cases it would slow down the write because
3484 these RPCs are unnecessary.
3486 Severity : enhancement
3488 Description: allow number of MDS service threads to be changed at module load
3489 Details : It is now possible to change the number of MDS service threads
3490 running. Adding "options mds mds_num_threads={N}" to the MDS's
3491 /etc/modprobe.conf will set the number of threads for the next
3492 time Lustre is restarted (assuming the "mds" module is also
3493 reloaded at that time). The default number of threads will
3494 stay the same, 32 for most systems.
3499 Description: OST crash if filesystem is unformatted or corrupt
3500 Details : If an OST is started on a device that has never been formatted
3501 or if the filesystem is corrupt and cannot even mount then the
3502 error handling cleanup routines would dereference a NULL pointer.
3507 Description: NULL pointer deref in llap_from_page.
3508 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3509 incorrect) mapping if the page was truncated/reclaimed while it was
3510 searched for. Check for this condition and skip such pages when
3511 doing readahead. Introduce extra check to llap_from_page() to
3512 verify page->mapping->host is non-NULL (so page is not anonymous).
3515 Frequency : Sometimes when using sys_sendfile
3517 Description: "page not covered by a lock" warnings from ll_readpage
3518 Details : sendfile called ll_readpage without right page locks present.
3519 Now we introduced ll_file_sendfile that does necessary locking
3520 around call to generic_file_sendfile() much like we do in
3524 Frequency : with certain MDS communication failures at client mount time
3526 Description: NULL pointer deref after failed client mount
3527 Details : a client connection request may delayed by the network layer
3528 and not be sent until after the PTLRPC layer has timed out the
3529 request. If the client fails the mount immediately it will try
3530 to clean up before the network times out the request. Add a
3531 reference from the request import to the obd device and delay
3532 the cleanup until the network drops the request.
3535 Frequency : occasionally during client (re)connect
3537 Description: assertion failure during client (re)connect
3538 Details : processing a client connection request may be delayed by the
3539 client or server longer than the client connect timeout. This
3540 causes the client to resend the connection request. If the
3541 original connection request is replied in this interval, the
3542 client may trip an assertion failure in ptlrpc_connect_interpret()
3543 which thought it would be the only running connect process.
3546 Frequency : only with obd_echo servers and clients that are rebooted
3548 Description: kernel BUG accessing uninitialized data structure
3549 Details : When running an obd_echo server it did not start the ping_evictor
3550 thread, and when a client was evicted an uninitialized data
3551 structure was accessed. Start the ping_evictor in the RPC
3552 service startup instead of the OBD startup.
3554 Severity : enhancement
3555 Bugzilla : 10193 (patchless)
3556 Description: Remove dependency on various unexported kernel interfaces.
3557 Details : No longer need reparent_to_init, exit_mm, exit_files,
3558 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3561 Frequency : rare (only users of deprecated and unsupported LDAP config)
3563 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3564 Details : LDAP apparently contains 'lustreName' attributes instead of
3565 'name'. A simple remapping of the name is sufficient.
3568 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3570 Description: waiting_locks_callback trips kernel BUG if client is evicted
3571 Details : Running with the dump_on_timeout debug flag turned on makes
3572 it possible that the waiting_locks_callback() can try to dump
3573 the Lustre kernel debug logs from an interrupt handler. Defer
3574 this log dumping to the expired_lock_main() thread.
3576 Severity : enhancement
3578 Description: Support NFS exporting on 2.6 kernels.
3579 Details : Implement non-rawops metadata methods for NFS server to use without
3580 changing NFS server code.
3583 Frequency : very rare (synthetic metadata workload only)
3585 Description: two racing renames might cause an MDS thread to deadlock
3586 Details : Running the "racer" program may cause one MDS thread to rename
3587 a file from being the source of a rename to being the target of
3588 a rename at exactly the same time that another thread is doing
3589 so, and the second thread has already enqueued these locks after
3590 doing a lookup of the target and is trying to relock them in
3591 order. Ensure that we don't try to re-lock the same resource.
3594 Frequency : only very large systems with liblustre clients
3596 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3597 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3598 all OSTs in parallel. This allows the eviction of stale liblustre
3599 clients to proceed much faster than if they were done in series,
3600 and also offers similar improvements for other set_info RPCs.
3605 Description: excessive CPU usage during initial read phase on client
3606 Details : During the initial read phase on a client, it would agressively
3607 retry readahead on the file, consuming too much CPU and impacting
3608 performance (since 1.4.5.8). Improve the readahead algorithm
3609 to avoid this, and also improve some other common cases (read
3610 of small files in particular, where "small" is files smaller than
3611 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3616 Description: MDS crash when receiving packet with unknown intent.
3617 Details : Do not LBUG in unknown intent case, just return -EFAULT
3619 Severity : enhancement
3620 Bugzilla : 9293, 9385
3621 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3622 Details : Do not serialize getattr (non-intent version) and statfs.
3625 Frequency : occasional, when OST network is overloaded/intermittent
3627 Description: client evicted by OST after bulk IO timeout
3628 Details : If a client sends a bulk IO request (read or write) the OST
3629 may evict the client if it is unresposive to its data GET/PUT
3630 request. This is incorrect if the network is overloaded (takes
3631 too long to transfer the RPC data) or dropped the OST GET/PUT
3632 request. There is no need to evict the client at all, since
3633 the pinger and/or lock callbacks will handle this, and the
3634 client can restart the bulk request.
3637 Frequency : Always when mmapping file with no objects
3639 Description: client crashes when mmapping file with no objects
3640 Details : Check that we actually have objects in a file before doing any
3641 operations on objects in ll_vm_open, ll_vm_close and
3647 Description: Request leak when working with deleted CWD
3648 Details : Introduce advanced request refcount tracking for requests
3649 referenced from lustre intent.
3651 Severity : Enhancement
3653 Description: Cache open file handles on client.
3654 Details : MDS now will return special lock along with openhandle, if
3655 requested and client is allowed to hold openhandle, even if unused,
3656 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3657 closing files for every read/write openration.
3659 Severity : Enhancement
3661 Description: Cache open negative dentries on client when possible.
3662 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3663 negative dentries on lock revocation.
3668 Description: Remounting a client read-only wasn't possible with a zconf mount
3669 Details : It wasn't possible to remount a client read-only with llmount.
3671 Severity : enhancement
3672 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3673 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3674 MPICH is included to add Lustre-specific ADIO interfaces.
3675 This is based closely on the UFS ADIO layer and only differs
3676 in file creation, in order to allow the OST striping to be set.
3677 This is user-contributed code and not supported by CFS.
3682 Description: extended inode attributes (immutable, append-only) work improperly
3683 when 2.4 and 2.6 kernels are used on client/server or vice versa
3684 Details : Introduce kernel-independent values for these flags.
3686 Severity : enhancement
3689 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3690 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3691 granularity. Now, Fractional MB granularity is be supported,
3692 this is very useful for low memory system.
3694 Severity : enhancement
3696 Description: Getattr by fid
3697 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3698 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3699 have locks and avoids getattr rpc requests for it on every stat.
3702 Frequency : Always, for filesystems larger than 2TB
3704 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3705 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3706 filesystems larger than 2TB to be mounted with mballoc enabled.
3709 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3711 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3712 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3713 filesystems larger than 2TB is mounted then the kernel will
3714 silently corrupt the start of the filesystem. CONFIG_LBD is
3715 enabled for all CFS-supported kernels, but the possibility of
3716 this happening with a modified kernel config exists.
3718 Severity : enhancement
3720 Description: add client O_DIRECT support for 2.6 kernels
3721 Details : It is now possible to do O_DIRECT reads and writes to files
3722 in the Lustre client mountpoint on 2.6 kernel clients.
3724 Severity : enhancement
3726 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3727 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3728 parallel, not waiting for response from every OST before sending
3729 a rpc to the next OST.
3734 Description: setattr vs write race when updating file timestamps
3735 Details : Client processes that update a file timestamp into the past
3736 right after writing to the file (e.g. tar) it is possible that
3737 the updated file modification time can be reset to the current
3738 time due to a race between processing the setattr and write RPC.
3740 Severity : enhancement
3742 Description: Bring 'lfs find' closer in line with regular Linux find.
3743 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3744 -print0 options and obtains all the needed info through the lustre
3747 Severity : enhancement
3749 Description: support up to 1024 configured devices on one node
3750 Details : change obd_dev array from statically allocated to dynamically
3751 allocated structs as they are first used to reduce memory usage
3756 Description: Flush dirty partially truncated pages during truncate
3757 Details : Immediatelly flush partially truncated pages in filter_setattr,
3758 this way we completely avoid having any pages in page cache on OST
3759 and can retire ugly workarounds during writes to flush such pages.
3764 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3765 Details : protect inode from truncation within vfs_unlink() context
3766 just take a reference before calling vfs_unlink() and release it
3767 when parent's i_sem is free.
3772 Description: last_id value checked outside lock on OST caused LASSERT failure
3773 Details : If there were multiple MDS->OST object precreate requests in
3774 flight, it was possible that the OST's last object id was checked
3775 outside a lock and incorrectly tripped an assertion. Move checks
3776 inside locks, and discard old precreate requests.
3779 Frequency : always, if extents are used on OSTs
3781 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3782 Details : OSTs using the extents format would not zero the high 16 bits of
3783 the index physical block number. This is not a problem for any
3784 OST filesystems smaller than 16TB, and no kernels support ext3
3785 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3786 new/modified files) and can be fixed for existing filesystems
3787 with e2fsprogs-1.39-cfs1.
3792 Description: import connection selection may be incorrect if timer wraps
3793 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3794 import connections to be ignored if the 32-bit jiffies counter
3795 wraps. Use a 64-bit jiffies counter.
3798 Frequency : very large clusters immediately after boot
3800 Description: LNET request buffers exhausted under heavy short-term load
3801 Details : If a large number of client requests are generated on a service
3802 that has previously never seen so many requests it is possible
3803 that the request buffer growth cannot keep up with the spike in
3804 demand. Instead of dropping incoming requests, they are held in
3805 the LND until the RPC service can accept more requests.
3808 Frequency : Sometimes during replay
3810 Description: Assertion failure in ll_local_open after replay.
3811 Details : If replay happened on an open request reply before we were able
3812 to set replay handler, reply will become not swabbed tripping the
3813 assertion in ll_local_open. Now we set the handler right after
3814 recognising of open request
3817 Frequency : very rare
3819 Description: kernel reports "badness in vsnprintf"
3820 Details : Reading from the "recovery_status" /proc file in small chunks
3821 may cause a negative length in lprocfs_obd_rd_recovery_status()
3822 call to vsnprintf() (which is otherwise harmless). Exit early
3823 if there is no more space in the output buffer.
3825 Severity : enhancement
3827 Description: clear OBD RPC statistics by writing to them
3828 Details : It is now possible to clear the OBD RPC statistics by writing
3829 to the "stats" file.
