1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.7.1.el5 (RHEL 5),
6 2.6.18-128.7.1.el5 (OEL 5),
7 2.6.22.14 vanilla (kernel.org).
8 * Client support for unpatched kernels:
9 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
10 2.6.16 - 2.6.21 vanilla (kernel.org)
11 * Recommended e2fsprogs version: 1.41.6.sun1
12 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
13 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
14 removed cwd "./" (refer to Bugzilla 14399).
15 * File join has been disabled in this release, refer to Bugzilla 16929.
17 Severity : enhancement
19 Description: Add support for OEL5.
23 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
27 Description: client could not reconnect to OST because of an active request.
28 Details : The client now retries to reconnect to the same server, if a
29 connect request failed with EBUSY or -EAGAIN. Test case 35b
30 conf-sanity.sh has been created.
34 Description: Don't allow make backward step on assignin osc next id.
35 Details : race between allocation next id and ll_sync thread can be cause
36 of set wrong osc next id and can be kill valid ost objects.
40 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
44 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
48 Description: Error handling in osc_statfs_interpret() has been improved.
49 Details : Check in osc_statfs_interpret() for EBADR.
53 Description: Do not send statfs() requests to OSTs disabled by administrator.
54 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
56 Severity : enhancement
58 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
60 Severity : enhancement
62 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
64 Severity : enhancement
66 Description: Enhancement for lfs(1) command to use numeric uid/gid.
70 Description: Improved cleanup procedure for conf-sanity test_46a.
74 Description: Do not start bulk transfer, if client tried to reconnect.
75 Details : Do not start bulk transfer, if client tried to reconnect,
76 but failed due to active rpcs. Check for exp_abort_active_req
81 Description: Prevent inconsistences between linux and lustre mount structures.
82 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
83 Make the sleep interruptible.
85 Severity : enhancement
87 Description: provide server to client comms path
88 Details : server to client communications path via new LDLM_SET_INFO rpc
92 Description: truncate starts GFP_FS allocation under transaction causing deadlock
93 Details : ldiskfs_truncate calls grab_cache_page which may start page
94 allocation under an open transaction. This may lead to
95 calling prune_icache with consequent lustre reentrance.
99 Description: send correctly lsm on open replay
100 Details : MDS is trust to LSM size on replay open, but client can set wrong size
105 Description: enable client interface failover
106 Details : When a child reconnects from another NID, properly update export
107 nid hash position and ldlm reverse import.
109 Severity : enhancement
111 Description: Communicate OST degraded/readonly state via statfs to MDS
112 Details : Flags in the statfs returned from OSTs indicate whether the
113 OST is in a degraded RAID state, or if the filesystem has
114 turned read-only after a filesystem error is detected.
119 Description: don't panic if EPROTO was hit when reading symlink
120 Details : correctly handling request reference in error cases.
122 Severity : enhancement
124 Description: Add LustreNetLink, a kernel-userspace communcation path.
126 Severity : enhancement
128 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
133 Description: access to llog context before init.
134 Details : move handling CATALOGS file at osc layer and forbid access to llog
138 Frequency : rare, only when client is writing to OST on local node
140 Description: Avoid deadlock for local client writes
141 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
142 memory freeing context. This allows OST threads to set the
143 PF_MEMALLOC flag on task structures in order to allocate memory
144 from reserved pools and complete IO.
145 Use GFP_HIGHUSER for OST allocations for non-local client writes,
146 so that the OST threads generate memory pressure and allow
147 inactive pages to be reclaimed.
149 Severity : enhancement
151 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
156 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
157 Details : move ldlm namespace creation in setup phase to avoid grab
158 _lprocfs_lock with cli_sem held.
162 Description: Temporarily disable grant shrink.
163 Details : Disable the feature for debugging.
166 Frequency : only when formatting test filesystems
168 Description: Unable to run several mkfs.lustre on loop devices at the same time
169 Details : mkfs.lustre returns error 256 on the concurrent loop devices
170 formatting. The solution is to proper handle the error.
172 Severity : enhancement
174 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
176 Severity : enhancement
178 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
183 Description: resolve race between obd_disconnect and class_disconnect_exports
184 Details : if obd_disconnect will be called to already disconnected export he
185 forget release one reference and osc module can't unloaded.
187 Severity : enhancement
189 Description: Allow tuning service thread via /proc
190 Details : For each service a new
191 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
192 created that can be used to set min/max thread counts, and get the
193 current number of running threads.