3834 Description: Client mtime is not the same on different clients after utimes
3835 Details : In some cases, the client was using the utimes() syscall on
3836 a file cached on another node. The clients now validate the
3837 ctime from the MDS + OSTs to determine which one is right.
3842 Description: Inability to activate failout mode
3843 Details : lconf script incorrectly assumed that in pythong string's numeric
3844 value is used in comparisons.
3847 Frequency : always with multiple stripes per file
3849 Description: Inefficient object allocation for mutli-stripe files
3850 Details : When selecting which OSTs to stripe files over, for files with
3851 a stripe count that divides evenly into the number of OSTs,
3852 the MDS is always picking the same starting OST for each file.
3853 Return the OST selection heuristic to the original design.
3858 Description: mount failures may take full timeout to return an error
3859 Details : Under some heavy load conditions it is possible that a
3860 failed mount can wait for the full obd_timeout interval,
3861 possibly several minutes, before reporting an error.
3862 Instead return an error as soon as the status is known.
3864 Frequency : quota enabled and large files being deleted
3866 Description: releasing more than 4GB of quota at once hangs OST
3867 Details : If a user deletes more than 4GB of files on a single OST it
3868 will cause the OST to spin in an infinite loop. Release
3869 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3874 Description: statfs data retrieved from /proc may be stale or zero
3875 Details : When reading per-device statfs data from /proc, in the
3876 {kbytes,files}_{total,free,avail} files, it may appear
3877 as zero or be out of date.
3880 Frequency : systems with MD RAID1 external journal devices
3882 Description: lconf's call to blkid is confused by RAID1 journal devices
3883 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3884 of returning all block devices that match the journal UUID.
3887 Frequency : always, for aggregate stripe size over 4GB
3889 Description: assertion fails when trying to use 4GB stripe size
3890 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3891 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3894 Frequency : always on ppc64
3896 Description: the first write on an ext3 filesystem with mballoc got stuck
3897 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3898 perform endianness conversion.
3900 ------------------------------------------------------------------------------
3902 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3904 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3905 INTEROPERATE with older versions automatically. Please read the
3906 user documentation before upgrading any part of a live system.
3907 * WARNING: Lustre networking configuration changes are required with
3908 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3911 * Support for kernels:
3912 2.6.9-22.0.2.EL (RHEL 4)
3913 2.6.5-7.244 (SLES 9)
3914 2.6.12.6 vanilla (kernel.org)
3917 Severity : enhancement
3918 Bugzilla : 7981/8208
3919 Description: Introduced Lustre Networking (LNET)
3920 Details : LNET is new networking infrastructure for Lustre, it includes
3921 a reorganized network configuration mode (see the user
3922 documentation for full details) as well as support for routing
3923 between different network fabrics. Lustre Networking Devices
3924 (LNDs) for the supported network fabrics have also been
3925 created for this new infrastructure.
3927 Severity : enhancement
3928 Description: Introduced Access control lists
3929 Details : clients can set ACLs on files and directories in order to have
3930 more fine-grained permissions than the standard Unix UGO+RWX.
3931 The MDS must be started with the "-o acl" mount option.
3933 Severity : enhancement
3934 Description: Introduced filesystem quotas
3935 Details : Administrators may now establish per-user quotas on the
3938 Severity : enhancement
3940 Description: Configuration change for the XT3
3941 The PTLLND is now used to run Lustre over Portals on the XT3
3942 The configure option(s) --with-cray-portals are no longer used.
3943 Rather --with-portals=<path-to-portals-includes> is used to
3944 enable building on the XT3. In addition to enable XT3 specific
3945 features the option --enable-cray-xt3 must be used.
3950 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3951 Details : A race between allocating a new llcd and re-getting the llcd_lock
3952 allowed another thread to grab newly-allocated llcd.
3954 Severity : enhancement
3956 Description: 2.6 OST async journal commit and locking fix to improve performance
3957 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3958 2.6 kernels are now async as they already were in 2.4 kernels so
3959 that they can commit concurrently with the network bulk transfer.
3960 For block-allocated files the filter allocation semaphore is held
3961 to avoid filesystem fragmentation during allocation. BKL lock
3962 removed for 2.6 xattr operations where it is no longer needed.
3967 Description: lconf incorrectly determined whether two IP networks could talk
3968 Details : In some more complicated routing and multiple-network
3969 configurations, lconf will avoid trying to make a network
3970 connection to a disjoint part of the IP space. It was doing the
3971 math incorrectly for one set of cases.
3976 Description: Fix for potential infinite loop processing records in an llog.
3977 Details : If an llog record is corrupted/zeroed, it is possible to loop
3978 forever in llog_process(). Validate the llog record length
3979 and skip the remainder of the block on error.
3982 Frequency : occasional (liblustre only)
3984 Description: liblustre could not open files whose last component is a symlink
3985 Details : sysio_path_walk() would incorrectly pass the open intent to
3986 intermediate path components.
3989 Frequency : rare (liblustre only with non-standard tuning)
3990 Bugzilla : 7201 (7350)
3991 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3992 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3993 after it has been released, only do this for referenced locks
3995 Severity : enhancement
3997 Description: specifying an (invalid) directory default stripe_size of -1
3998 would reset the directory default striping
3999 Details : stripe_size -1 was used internally to signal directory stripe
4000 removal, now use "all default" to signal dir stripe removal
4001 as a directory striping of "all default" is not useful
4004 Frequency : common for large clusters running liblustre clients
4006 Description: doing an ls when liblustre clients are running is slow
4007 Details : sending a glimpse AST to a liblustre client waits for every AST
4008 to time out, as liblustre clients will not respond. Since they
4009 cannot cache data we refresh the OST lock LVB from disk instead.
4011 Severity : enhancement
4013 Description: doing an ls at the same time as file IO can be slow
4014 Details : enqueue and other "small" requests can be blocked behind many
4015 large IO requests. Create a new OST IO portal for non-IO
4016 requests so they can be processed faster.
4019 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4021 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4022 created file with mode 000
4023 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4024 allow an ftruncate() as a non-root user to a file with mode 000.
4025 Lustre now respects this flag to disable mode checks when
4026 truncating a file owned by the user
4029 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4032 Description: Revoking locks from clients that went dead or catatonic might take
4034 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4035 cancellation of such locks instant on servers without waiting for
4036 any reply from clients. Clients drops these locks when cancel
4037 notification from server is received without replying.
4040 Frequency : liblustre-only, when liblustre client dies or becomes busy
4042 Description: Doing ls on Linux clients can take a long time with active
4044 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4045 granting such locks to it in the first place if possible. Locks
4046 are taken by proxy on the OST during the read or write and
4047 dropped immediately afterward. Add connect flags handling, do
4048 not grant locks to liblustre clients for glimpse ASTs.
4050 Severity : enhancement
4052 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4053 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4054 sequential reads and behaves badly with applications doing random
4055 reads. Improve it by reading ahead at least read region, and
4056 avoiding excessive large RPC for small reads.
4058 Severity : enhancement
4060 Description: Creating more than 1000 files for a single job may cause a load
4061 imbalance on the OSTs if there are also a large number of OSTs.
4062 Details : qos_prep_create() uses an OST index reseed value that is an
4063 even multiple of the number of available OSTs so that if the
4064 reseed happens in the middle of the object allocation it will
4065 still utilize the OSTs as uniformly as possible.
4070 Description: OST or MDS may oops in ping_evictor_main()
4071 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4072 but doesn't restart at beginning of obd_exports_timed list
4075 Severity : enhancement
4077 Description: improve by-nid export eviction on the MDS and OST
4078 Details : allow multiple exports with the same NID to be evicted at one
4079 time without re-searching the exports list.
4082 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4084 Description: MDS may oops in groups_free()
4085 Details : in rare race conditions a newly allocated group_info struct is
4086 freed again, and this can be NULL. The 2.4 compatibility code
4087 for groups_free() checked for a NULL pointer, but 2.6 did not.
4090 Frequency : common for liblustre clients doing little filesystem IO
4091 Bugzilla : 9352, 7313
4092 Description: server may evict liblustre clients accessing contended locks
4093 Details : if a client is granted a lock or receives a completion AST
4094 with a blocking AST already set it would not reply to the AST
4095 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4098 Frequency : lfs setstripe, only systems with more than 160 OSTs
4100 Description: unable to set striping with a starting offset beyond OST 160
4101 Details : llapi_create_file() incorrectly limited the starting stripe
4102 index to the maximum single-file stripe count.
4105 Frequency : LDAP users only
4107 Description: lconf did not handle in-kernel recovery with LDAP properly
4108 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4110 Severity : enhancement
4112 Description: bind OST threads to NUMA nodes to improve performance
4113 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4114 node and do their allocations there to localize memory access
4116 Severity : enhancement
4118 Description: llmount can determine client NID directly from Myrinet (GM)
4119 Details : the client NID code from gmnalnid was moved directly into
4120 llmount, removing the need to use this or specifying the
4121 client NID explicitly when mounting GM clients with zeroconf
4124 Frequency : if client is started with down MDS
4126 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4127 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4128 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4129 ptlrpc_import_delayed_req() because we didn't check if the
4130 request was interrupted, and we also didn't break out of the
4131 event loop if there was no timeout
4136 Description: data loss during non-page-aligned writes to a single file from
4137 both multiple nodes and multiple threads on one node at same time
4138 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4139 inconsistency led to false short-reads, that were cached and later
4140 used by ->prepare_write() to fill in partially written page,
4141 leading to data loss.
4144 Frequency : always, if lconf --abort_recovery used
4146 Description: lconf --abort_recovery fails with 'Operation not supported'
4147 Details : lconf was attempting to abort recovery on the MDT device and not
4150 Severity : enhancement
4152 Description: remove cleanup logs
4153 Details : replace lconf-generated cleanup logs with lustre internal
4154 cleanup routines. Eliminates the need for client-cleanup and
4157 Severity : enhancement
4159 Description: add support for EAs (user and system) on lustre filesystems
4160 Details : it is now possible to store extended attributes in the Lustre
4161 client filesystem, and with the user_xattr mount option it
4162 is possible to allow users to store EAs on their files also
4164 Severity : enhancement
4166 Description: Add possibility (config option) to show minimal available OST free
4168 Details : When compiled with --enable-mindf configure option, statfs(2)
4169 (and so, df) will return least minimal free space available from
4170 all OSTs as amount of free space on FS, instead of summary of
4171 free spaces of all OSTs.
4173 Severity : enhancement
4175 Description: do not expand extent locks acquired on OST-side
4176 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4177 by server: they are not cached anyway.
4180 Frequency : when mmap is used/binaries executed from Lustre
4182 Description: Unmmap pages before throwing them away from read cache.
4183 Details : llap_shrink cache now attempts to unmap pages before discarding
4184 them (if unmapping failed - do not discard). SLES9 kernel has
4185 extra checks that trigger if this unmapping is not done first.
4190 Description: lconf didn't resolve symlinks before checking to see whether a
4191 given mountpoint was already in use
4194 Frequency : when migrating failover services
4195 Bugzilla : 6395, 9514
4196 Description: When migrating a subset of services from a node (e.g. failback
4197 from a failover service node) the remaining services would
4198 time out and evict clients.
4199 Details : lconf --force (implied by --failover) sets the global obd_timeout
4200 to 5 seconds in order to quickly disconnect, but this caused
4201 other RPCs to time out too quickly. Do not change the global
4202 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4204 Severity : enhancement
4205 Frequency : if MDS is started with down OST
4206 Bugzilla : 9439,5706
4207 Description: Allow startup/shutdown of an MDS without depending on the
4208 availability of the OSTs.
4209 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4210 Add appropriate locking and lov-osc refcounts for safe
4211 cleaning. Add osc abort_inflight calls in case the
4212 synchronize never started.
4215 Frequency : occasional (Cray XT3 only)
4217 Description: root not authorized to access files in CRAY_PORTALS environment
4218 Details : The client process capabilities were not honoured on the MDS in
4219 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4220 been accessed by an authorized user then root was able to access
4221 the file on the local client also. The root user capabilities
4222 are now allowed on the MDS, as this environment has secure UID.
4225 Frequency : occasional
4227 Description: ldiskfs "too long searching" message happens too often
4228 Details : A debugging message (otherwise harmless) prints too often on
4229 the OST console. This has been reduced to only happen when
4230 there are fragmentation problems on the filesystem.
4235 Description: Division by zero in statfs when all OSCs are inactive
4236 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4237 lov_statfs divides by value returned by lov_get_stripecnt().
4241 Bugzilla : 9489, 3273
4242 Description: First write from each client to each OST was only 4kB in size,
4243 to initialize client writeback cache, which caused sub-optimal
4244 RPCs and poor layout on disk for the first writen file.
4245 Details : Clients now request an initial cache grant at (re)connect time
4246 and so that they can start streaming writes to the cache right
4247 away and always do full-sized RPCs if there is enough data.
4248 If the OST is rebooted the client also re-establishes its grant
4249 so that client cached writes will be honoured under the grant.
4254 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4255 Details : Now I/O RPCs go to different portal number and (presumably) fast
4256 lock requests (and glimses) and other RPCs get their own service
4257 threads pool that should be able to service those RPCs
4260 Severity : enhancement
4262 Description: Ability to exchange lustre version between client and servers and
4263 issue warnings at client side if client is too old. Also for
4264 liblustre clients there is ability to refuse connection of too old
4266 Details : New 'version' field is added to connect data structure that is
4267 filled with version info. That info is later checked by server and
4271 Frequency : rare, liblustre only.
4272 Bugzilla : 9296, 9581
4273 Description: Two simultaneous writes from liblustre at offset within same page
4274 might proceed at the same time overwriting eachother with stale
4276 Details : I/O lock withing llu_file_prwv was released too early, before data
4277 actually was hitting the wire. Extended lock-holding time until
4278 server acknowledges receiving data.
4281 Frequency : extremely rare. Never observed in practice.
4283 Description: avoid generating lustre_handle cookie of 0.
4284 Details : class_handle_hash() generates handle cookies by incrementing
4285 global counter, and can hit 0 occasionaly (this is unlikely, but
4286 not impossible, because initial value of cookie counter is
4287 selected randonly). Value of 0 is used as a sentinel meaning
4288 "unassigned handle" --- avoid it. Also coalesce two critical
4289 sections in this function into one.
4291 Severity : enhancement
4293 Description: allow liblustre clients to delegate truncate locking to OST
4294 Details : To avoid overhead of locking, liblustre client instructs OST to
4295 take extent lock in ost_punch() on client's behalf. New connection
4296 flag is added to handle backward compatibility.
4298 Severity : enhancement
4299 Bugzilla : 4928, 7341, 9758
4300 Description: allow number of OST service threads to be specified
4301 Details : a module parameter allows the number of OST service threads
4302 to be specified via "options ost ost_num_threads={N}" in the
4303 OSS's /etc/modules.conf or /etc/modprobe.conf.
4307 Bugzilla : 6146, 9635, 9895
4308 Description: servers crash with bad pointer in target_handle_connect()
4309 Details : In rare cases when a client is reconnecting it was possible that
4310 the connection request was the last reference for that export.
4311 We would temporarily drop the export reference and get a new
4312 one, but this may have been the last reference and the export
4313 was just destroyed. Get new reference before dropping old one.
4315 Severity : enhancement
4316 Frequency : if client is started with failover MDS
4318 Description: Allow multiple MDS hostnames in the mount command
4319 Details : Try to read the configuration from all specified MDS
4320 hostnames during a client mount in case the "primary"
4323 Severity : enhancement
4325 Description: Stop sending data to evicted clients as soon as possible.
4326 Details : Check if the client we are about to send or are sending data to
4327 was evicted already. (Check is done every second of waiting,
4328 for which l_wait_event interface was extended to allow checking
4329 of exit condition at specified intervals).
4332 Frequency : rare, normally only when NFS exporting is done from client
4334 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4336 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4339 Frequency : always, if srand() is called before liblustre initialization
4341 Description: Liblustre uses system PRNG disturbing its usage by user application
4342 Details : Introduce internal to lustre fast and high-quality PRNG for
4343 lustre usage and make liblustre and some other places in generic
4344 lustre code to use it.
4346 Severity : enhancement
4347 Bugzilla : 9477, 9557, 9870
4348 Description: Verify that the MDS configuration logs are updated when xml is
4349 Details : Check if the .xml configuration logs are newer than the config
4350 logs stored on the MDS and report an error if this is the case.
4351 Request --write-conf, or allow starting with --old_conf.
4353 Severity : enhancement
4355 Description: Handle symlinks in the path when checking if Lustre is mounted.
4356 Details : Resolve intermediate symlinks when checking if a client has
4357 mounted a filesystem to avoid duplicate client mounts.
4362 Description: lconf can hit an error exception but still return success.
4363 Details : The lconf command catches the Command error exception at the top
4364 level script context and will exit with the associated exit
4365 status, but doesn't ensure that this exit status is non-zero.
4370 Description: failure of ptlrpc thread startup can cause oops
4371 Details : Starting a ptlrpc service thread can fail if there are a large
4372 number of threads or the server memory is very fragmented.
4373 Handle this without oopsing.
4376 Frequency : always, only if liblustre and non-default acceptor port was used
4378 Description: liblustre cannot connect to servers with non-default acceptor port
4379 Details : tcpnal_set_default_params() was not called and was therefore
4380 ignoring the environment varaible TCPNAL_PORT, as well as other
4381 TCPNAL_ environment variables
4386 Description: two objects could be created on the same OST for a single file
4387 Details : If an OST is down, in some cases it was possible to create two
4388 objects on a single OST for a single file. No problems other
4389 than potential performance impact and spurious error messages.
4393 Bugzilla : 5681, 9562
4394 Description: Client may oops in ll_unhash_aliases
4395 Details : Client dcache may become inconsistent in race condition.
4396 In some cases "getcwd" can fail if the current directory is
4402 Description: Inode refcounting problems in NFS export code
4403 Details : link_raw functions used to call d_instantiate without obtaining
4404 extra inode reference first.
4408 Bugzilla : 9942, 9903
4409 Description: Referencing freed requests leading to crash, memleaks with NFS.
4410 Details : We used to require that call to ll_revalidate_it was always
4411 followed by ll_lookup_it. Also with revalidate_special() it is
4412 possible to call ll_revalidate_it() twice for the same dentry
4413 even if first occurence returned success. This fix changes semantic
4414 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4415 reference on a request referred from the intent.
4416 ll_intent_release() then releases such a request.
4419 Frequency : rare, normally benchmark loads only
4421 Description: unlinked inodes were kept in memory on the client
4422 Details : If a client is repeatedly creating and unlinking files it
4423 can accumulate a lot of stale inodes in the inode slab cache.
4424 If there is no other client load running this can cause the
4425 client node to run out of memory. Instead flush old inodes
4426 from client cache that have the same inode number as a new inode.
4429 Frequency : SLES9 2.6.5 kernel and long filenames only
4430 Bugzilla : 9969, 10379
4431 Description: utime reports stale NFS file handle
4432 Details : SLES9 uses out-of-dentry names in some cases, which confused
4433 the lustre dentry revalidation. Change it to always use the
4437 Frequency : rare, unless heavy write-truncate concurrency is continuous
4438 Bugzilla : 4180, 6984, 7171, 9963, 9331
4439 Description: OST becomes very slow and/or deadlocked during object unlink
4440 Details : filter_destroy() was holding onto the parent directory lock
4441 while truncating+unlinking objects. For very large objects this
4442 may block other threads for a long time and slow overall OST
4443 responsiveness. It may also be possible to get a lock ordering
4444 deadlock in this case, or run out of journal credits because of
4445 the combined truncate+unlink. Solution is to do object truncate
4446 first in one transaction without parent lock, and then do the
4447 final unlink in a new transaction with the parent lock. This
4448 reduces the lock hold time dramatically.
4451 Frequency : rare, 2.4 kernels only
4453 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4454 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4455 before unmounting their filesystem in order to avoid deadlock.
4456 The kernel_locked() function in 2.4 kernels only checks whether
4457 the kernel lock is held, not whether it is this process that is
4458 holding it as 2.6 kernels do.
4463 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4464 Details : The client ptlrpc code may be trying to reconnect to a down
4465 server before a previous connection attempt has timed out.
4466 Increase the reconnect interval to be longer than the connection
4467 timeout interval to avoid sending duplicate connections to
4471 Frequency : echo_client brw_test command
4473 Description: fix echo_client to work with OST preallocated code
4474 Details : OST preallocation code (5137) didn't take echo_client IO path
4475 into account: echo_client calls filter methods outside of any
4476 OST thread and, hence, there is no per-thread preallocated
4477 pages and buffers to use. Solution: hijack pga pages for IO. As
4478 a byproduct, this avoids unnecessary data copying.
4482 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4483 Description: Client can oops in mdc_commit_close() after open replay
4484 Details : It was possible for the MDS to return an open request with no
4485 transaction number in mds_finish_transno() if the client was
4486 evicted, but without actually returning an error. Clients
4487 would later try to replay that open and may trip an assertion
4488 Simplify the client close codepath, and always return an error
4489 from the MDS in case the open is not successful.
4492 Frequency : rare, 2.6 OSTs only
4494 Description: OST may deadlock under high load on fragmented files
4495 Details : If there was a heavy load and highly-fragmented OST filesystems
4496 it was possible to have all the OST threads deadlock waiting on
4497 allocation of biovecs, because the biovecs were not released
4498 until the entire RPC IO was completed. Instead, release biovecs
4499 as soon as they are complete to ensure forward IO progress.
4501 Severity : enhancement
4503 Description: Support for specifying external journal device at mount
4504 Details : If an OST or MDS device is formatted with an external journal
4505 device, this device major/minor is stored in the ext3 superblock
4506 and may not be valid for failover. Allow detecting and
4507 specifying the external journal at mount time.
4512 Description: Mounting an MDS with pending unlinked files may cause oops
4513 Details : target_finish_recovery() calls mds_postrecov() which returned
4514 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4515 considers this an error and immediately begins cleaning up the
4516 lov, just after starting the mds_lov process
4518 Severity : enhancement
4520 Description: Implement 'lfs df' to report actual free space on per-OST basis
4521 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4522 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4523 usage of objects; '-h' to report in human readable format.
4525 ------------------------------------------------------------------------------
4527 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4534 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4535 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4536 at startup during orphan unlinking. Instead, skip buddy update
4537 before setup, as it will be regenerated after recovery is complete.
4542 Description: If an OST is inactive, its locks might reference stale inodes.
4543 Details : lov_change_cbdata() must iterate over all namespaces, even if
4544 they are inactive to clear inode references from the lock.
4546 Severity : enhancement
4547 Frequency : occasional, if non-standard max_dirty_mb used
4549 Description: Client will block write RPCs if not enough grant
4550 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4551 then the client will block writes while waiting for another RPC
4552 to complete instead of consuming its dirty limit. With change
4553 we get improved performance when max_dirty_mb is small.
4555 Severity : enhancement
4556 Bugzilla : 3389, 6253
4557 Description: Add support for supplementary groups on the MDS.
4558 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4559 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4560 lookups for user supplementary groups into a cache.
4564 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4565 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4566 for liblustre O_EXCL is enforced by MDS.
4570 Description: readdir never returns NULL in liblustre.
4571 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4572 dentry in struct dirent.
4576 Frequency : liblustre only, depends on application IO pattern
4577 Description: liblustre clients evicted if not contacting servers
4578 Details : Don't put liblustre clients into the ping_evictor list, so
4579 they will not be evicted by the pinger ever.
4581 Severity : enhancement
4583 Description: Add ability to evict clients by NID from MDS.
4584 Details : By echoing "nid:$NID" string into
4585 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4586 $NID would be instantly evicted from this MDS and from all active
4587 OSTs connected to it.
4591 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4592 Details : lookup_it_finish() used to query file size from OSTs that was not
4597 Description: service threads change working directory to that of init
4598 Details : Starting lustre service threads may pin the working directory
4599 of the parent thread, making that filesystem busy. Threads
4600 now change to the working directory of init to avoid this.
4604 Frequency : during shutdown only
4605 Description: shutdown with a failed MDS or OST can cause unmount to hang
4606 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4611 Frequency : 2.6 only
4612 Description: chmod/chown may include an extra supplementary group
4613 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4614 supplementary group and if none is specified this is used.
4617 Bugzilla : 5479 (6816)
4619 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4620 Details : If lookup is in progress on a file that is unlinked we might try
4621 to revalidate the inode and fail in revalidate after lookup is
4622 complete and ll_file_open() enqueues the open again but
4623 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4626 Frequency : always, if lconf --abort_recovery used
4628 Description: lconf --abort_recovery fails with 'Operation not supported'
4629 Details : lconf was attempting to abort recovery on the MDT device and not
4632 ------------------------------------------------------------------------------
4634 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4639 Frequency : rare (only unsupported configurations with a node running as an
4641 Bugzilla : 6514, 5137
4642 Description: Mounting a Lustre file system on a node running as an OST could
4644 Details : OSTs now preallocates memory needed to write out data at
4645 startup, instead of when needed, to avoid having to
4646 allocate memory in possibly low memory situations.
4647 Specifically, if the file system is mounted on on OST,
4648 memory pressure could force it to try to write out data,
4649 which it needed to allocate memory to do. Due to the low
4650 memory, it would be unable to do so and the node would
4651 become unresponsive.
4653 Severity : enhancement
4655 Description: Addition of lconf --service command line option
4656 Details : lconf now accepts a '--service <arg>' option, which is
4657 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4659 Severity : enhancement
4661 Description: Failover mode is now the default for OSTs.
4662 Details : By default, OSTs will now run in failover mode. To return to
4663 the old behaviour, add '--failout' to the lmc line for OSTs.
4665 Severity : enhancement
4667 Description: Health checks are now provided for MDS and OSTs
4668 Details : Additional detailed health check information on MSD and OSTs
4669 is now provided through the procfs health_check value.
4672 Frequency : occasional, depends on IO load
4674 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4675 after numerous create/delete cycles
4676 Details : The ext3 inode allocation policy would not allocate new inodes
4677 very well on the OSTs because there are no new directories
4678 being created. Instead we look for groups with free space if
4679 the parent directories are nearly full.
4684 Description: Network or server problems during mount may cause partially
4685 mounted clients instead of returning an error.
4686 Details : The config llog parsing code may overwrite the error return
4687 code during mount error handling, returning success instead
4693 Description: MDS can fail to allocate large reply buffers
4694 Details : After long uptimes the MDS can fail to allocate large reply
4695 buffers (e.g. zconf client mount config records) due to memory
4696 fragmentation or consumption by the buffer cache. Preallocate
4697 some large reply buffers so that these replies can be sent even
4698 under memory pressure.
4702 Frequency : rare (liblustre)
4703 Description: fsx running with liblustre complained that using truncate() to
4704 extend the file doesn't work. This patch corrects that issue.
4705 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4706 ATTR_SIZE and lsm use in llu_setattr_raw.
4710 Frequency : rare, only 2.6 kernels
4711 Description: Unusual file access patterns on the MDS may result in inode
4712 data being lost in very rare circumstances.
4713 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4714 code in the 2.6 kernel caused the inode and/or EA data not to
4715 be read from disk, causing single-file corruption.
4719 Frequency : rare, only 2.6 filesystems using extents
4720 Description: Heavy concurrent write and delete load may cause data corruption.
4721 Details : It was possible under high-load situations to have an extent
4722 metadata block in the block device cache from a just-unlinked
4723 file overwrite a newly-allocated data block. We now unmap any
4724 metadata buffers that alias just-allocated data blocks.
4728 Frequency : filesystems with default stripe_count larger than 77
4729 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4730 Details : lconf specifies an inode size of 4096 bytes when the default
4731 stripe_count is larger than 77. This conflicts with the default
4732 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4733 this case to avoid pinning too much memory for large EAs.
4735 ------------------------------------------------------------------------------
4737 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4742 Frequency : rare (extremely heavy IO load with hundreds of clients)
4744 Description: Client is evicted, gets IO error writing to file
4745 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4746 caused clients to be evicted for AST timeouts. The fixes in
4747 bug 5192 mean we no longer need to have such short AST timeouts
4748 so ldlm_timeout has been increased.
4751 Frequency : occasional during --force or --failover shutdown under load
4752 Bugzilla : 5949, 4834
4753 Description: Server oops/LBUG if stopped with --force or --failover under load
4754 Details : a collection of import/export refcount and cleanup ordering
4755 issues fixed for safer force cleanup
4758 Frequency : only filesystems larger than 120 OSTs
4759 Bugzilla : 5990, 6223
4760 Description: lfs getstripe would oops on a very large filesystem
4761 Details : lov_getconfig used kfree on vmalloc'd memory
4764 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4765 Bugzilla : 6242, 6243
4766 Description: reading from files that had been truncated to a non-zero size
4767 but never opened returned no data
4768 Details : ll_file_read() reads zeros from no-object files to EOF
4773 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4774 Details : There's little harm in aborting MDS/OSS recovery and letting it
4775 try again, so I removed the LASSERT and return an error instead.
4777 Severity : enhancement
4779 Description: New debugging infrastructure for tracking down data corruption
4780 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4781 (b) cover more of the client-side code path, and (c) try to narrow
4782 down where problems occurred
4786 Bugzilla : 3819, 4364, 4397, 6313
4787 Description: Racing close and eviction MDS could cause assertion in mds_close
4788 Details : It was possible to get multiple mfd references during close and
4789 client eviction, leading to one thread referencing a freed mfd.
4791 Severity: : enhancement
4792 Bugzilla : 3262, 6359
4793 Description: Attempts to reconnect to servers are now more aggressive.
4794 Details : This builds on the enhanced upcall-less recovery that was added
4795 in 1.4.2. When trying to reconnect to servers, clients will
4796 now try each server in the failover group every 10 seconds. By
4797 default, clients would previously try one server every 25 seconds.
4802 Description: After recovery, certain operations trigger a failed
4803 assertion on a client.
4804 Details : Failing over an mds, using lconf -d --failover, while a
4805 client was doing a readdir() call would cause the client to
4806 LBUG after recovery completed and the readdir() was resent.
4808 Severity : enhancement
4810 Description: Default groups are now added by lconf
4811 Details : You can now run lconf --group <servicename> without having to
4812 manually add groups with lmc.
4815 Frequency : occasional
4817 Description: Nodes with an elan id of 0 trigger a failed assertion
4820 Frequency : always when accessing e.g. tty/console device nodes
4822 Description: tty and some other devices nodes cannot be used on lustre
4823 Details : file's private_data field is used by device data and lustre
4824 values in there got lost. New field was added to struct file to
4825 store fs-specific private data.
4828 Frequency : when exporting Lustre via NFS
4830 Description: NFSD failed occasionally when looking up a path component
4831 Details : NFSD is looking up ".." which was broken in ext3 directories
4832 that had grown large enough to become hashed.
4835 Frequency : Clusters with multiple interfaces not on the same subnet
4837 Description: Nodes will repeatedly try to reconnect to an interface which it
4838 cannot reach and report an error to the log.
4839 Details : Extra peer list entries will be created by lconf with some peers
4840 unreachable. lconf now validates the peer before adding it.
4843 Frequency : Only if a default stripe is set on the filesystem root.
4845 Description: Setting a default stripe on the filesystem root prevented the
4846 filesystem from being remounted.
4847 Details : The client was sending extra request flags in the root getattr
4848 request and did not allocate a reply buffer for the dir EA.
4851 Frequency : occasional, higher if lots of files are accessed by one client
4852 Bugzilla : 6159, 6097
4853 Description: Client trips assertion regarding lsm mismatch/magic
4854 Details : While revalidating inodes the VFS looks up inodes with ifind()
4855 and in rare cases can find an inode that is being freed.
4856 The ll_test_inode() code will free the lsm during ifind()
4857 when it finds an existing inode and then the VFS later attaches
4858 this free lsm to a new inode.
4862 Bugzilla : 6422, 7030
4863 Description: MDS deadlock between mkdir and client eviction
4864 Details : Creating a new file via mkdir or mknod (starting a transaction
4865 and getting the ns lock) can deadlock with client eviction
4866 (gets ns lock and trying to finish a synchronous transaction).
4869 Frequency : occasional
4870 Description: While starting a server, the fsfilt_ext3 module could not be
4872 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4873 kernels. Previously, lconf would still use the ext3 name
4874 when trying to load modules. Now, it will correctly use
4875 ext3 on 2.4 and ldiskfs on 2.6.
4877 Severity : enhancement
4878 Description: The default stripe count has been changed to 1
4879 Details : The interpretation of the default stripe count (0, to lfs
4880 or lmc) has been changed to mean striping across a single
4881 OST, rather than all available. For general usage we have
4882 found a stripe count of 1 or 2 works best.
4884 Severity : enhancement
4885 Description: Add support for compiling against Cray portals.
4886 Details : Conditional compiling for some areas that are different
4890 Frequency : occasional
4891 Bugzilla : 6409, 6834
4892 Description: Creating files with an explicit stripe count may lead to
4893 a failed assertion on the MDS
4894 Details : If some OSTs are full or unavailable, creating files may
4895 trigger a failed assertion on the MDS. Now, Lustre will
4896 try to use other servers or return an error to the
4900 Frequency : occasional
4902 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4903 caused lock timeout during readahead (since 1.4.2).
4904 Details : Processes doing readahead might match a lock that hasn't been
4905 granted yet if there are overlapping and conflicting lock
4906 requests. The readahead process waits on ungranted lock
4907 (original lock is CBPENDING), while OST waits for that process
4908 to cancel CBPENDING read lock and eventually evicts client.
4910 Severity : enhancement
4912 Description: Initial enabling of flock support for clients
4913 Details : Implements fcntl advisory locking and file status functions.
4914 This feature is provided as an optional mount flag (default
4915 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4916 locking are implemented yet, and those that are are not guaranteed
4917 to be completely correct in production environments.
4918 mount -t lustre -o [flock|noflock] ...
4921 Frequency : occasional
4923 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4924 assertion in the ext3 JBD (journaling) layer.
4925 Details : The b_committed_data struct is protected by the big kernel lock
4926 in 2.4 kernels, serializing journal_commit_transaction() and
4927 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4928 access to this struct. In 2.6 kernels there is finer grained
4929 locking to improve SMP performance of the JBD layer.
4933 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4938 Description: merge and cleanup kernel patches.
4941 -----------------------------------------------------------------------------
4943 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4945 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4946 versions of Lustre. Please update all servers and clients to
4947 version 1.4.2 or later at the same time. You must also run
4948 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4949 to update the configuration logs.
4951 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4952 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4953 - swab llog records of type '0' so we get proper header size/idx (5861)
4954 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4955 - fix rename of one directory over another leaking an inode (5953)
4956 - avoid SetPageDirty on 2.6 (5981)
4957 - don't re-add just-being-destroyed locks to the waiting list (5653)
4958 - when creating new directories, inherit the parent's custom
4959 striping settings if present parent (3048)
4960 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4961 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4962 - don't hold client locks on temporary worklist from l_lru (5666)
4963 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4964 - automatically evict dead clients (5921)
4965 - Update file size properly in create+truncate+fstat case (6196)
4966 - Do not unhash mountpoint dentries, do not allow removal of
4968 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4969 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4970 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4971 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4972 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4973 - Correctly handle reads of files with no objects (6243)
4974 - lctl recover will also mark a device active if deactivate used (5933)
4976 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4977 - allow --write-conf on an MDS with different nettype than client (5619)
4978 - don't write config llogs to MDS for mounts not from that MDS (5617)
4979 - lconf should create multiple TCP connections from a client (5201)
4980 - init scripts are now turned off by default; run chkconfig --on
4981 lustre and chkconfig --on lustrefs to use them
4982 - upcalls are no longer needed for clients to recover to failover
4984 - add --abort-recovery option to lconf to abort recovery on device
4986 - add support for an arbitrary number of OSTs (3026)
4987 - Quota support protocol changes.
4988 - forward compatibility changes to wire structs (6007)
4989 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4990 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4991 - improved "lustre status" script
4992 - initialize blocksize for non-regular files (6062)
4993 - added --disable-server and --disable-client configure options (5782)
4994 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4995 - Vanilla 2.4.29 support
4996 - increase maximum number of obd devices to 520 (6242)
4997 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4998 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4999 SLES 9 2.6.5 kernels
5000 - limit stripes per file to 160 (the maximum EA size) (6093)
5002 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5005 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5006 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5007 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5008 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5009 - small ext3 extents cleanups and fixes (5733)
5010 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5011 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5012 - avoid needless client->OST connect, fix handle mismatch (5317)
5013 - fix DLM error path that led to out-of-sync client, long delays (5779)
5014 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5015 - fix several locking issues related to i_size (5492,5624,5654,5672)
5016 - don't move pending lock onto export if it is already evicted (5683)
5017 - fix kernel oops when creating .foo in unlinked directory (5548)
5018 - fix deadlock in obdfilter statistics vs. object create (5811)
5019 - use time_{before,after} to avoid timer jiffies wrap (5882)
5020 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5021 - Do not leak request if server was not able to process it (5154)
5022 - If mds_open unable to find parent dir, make that negative lookup(5154)
5023 - don't create new directories with extent-mapping (5909, 5936)
5025 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5026 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5027 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5028 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5029 - support for RHEL4 kernel on i686 (5773)
5030 - provide error messages when incompatible logs are encountered (5898)
5032 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5033 * version 1.4.0.10 (1.4.1 release candidate 1)
5035 - don't keep a lock reference when lock is not granted (4238)
5036 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5037 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5038 - drop import inflight refcount on signal_completed_replay error (5255)
5039 - unlock page after async write error during send (3677)
5040 - handle missing objects in filter_preprw_read properly (5265)
5041 - no transno return for symlink open, don't save no-trasno open (3440)
5042 - don't try to complete elan receive that already failed (4012)
5043 - free RPC server reply state on error (5406)
5044 - clean up thread from ptlrpc_start_thread() on error (5160)
5045 - readahead could read extra page into cache that wasn't ejected (5388)
5046 - prevent races in class_attach/setup/cleanup/detach (5260)
5047 - don't dereference de->d_inode after l_dput of de (5458)
5048 - use "int" for stripe value returned from lock_to_stripe (5544)
5049 - mballoc allocation and error-checking fixes in 2.6 (5504)
5050 - block device patches to fix I/O request sizes in 2.6 (5482)
5051 - look up hostnames for IB nals (5602)
5052 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5053 - don't start multiple acceptors for the same port (5277)
5054 - fix incorrect LASSERT in mds_getattr_name (5635)
5055 - export a proc file for general "ping" checking (5628)
5056 - fix "lfs check" to not block when the MDS is down (5628)
5058 - service request history (4965)
5059 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5060 - create an "evict_client" /proc entry on OSTs, like the MDS has
5061 - fix mount usage message, return errors per mount(8) (5168)
5062 - change grep [] to grep "[]" in tests so they work in more UMLs
5063 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5064 - remove ancient LOV_MAGIC_V0 EA support (5047)
5065 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5066 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5067 - allow readahead allocations to fail when low on memory (5383)
5068 - mmap locking landed again, after considerable improvement (2828)
5069 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5071 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5074 - send OST transaction number in read/write reply to free req (4966)
5075 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5076 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5077 - the watchdog thread now runs as interruptible (5246)
5078 - flock/lockf fixes (but it's still disabled, pending 5135)
5079 - don't use EXT3 constants in llite code (5094)
5080 - memory shortage at startup could cause assertion (5176)
5082 - reorganization of lov code
5083 - single portals codebase
5085 - add extents/mballoc support (5025)
5086 - direct I/O reads in the obdfilter (4048)
5087 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5089 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5092 - send OST transaction number in read/write reply to free req (4966)
5093 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5094 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5095 - the watchdog thread now runs as interruptible (5246)
5096 - handle missing objects in filter_preprw_read properly (5265)
5097 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5098 - drop import inflight refcount on signal_completed_replay error (5255)
5099 - unlock page after async write error during send (3677)
5100 - return original error code on reconstructed replies (3761)
5101 - no transno return for symlink open, don't save no-trasno open (3440)
5103 - add pid to ldlm debugging output (4922)
5104 - bump the watchdog timeouts -- we can't handle 30sec yet
5105 - extra debugging for orphan dentry/inode bug (5259)
5107 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5110 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5111 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5112 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5113 - fix NULL dereference in /proc/sys/portals/routes (4827)
5114 - allow failed mdc_close() operations to be interrupted (4561)
5115 - stop precreate on OST before MDS would time out on it (4778)
5116 - don't send partial-page writes before EOF from client (4410)
5117 - discard client grant for sub-page writes on large-page clients (4520)
5118 - don't free dentries not owned by NFS code, check generation (4806)
5119 - fix lsm leak if mds_create_objects() fails (4801)
5120 - limit debug_daemon file size, always print CERROR messages (4789)
5121 - use transno after validating reply (3892)
5122 - process timed out requests if import state changes (3754)
5123 - update mtime on OST during writes, return in glimpse (4829)
5124 - add mkfsoptions to LDAP (4679)
5125 - use ->max_readahead method instead of zapping global ra (5039)
5126 - don't interrupt __l_wait_event() during strace
5128 - add software watchdogs to catch hung threads quickly (4941)
5129 - make lustrefs init script start after nfs is mounted
5130 - fix CWARN/ERROR duplication (4930)
5131 - return async write errors to application if possible (2248)
5132 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5133 - print NAL number in %x format (4645)
5134 - update barely-supported suse-2.4.21-171 series (4842)
5135 - support for sles 9 %post scripts
5136 - support for building 2.6 kernel-source packages
5137 - support for sles km_* packages
5139 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5142 - ignore -ENOENT errors in osc_destroy (3639)
5143 - notify osc create thread that OSC is being cleaned up (4600)
5144 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5145 - reconstruct ost_handle() like mds_handle() (4657)
5146 - create a new thread to do import eviction to avoid deadlock (3969)
5147 - let lconf resolve symlinked-to devices (4629)
5148 - don't unlink "objects" from directory with default EA (4554)
5149 - hold socknal file ref over connect in case target is down (4394)
5150 - allow more than 32000 subdirectories in a single directory (3244)
5151 - fix blocks count for O_DIRECT writes (3751)
5152 - OST returns ENOSPC from object create when no space left (4539)
5153 - don't send truncate RPC if file size isn't changing (4410)
5154 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5155 - bind to privileged port in socknal and tcpnal (3689)
5157 - rate limit CERROR/CWARN console message to avoid overload (4519)
5158 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5159 - basic mmap support (3918)
5160 - kernel patch series update from b1_4 (4711)
5162 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5165 - avoid crash during MDS cleanup with OST shut down (2775)
5166 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5167 - don't use bad inodes on the MDS (3744)
5168 - dynamic object preallocation to improve recovery speed (4236)
5169 - don't hold spinlock over lock dumping or change debug flags (4401)
5170 - don't zero obd_dev when it is force cleaned (3651)
5171 - print grants to console if they go negative (4431)
5172 - "lctl deactivate" will stop automatic recovery attempts (3406)
5173 - look for existing locks in ldlm_handle_enqueue() (3764)
5174 - don't resolve lock handle twice in recovery avoiding race (4401)
5175 - revalidate should check working dir is a directory (4134)
5177 - don't always mark "slow" obdfilter messages as errors (4418)
5179 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5182 - don't close LustreDB during write_conf until it is done (3860)
5183 - fix typo in lconf for_each_profile (3821)
5184 - allow dumping logs from multiple threads at one time (3820)
5185 - don't allow multiple threads in OSC recovery (3812)
5186 - fix debug_size parameters (3864)
5187 - fix mds_postrecov to initialize import for llog ctxt (3121)
5188 - replace config semaphore with spinlock (3306)
5189 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5190 - don't allow enqueue to complete on a destroyed export (3822)
5191 - down write_lock before checking llog header bitmap (3825)
5192 - recover from lock replay timeout (3764)
5193 - up llog sem before sending rpc (3652)
5194 - reduce ns lock hold times when setting kms (3267)
5195 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5196 - fix NULL deref and obd_dev leak on setup error (3312)
5197 - replace some LBUG about llog ops with error handling (3841)
5198 - don't match INVALID dentries from d_lookup and spin (3784)
5199 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5200 - fix invalid assertion in ptlrpc_set_wait (3880)
5202 - add libwrap support for the TCP acceptor (3996)
5203 - add /proc/sys/portals/routes for non-root route listing (3994)
5204 - allow setting MDS UUID in .xml (2580)
5205 - print the stack of a process that LBUGs (4228)
5207 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5210 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5211 - make sure to unset replay flag from failed open requests (3440)
5212 - if default stripe count is 0, use OST count for inode size (3636)
5213 - update parent mtime/ctime on client for create/unlink (2611)
5214 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5215 - free recovery state on server during a forced cleanup (3571)
5216 - unregister_reply for resent reqs (3063)
5217 - loop back devices mounting and status check on 2.6 (3563)
5218 - fix resource-creation race that can provoke i_size == 0 (3513)
5219 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5220 - more debugging for page-accounting assertion (3746)
5221 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5222 - avoid deadlock after precreation failure (3758)
5223 - fix race and lock order deadlock in orphan handling (3450, 3750)
5224 - add validity checks when grabbing inodes from l_ast_data (3599)
5226 - add /proc/.../recovery_status to obdfilter (3428)
5227 - lightweight CDEBUG infrastructure, debug daemon (3668)
5228 - change default OSC RPC parameters to be better on small clusters
5229 - turn off OST read cache for files smaller than 32MB
5230 - install man pages and include them in rpms (3100)
5231 - add new init script for (un)mounting lustre filesystems (2593)
5232 - run chkconfig in %post for init scripts (3701)
5233 - drop scimac NAL (unmaintained)
5235 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5238 - clean kiobufs before and after use (3485)
5239 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5240 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5241 - change init script to fail more clearly if not run as root (1528)
5242 - allow clients to reconnect during replay (1742)
5243 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5244 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5245 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5246 - fix cond_resched (3554)
5247 - only evict unfinished clients after recovery (3515)
5248 - allow bulk resend, prevent data loss (3570)
5249 - dynamic ptlrpc request buffer allocation (2102)
5250 - don't allow unlinking open directory if it isn't empty (2904)
5251 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5252 - remove extraneous obd dereference causing LASSERT failure (3334)
5253 - don't use get_cycles() when creating temp. files on the mds (3156)
5254 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5255 - handle EEXIST for set-stripe, set proper directory name (3336)
5257 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5258 - fix ksocknal_fmb_callback() error messages (2918)
5260 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5263 - don't copy lvb into (possibly NULL) reply on error (2983)
5264 - don't deref dentry after dput, don't free lvb on error (2922)
5265 - use the kms to determine writeback rpc length (2947)
5266 - increment oti_logcookies when osc is inactive (2948)
5267 - update client's i_blocks count via lvb messages (2543)
5268 - handle intent open/close of special files properly (1557)
5269 - mount MDS with errors=remount-ro, like obdfilter (2009)
5270 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5271 - don't use cancelling-locks' kms values (2947)
5272 - use highest lock extent for kms, not last one (2925)
5273 - don't dereference ERR_PTR() dentry in error handling path (3107)
5274 - fix thread race in portals_debug_dumplog() (3122)
5275 - create lprocfs device entries at setup instead of at attach (1519)
5276 - common AST error handler, don't evict client on completion race (3145)
5277 - zero nameidata in detach_mnt in 2.6 (3118)
5278 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5279 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5280 - initialize RPC timeout timer earlier for 2.6 (3219)
5281 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5282 - print nal/nid for unknown nid (3258)
5283 - additional checks for oscc recovery before doing precreate (3284)
5284 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5285 - don't crash in mdc_close for bad permissions on open (3285)
5286 - zero i_rdev for non-device files (3147)
5287 - clear page->private before handing to FS, better assertion (3119)
5288 - tune the read pipeline (3236)
5289 - fix incorrect decref of invalidated dentry (2350)
5290 - provide read-ahead stats and refine rpc in flight stats (3328)
5291 - don't hold journal transaction open across create RPC (3313)
5292 - update atime on MDS at close time (3265)
5293 - close LDAP connection when recovering to avoid server load (3315)
5294 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5295 - don't leak open file on MDS after open resend (3325)
5296 - serialize filter_precreate and filter_destroy_precreated (3329)
5297 - loop device shouldn't call sync_dev() for nul device (3092)
5298 - clear page cache after eviction (2766)
5299 - resynchronize MDS->OST in background (2824)
5300 - refuse to mount the same filesystem twice on same mountpoint (3394)
5301 - allow llmount to create routes for mounting behind routers (3320)
5302 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5303 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5304 - fix rare problem with rename on htree directories (3417)
5306 - allow default OST striping configuration per directory (1414)
5307 - fix compilation for qswnal for 2.6 kernels (3125)
5308 - increase maximum number of MDS request buffers for large systems
5309 - change liblustreapi to be useful for external progs like lfsck (3098)
5310 - increase local configuration timeout for slow disks (3353)
5311 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5313 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5316 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5317 - don't overwrite extent policy data in reply if lock was blocked (2901)
5318 - drop filter export grants atomically with removal from device (2663)
5319 - del obd_self_export from work_list in class_disconnect_exports (2908)
5320 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5321 - swab reply message in mdc_close, other PPC fixes (2464)
5322 - fix destroying of named logs (2325)
5323 - overwrite old logs when running lconf --write_conf (2264)
5324 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5325 - fix race in target_handle_connect (2898)
5326 - mds_reint_create() should take same inode create lock (2926)
5327 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5328 - don't close files for self_export to avoid uninitialized obd (2936)
5329 - allow MDS with the same name as client node (2939)
5330 - hold dentry reference for closed log files for unlink (2325)
5331 - reserve space for all logs during transactions (2059)
5332 - don't evict page beyond end of stripe extent (2925)
5333 - don't oops on a deleted current working directory (2399)
5334 - handle hard links to targets without a parent properly (2517)
5335 - don't dereference NULL lock when racing during eviction (2867)
5336 - don't grow lock extents when lots of conflicting locks (2919)
5338 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5341 - account for cache space usage on clients to avoid data loss (974)
5342 - lfsck support in lustre kernel code (2349)
5343 - reduce journal credits needed for BRW writes (2370)
5344 - orphan handling to avoid losing space on client/server crashes
5345 - ptlrpcd can be blocked, stopping ALL progress (2477)
5346 - use lock value blocks to assist in proper KMS, faster stat (1021)
5347 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5348 - recovery for initial connections (2355)
5349 - fixes for mds_cleanup_orphans (1934)
5350 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5351 - block all file creations until orphan recovery completes (1901)
5352 - client remove rq_connection from request struct (2423)
5353 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5354 - recovery timer race (2670)
5355 - mdc_close recovey bug (2532)
5356 - ptlrpc cleanup bug (2710)
5357 - mds timeout on local locks (2588)
5358 - namespace lock held during RPCs (2431)
5359 - handle interrupted sync write properly (2503)
5360 - don't try to handle a message that hasn't been replied to (2699)
5361 - client assert failure during cleanup after abort recovery (2701)
5362 - leak mdc device after failed mount (2712)
5363 - ptlrpc_check_set allows timedout requests to complete (2714)
5364 - wait for inflight reqs when ptlrpcd finishes (2710)
5365 - make sure unregistered services are removed from the srv_list
5366 - reset bulk XID's when resending them (caught by 1138 test)
5367 - unregister_bulk after timeout
5368 - fix lconf error (2694)
5369 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5370 - readahead locks pages, leaves pending causing memory pressure (2673)
5371 - increase OST request buffers to 4096 on large machines (2729)
5372 - fix up permission of existing directories in simple_mkdir (2661)
5373 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5374 - don't assign transno to errored transactions (2742)
5375 - don't delete objects on OST if given a bogus objid from MDS (2751)
5376 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5377 - if rq_no_resend, then timeout request after recovery (2432)
5378 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5379 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5380 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5381 - MDS recovery shouldn't delete objects that it has given out (2730)
5382 - if enqueue arrives after completion, don't clobber LVB (2819)
5383 - don't unlock pages twice when trigger_group_io returns error (2814)
5384 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5385 - don't write pages to disk if there was an error (1450)
5386 - don't ping imports that have recovery disabled (2676)
5387 - take buffered bytes into account when balancing socknal conn (2817)
5388 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5389 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5390 - remove little-endian swabbing of llog records (1987)
5391 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5392 - retry reposting request buffers if they fail (1191)
5393 - grow extent at grant time to avoid granting a revoked lock (2809)
5394 - lock revoke doesn't evict page if covered by a second lock (2765)
5395 - disable VM readahead to avoid reading outside lock extents (2805)
5397 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5398 - updated kernel patches for hp-2.4.20 kernel (2681)
5400 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5403 - fix truncated write corruption (2366)
5404 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5406 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5407 - ensure len(uuid) < 37 in lmc (1171)
5408 - fix ia64 OOPS in llog_test (2255)
5409 - zero end of page at obdfilter for partial page writes (2648)
5410 - don't leave stale dentries around after renames (bug 2428)
5411 - fix timeouts when evicting a client with a single lock held (2642)
5412 - set deadline for the initial HELLO message to drain (2634)
5413 - print out dotted-quad IP addresses in the socknal (2302)
5415 - additional debugging for MDS client eviction problem (2443)
5416 - fix mkfsoptions support for osts (2603, 2604)
5418 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5421 - add series for the vanilla 2.6.0 kernel
5422 - add series for the vanilla 2.4.24 kernel
5423 - add series for a cray x86/64 UL kernel drop
5424 - fix xattr patches for the vanilla 2.4.19 series
5426 - generate true UUIDs in lmc (1171)
5427 - have portals stack dumping break in UML (2466)
5428 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5429 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5430 - wake callback waiting threads on client eviction (2460)
5431 - Add --ptldebug and --subsystem to lmc (1719)
5432 - update assertion to allow safe interrupt allocation
5433 - set rq_no_resend for cancel requests (2432)
5434 - recalculate ptlrpcd timeout after resend (2494)
5435 - call vfs_rmdir when removing pending directories (2368)
5436 - fix renaming a file to itself (2429)
5437 - lmc creates a default one-stripe lov (2454)
5438 - expand procfs space to handle large clusters (2326)
5439 - increase UML stack to avoid overflow
5440 - update lconf's list of debug and subsystem masks
5441 - fix lfs find --obd (2510)
5442 - /proc tunable for disabling filter read caching (2591)
5443 - stop rpm packages from altering slapd.conf (2301)
5444 - disable nagle in the socknal under 0conf (2578)
5445 - choose mds inode size based on stripe count (2572)
5446 - fix kernel-source rpm problems (2516)
5448 - add --disable-doc to avoid pdf generation (2421)
5449 - update documentation, tests, type-os, comments
5450 - avoid format warnings on ia64
5451 - remove the TOE NAL
5452 - tiny code cleanups by removing unused fields
5454 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5457 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5458 - remove the most blatant lies from BUILDING file (2371)
5459 - change default debug level to reasonable production setting
5460 - reduce client side cache size to reduce cache flush time
5461 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5462 - make TCP zerocopy and pinger support enabled by default (2476)
5463 - sync writes completed after process exits caused crashes (2319)
5464 - maintain correct mount count on the MDS (2356)
5465 - backout 1557, because 2316 wasn't really fixed
5466 - better file I/O statistics gathering in /proc
5467 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5468 - another kernel patch to fix zero-copy TCP function export
5469 - don't take duplicate lock when processing re-sent getattr (2420)
5470 - lctl uses obd_self_export instead of creating new conn (2353)
5471 - MDS/OST recovery case which requires object creation asserted (2425)
5472 - move lfs from /usr/sbin to /usr/bin in packages
5473 - fix race between mds_client_add and mds_client_free (2417)
5474 - use kmalloc instead of slabs in portals (2430)
5475 - don't create duplicate records when a failover MDS is present (2442)
5476 - remove unnecessary mount age check (2332)
5477 - don't remove directory inodes from locks prematurely (2451)
5478 - don't break if MDS service name is the same as hostname (2103)
5479 - fix races in client write RPC generation when cache full (2482)
5481 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5484 - remove now-unused request->rq_obd (278)
5485 - if an allocation fails, print out how much memory we've used (1933)
5486 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5487 - add the "configurable stack size" patch to most series files (1256)
5488 - ability to write large log records, for 100+ OST configs (2306)
5489 - fix NULL deref when filter_prep fails (2314)
5490 - fix operator precedence error in filter_sync
5491 - dynamic allocation of socknal TX descriptors (2315)
5492 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5493 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5494 - ensure that CWARN messages in Portals always get to the syslog
5495 - __init/__exit are not for prototype decls (ldlm_init/exit)
5496 - x86-64 compile warning fixes
5497 - fix gateway LMC keyword conflict (2318)
5498 - fix MDS lock inversions in getattr/reint paths (1844)
5499 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5500 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5501 - DLM race condition prevented some lock evictions (2328)
5502 - ENOMEM detection and retry on socknal sends (2230)
5503 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5504 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5505 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5506 - better error messages when a client is rejected during recovery (1505)
5507 - avoid cancelling locks which were never granted, after failure (2330)
5508 - fix i_sem/journal inversion in mds_client_add (2333)
5509 - fix truncate/getattr lock cycle deadlock (2334)
5510 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5511 - fix two rare exit paths which could leak an l_lock() ref (2321)
5512 - fencepost error in MDS/OST orphan recovery (2226)
5513 - make log record alignment 8 bytes (1988)
5514 - lstripe now fails when requested offset > ost_count (2237)
5515 - ensure that all kernel series have a complete list.h (1607)
5516 - fix crashes in special-file operations (2316)
5517 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5519 - allow configurable automake binary, for testing new versions
5520 - small update to the lfs documentation
5522 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5524 * fix negative export reference count in fsfilt_sync (2312)
5526 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5527 * release candidate 0.9.1
5529 - orphans are moved into the PENDING directory for possible recovery
5530 - replayed opens now open by fid for orphan/rename safety (1042)
5531 - last close of an orphan inode generates a transno (683)
5532 - chdir() and mount() now pin the directory entry (1020)
5533 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5534 - discard very old requests without processing them (1502)
5535 - remove some common, well-understood CERRORs (1505)
5536 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5537 - clear "grant" flags in OST replies until OST grant code lands (1644)
5538 - fix read performance by not clobbering i_blksize on client (1598)
5539 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5540 - make LRU size a /proc tunable, clears locks when reduced (707)
5541 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5542 - server locks take references on exports now (1558)
5543 - build fixes for 2.4.20-rh trees (1663)
5544 - return an error from lov_create if all OSCs are inactive (1751)
5545 - fix import levels when a reconnect happens without a timeout (1597)
5546 - exit early from mds_open if we get a lookup error (1749)
5547 - partial page read at EOF wouldn't wait for disk before sending (1642)
5548 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5549 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5550 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5551 - write/truncate lock inversion (1639)
5552 - Don't auto-load obdclass, portals modules during cleanup (1495)
5553 - fix timestamps from jumping to "now" (1763)
5554 - extra journal assertions (1648)
5555 - add an extra multiunlink test (1771)
5556 - fix read_record/write_record API (1776)
5557 - fix leak of offset_extent, possible incorrect i_size later (1772)
5558 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5559 - Debugging for the kqswnal_get_idle_tx problems (1820)
5560 - Allow recovery to be attempted multiple times (1536)
5561 - Write out MDS last_rcvd file after it is first created (1600)
5562 - Fix tx_descriptor leak in failed transmit situations (1827)
5563 - ext3 journaling fixes for assertion failure after IO error (1871)
5564 - class_export_put() on freed export after completion AST error (1896)
5565 - Fix revalidate looping in VFS (1322)
5566 - Don't access a freed export during MDS_REINT timeout (1521)
5567 - Add open-unlink recovery support on the MDS (1673,1764)
5568 - Return an error if no MDS data was read from last_rcvd (1946)
5569 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5570 - Don't setup a disk device that doesn't match exported UUID (317)
5571 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5572 - avoid committing NULL handle in force close
5573 - local.sh is now a one-stripe LOV configuration
5574 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5575 - fix ext3 htree duplicate directory entry corruption (1516)
5576 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5577 - update to new LOV EA format (2097)
5578 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5580 2003-06-15 Phil Schwan <phil@clusterfs.com>
5583 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5584 - per-import/export recovery handling (958, 931, 959)
5585 - multiple last-rcvd slots, for serving multiple FSes (949)
5586 - connections are again shared between multiple imp/exports (963, 964)
5587 - "umount -f" would hang if any requests needed to be sent (393, 978)
5588 - avoid pinning large req buffer by copying for queued messages (989)
5589 - add "uuid" to "lctl device" command to help upcalls (991)
5590 - "open" RPCs with transnos would confuse recovery counters (1037)
5591 - do proper endian conversion of all wire messages (288, 340, 891)
5592 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5593 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5594 - fix ptlrpc_connection leak in target_handle_connect (1174)
5595 - fix import refcounting bug in OST and MDS cleanup (1134)
5596 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5597 - fix very unlikely obd_types race condition (501)
5598 - remove osc_open hack for echo_client (1187)
5599 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5600 - a failure in read_inode2 leads to deadlock (1139)
5601 - cancel ack-locks as soon as transaction is committed (1072)
5602 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5603 - make sure to commitrw after any preprw to avoid deadlock (1162)
5604 - failing to execute a file in a lustre FS would lock inode (1203)
5605 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5606 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5607 - obd_uuid2tgt would walk off the end of the list (1255)
5608 - on IA64 the osc would give portals incorrect bulk size (1258)
5609 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5610 - fix lock inversion caused by new llite matching code (1282)
5611 - limit the number of dirty pages on a client to 10MB (1286)
5612 - timed out locks were not being corrected cancelled (1289)
5613 - fix O_DIRECT above 4GB on IA-32 (1292)
5614 * major user-visible changes
5615 - fail out/fail over policy now controlled by the upcall (993)
5617 - add OBD_PING to check server availability and failure (954)
5618 - lustre messages are now sent in sending host order (288, 340, 891)
5619 - add eadatalen to MDS getattr reply (340)
5620 - OST read replies may contain second buffer, with per-page status (593)
5622 2003-03-11 Phil Schwan <phil@clusterfs.com>
5625 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5626 - failed open()s could cause deadlock; fixed (867, 869)
5627 - stop cancelling OST locks when files are closed (481)
5628 - overlapping XID spaces caused network corruption (851, 853)
5629 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5630 - setattr_raw added, to do single-RPC, server-side setattrs
5631 - lmc/lconf syntax change for OST UUIDs
5632 - fix crashy race condition between ptlrpc_free_req and osc_close
5633 - don't use request in mdc_enqueue if we hit a timeout (889)
5634 - don't set the inode i_size for regular files from the MDS (896)
5635 - handle out of order completion AST (842)
5636 - don't LBUG if a lock request times out after receiving AST (913)
5637 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5638 - if a bad lock AST arrives, send an error instead of dropping entirely
5639 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5640 - fix leak in bulk IO when only partially completed (899, 900, 926)
5641 - fix O_DIRECT for ia64 (55)
5642 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5643 - C-z after timeout could hang a process forever; fixed (977)
5645 - client-side I/O cache (678, 924, 929, 941, 970)
5647 - READPAGE and SETATTRs which don't take server-side locks get
5650 2003-02-11 Phil Schwan <phil@clusterfs.com>
5653 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5654 - Fix fsfilt_extN_readpage() to read a full page of directory
5655 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5656 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5657 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5658 - Most intent operations were removed, in favour of a new RPC mode
5659 that does a single RPC to the server and bypasses most of the VFS
5660 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5661 - Aggressively cancel local locks on DLM servers
5662 - mds_reint_unlink sends EA to the client if it's the last nlink.
5663 client uses that EA to unlink OST objects.
5664 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5665 - recursive symlinks were fixed (439)
5666 - fixed NULL deref in DEBUG_REQ
5667 - filter_update_lastobjid no longer calls sync, which annoyed extN
5668 - fixed multi-client small-writes to a single file problem (445)
5669 - fixed mtime updates during file writes (607)
5670 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5671 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5672 - fixed Linux OST size reporting problem (444, 656)
5673 - OST now updates object mtime with writes or setattr (607, 619)
5674 - client verifies file size before zeroing page past EOF (445)
5675 - OST now writes last allocated objid to disk with allocation (108)
5676 - LOV on echo now works (409)
5678 - mds_reint_unlink sends a new buffer, with the EA included. this
5679 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5680 set if a regular file was being unlinked, and it was the last link
5681 - use PtlGet from the target for bulk writes (315)
5682 - OST now updates object mtime with writes or setattr (607, 619)
5683 - LDLM now has a grant-time callback to revalidate locked items, if
5685 - Many MDS operations were reorganized to combat race conditions
5687 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5688 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5690 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5693 - Fully reactivate OST imports after reconnection (512, others)
5694 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5695 - More graceful error handling for truncating on dead OST (515)
5696 - Don't error out unless we're actually accessing dead stripes (474)
5697 - Fix garbage sizes when stripes are missing (410)
5698 - LRU counters were broken, causing constant lock purge (433, 432)
5699 - garbage on read from stripes with failed OSTs (441)
5700 - mark OSCs as active before reconnecting during recovery (438)
5701 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5702 - lfind did not preserve OST order in output (443)
5703 - symlinks cause hung clients, incorrect data (439)
5704 - stop dereferencing request after dropping refcount (457)
5705 - don't LASSERT(spin_is_locked) on non-SMP (455)
5706 - fixes for many rename() bugs
5707 - fstat didn't correctly synchronize attributes (399)
5708 - server must handle lock cancellation during blocking AST prep (487)
5709 - bulk descriptors were free()d too soon (511)
5710 - fix paths in lconf, which would load incorrect modules (451, 507)
5711 - fix confusing lconf 'host not found' error message (386)
5712 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5713 - fix race condition in mdc_blocking_ast() for inode access (526)
5714 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5715 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5716 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5717 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5718 - fix lstripe in several ways (kernel side) (527)
5719 - fix request leak in ldlm_cli_enqueue (262)
5720 - incorrect OSC was marked inactive after OST failure
5721 - call mds_fs_cleanup before unmounting filesystem (524)
5722 - fix races between taking ns_lock and ldlm_lock_change_resource
5723 - fix races updating LOV export open file list
5724 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5725 - fix recovery NULL deref in ldlm_cli_cancel_unused
5726 - fix some DLM races by using new hash table for lock handles (419)
5727 - permit the client to specify desired inodes, at replay
5728 - duplicate requests when we queue them for replay reintegration
5729 - fix last_rcvd offset calculation
5730 - sync after each recovered transaction, so we always make progress
5731 - never, not always, ERESTART requests without transnos
5732 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5734 - skip replay if the MDS says that the client is already connected
5735 - don't check for a recovery-enabled export to match lctl's UUID
5736 - don't INC_USE_COUNT for phantom exports
5737 - don't crash when cleaning up phantom exports (567)
5738 - don't double-finish or set replay data for errored mdc_open requests
5739 - abort requests when they time out, so we don't get old replies
5740 - send/receive replies for AST messages again
5741 - if the client says that it doesn't have the lock, cancel it on the
5743 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5744 mark it as destroyed, it will all work out when decref is called
5745 - fix module use counts (22, 581)
5747 - ASTs now expect a reply (server cancels lock on error reply)
5749 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5752 - fix many simultaneous client startup (392)
5753 - fix dentry->d_it clobbering
5754 - credentials weren't being shipped for readdir/getattr operations
5755 - remove invalid assertions triggered during some concurrent MD
5757 - proper Lustre versions added (336, 389)
5758 - fix memory leak for create error case (398)
5759 - fix LOV locking bug that would get cli/srv out of sync
5760 - fix echo client over LOV (409)
5761 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5762 - fix double-O_EXCL intent crash (424)
5763 - avoid sending multiple lock CANCELs (352)
5765 - MDS can do multi-client recovery (modulo bugs in new code)
5767 - many updates, edits, cleanups
5769 2002-11-18 Phil Schwan <phil@clusterfs.com>
5772 - fix null d_it dereference (346)
5773 - fix full OST/dbench hang (333)
5774 - fix permission problem with file removal (286)
5775 - fix removal of OSCs from LOV when they fail
5776 - fix NULL deref during bulk timeout (214)
5777 - fix problems related to multiple filesystems on one MDS (241)
5778 - fixed serious subtle metadata locking bugs
5779 - free locks on clients when inodes are removed due to memory
5781 - fix inode pointer in lock data (285)
5782 - partial support for multiple MDS on a single host (241)
5783 - data locks weren't cancelled at clear_inode time (290, 311)
5784 - intent locks could lead to unbounded lock growth (205)
5785 - added a maximum lock count, an LRU list, and a flusher
5786 - fix multiple rename (365)
5787 - properly abstracted the echo client
5788 - OSC locked 1 byte too many; fixed
5789 - rewrote brw callback code:
5790 - fixed recovery bugs related to LOVs (306)
5791 - fixed too-many-pages-in-one-write crash (191)
5792 - fixed (again) crash in sync_io_timeout (214)
5793 - probably fixed callback-related race (385)
5795 - Add capability to MDS protocol
5796 - LDLM cancellations and callbacks on different portals
5798 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5801 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5802 - defer open object destruction to close time (601981, 138)
5803 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5804 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5805 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5806 - don't call obd_setattr in ll_file_release for destroyed objects
5807 * protocol change to lustre_msg: move |version| and add |flags|
5808 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5809 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5810 * added replay of create, unlink, link and rename operations during
5811 MDS failover; recovery should be much more robust now
5812 * remove failed OSCs from LOVs (only lov_create uses this so far)
5813 * the lustre-HOWTO was brought (more) up to date (582544)
5815 2002-10-23 Phil Schwan <phil@clusterfs.com>
5818 - in-use dentries weren't being reused properly (617851)
5819 - prevent multiple LDLM setup (599178)
5820 - fix LOV size calculations for truncate (617853)
5821 - fix client handling of MDS intent errors (POSIX)
5822 - fix permission bug in lovstripe.c test (624321)
5823 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5824 - truncate past end of file could corrupt data
5825 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5826 - a race in recovery could return ETIMEDOUT to apps (623947)
5827 - building outside the source directory was fixed
5828 * the lustre-HOWTO was brought (more) up to date (582544)
5829 * major progress was made on recovery functionality
5831 2002-10-10 Phil Schwan <phil@clusterfs.com>
5834 - recovery deadlock fix
5835 - rm -rf causes LBUG fix (617817)
5836 - file open by multiple tasks fix (618962)
5837 - directory permissions bugs (602707 and 620007)
5838 - journal_stop fixed with locking (611313)
5839 - O_APPEND failures resolved (618273, perhaps 614459)
5840 - lconf PATH fix (619770)
5841 - IA64 build fix (621450)
5842 - RPC buffer sizes scale with amount of memory
5844 2002-10-01 Phil Schwan <phil@clusterfs.com>
5847 - locks would be cancelled without throwing away data pages,
5848 resulting in inconsistent data (605627)
5849 - inode attributes were not always being refreshed (605627, 612449)
5850 - lconf now continues to cleanup after lctl reports an error
5851 - MDS now enforces user permissions (602707)
5852 - lprocfs cleanup fixed, but not yet enabled (614157)
5853 - fixed infinite server hang, should a client not respond to an AST
5854 - avoid going into recovery if user calls readlink() with a buffer
5855 that's too small (613941)
5856 - AST RPCs no longer require replies (614867) -- this may be changed
5857 - don't crash server if client sends an IOV that's too big (611336)
5858 - fixed lock conversion deadlock (611892)
5859 - fixed the following of symlinks (614622)
5860 * recovery: the server can remove locks from a client that dies, other
5861 clients can make progress
5862 * more extN patch fixes
5863 * compile-time configurable ptlrpc buffer allocations
5865 - collaborative read cache document
5866 - Lustre Lite Performance CDR document-in-progress
5868 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5871 - fix typo in patch-2.4.18
5873 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5876 - clear ptlrpc request each time in handle_incoming_request()
5877 - unlink of files now destroys the object on the OST
5879 2002-09-19 Peter Braam <braam@clusterfs.com>
5881 * add hard link support
5882 * change obdfile creation method
5883 * kernel patch changed
5885 2002-09-19 Peter Braam <braam@clusterfs.com>
5888 - stack overflow bug in extN fixed
5890 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5892 * documentation updates
5893 - add man pages for config tools
5894 - update tests/README to describe testing with new config tools
5895 - finish metadata API descriptions
5896 * bug fixes and cleanups
5897 - statfs workaround for 16TB limit
5898 - LOV stripe allocation improved, can stripe on subset of OSTs
5899 - LOV file size/IO offset was wrong for files > 4GB in size
5900 - object EA data was being dropped, caused files to be unreadable
5901 - memory overflow with non-LOV OST caused memory corruption
5902 - fixed regression tests to work with new config tools, obdfilter
5903 - fixed bug when directory size became larger than 1 block
5904 - fixed bug (for single client case) when PWD was deleted
5905 - invalidate local directory pages when doing intent-based ops
5906 - avoid LDLM oops when lock callback contained bad data
5908 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5910 * documentation updates
5911 * bug fixes and cleanups
5912 - configuration tools
5915 - 64-bit compile warnings
5916 - 64-bit internal statfs data
5918 * test_brw on persistent OST devices
5920 * lprocfs (disabled)
5922 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5924 * documentation updates
5925 * bug fixes and cleanups
5926 * configuration tools
5928 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5931 * many small fixes to 0_5_4
5932 * io/network handling
5933 * thinkos in MDS operations
5935 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5938 * crucial basic fixes to 0.5.3
5939 * IOR, Iozone work over Elan
5942 2002-08-07 Phil Schwan <phil@clusterfs.com>
5943 * version 0_5_3, our first alpha
5944 * we use the new Portals iovs
5945 * documentation updates
5946 * bug fixes and cleanups
5947 * small changes in the DLM wire protocol
5949 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5950 * version 0_5_1 with some initial stability,
5951 * locking on MD and file I/O.
5952 * documentation updates
5953 * several bug fixes since 0.5.0
5954 * small changes in wire protocol
5956 2002-07-18 Phil Schwan <phil@clusterfs.com>
5958 * delivered as Lustre Light Alpha
5959 * fixed a crash after handling invalid MDS requests
5960 * fixed directory pages for architectures with non-4k pages sizes
5962 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5963 * release version v0_4_4
5964 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5966 * Fixes a number of interruption problems with OST operations.
5967 * Update documentation for portals header changes
5968 * Move all wire protocol structs/defines to lustre_idl.h
5969 * Fixes symlink length bug.
5970 * Add tcpdump to repository.
5972 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5973 * release version v0_4_3
5974 * Fixes statfs for inodes on extN.
5975 * Fixes bug in runtests which would delete /etc/hosts.
5976 * Use 64-bit object IDs wherever possible (not into VFS though)
5977 Remove ost_get_info, which is unused by lustre, and out of date.
5979 2002-07-03 Peter Braam <braam@clusterfs.com>
5980 * release version v0_4_2 Fixes a lookup error (type not passed)
5981 * move forward to head of Portals
5982 * move forward to latest Lustre kernel
5984 2002-06-25 Peter Braam <braam@clusterfs.com>
5985 * release version v0_4_1. Hopefully stable on single node use.