198 Descriptoin: don't return error if have partially created objects for file.
199 Details : lov_update_create_set uses set->set_success as index for created
200 objects, so if some requests will be failed, they will have hole
201 at end of array and qos_shrink_lsm for allocate correct lsm.
203 Severity : enhancement
205 Description: Update OFED support to 1.4.1
210 Description: Reduce small size read RPC
211 Details : Set read-ahead limit for every file and only do read-ahead when
212 available read-ahead pages are bigger than 1M to avoid small size
215 Severity : enhancement
217 Description: move AT tunable parameters for more consistent usage
218 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
220 Severity : enhancement
222 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
223 Details : allow skip disconnected ost for send statfs request and hide error
229 Description: fix racy locking of mballoc block bitmaps causing BUG
230 Details : The locking of the mballoc buddy bitmap and the in-memory
231 block bitmap was using two different spin locks in some
232 cases. This made it possible to incorrectly access the
233 mballoc bitmap while another process was modifying it,
234 causing a sanity assertion to fail. While no on-disk corruption
235 was reported, there was some risk of this happening.
238 Frequency : rare, on llog test 6
240 Descriptoin: don't allow connect to already connected import
241 Details : allowing connect to already connected import is hide connecting problem.
246 Descriptoin: allow kill process which wait statahead result
247 Details : for some reasons 'ls' can stick in waiting result from statahead,
248 in this case need way for kill this process.
250 Severity : enhancement
252 Description: Add state history info file, enhance import info file
253 Details : Track import connection state changes in a new osc/mdc proc file;
254 add overview-type data to the osc/mdc import proc file.
256 Severity : enhancement
258 Description: MDS create should not wait for statfs RPC while holding DLM lock.
260 Severity : enhancement
262 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
267 Descriptoin: OST grant too much space to client even there are not enough space.
268 Details : Client will shrink its grant cache to OST if there are no write
269 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
270 this grant cache if there are already not enough avaible space
271 (left_space < total_clients * 32M).
274 Frequency : start MDS on uncleanly shutdowned MDS device
276 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
277 Details : stay in waiting mds<>ost recovery finished produce random bugs
278 due race between two ll_sync thread for one lov target. send
279 ACTIVATE event only if connect realy finished and import have
283 Frequency : rare, connect and disconnect target at same time
285 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
286 Details : don't call obd_disconnect under lov_lock. this long time
287 operation and can block ptlrpcd which answer to connect request.
292 Descriptoin: don't lose wakeup for imp_recovery_waitq
293 Details : recover_import_no_retry or invalidate_import and import_close can
294 both sleep on imp_recovery_waitq, but we was send only one wakeup
298 Frequency : always with long access acl
300 Descriptoin: mds can't pack reply with long acl.
301 Details : mds don't control size of acl but they limited by reint/getattr
304 Severity : enhancement
306 Description: Update to SLES10 kernel-2.6.16.60-0.33.
308 Severity : enhancement
310 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
313 Frequency : start MDS on uncleanly shutdowned MDS device
315 Descriptoin: aborting recovery hang on MDS
316 Details : don't throttle destroy RPCs for the MDT.
319 Frequency : on remount
321 Description: external journal device not working after the remount
322 Details : clear dev_rdonly flag for external journal devices in
328 Description: shutdown vs evict race
329 Details : client_disconnect_export vs connect request race.
330 if client will evicted at this time - we start invalidate
331 thread without referece to import and import can be freed
335 Frequency : rare, need acl's on inode.
337 Description: client can't handle ost additional correctly
338 Details : if ost was added after client connected to mds client can have
339 hit lnet_try_match_md ... to big messages to wide striped files.
340 in this case need teach client to handle config events about add
341 lov target and update client max ea size at that event.
343 Severity : enhancement
345 Description: Changelogs
346 Details : Changelogs are a lightweight mechanism to track filesystem
347 metadata and namespace changes. The changelog is recorded
348 permanently on the MDTs, and is periodically "consumed" / purged
349 when records are no longer needed.
351 Severity : enhancement
353 Description: compact fld format with extents
354 Details : Store range of seq rather than every seq in FLD. Seq
355 controller update FLD rather than clients. In Case of CMD, mdt0
356 has FLD, all other metadata server act as non persistent proxy
357 for FLD queries and cache fld entries in fld cache.
362 Description: don't skip ost target if they assigned to file
363 Details : Drop slow OSCs if we can, but not for requested start idx.
364 This means "if OSC is slow and it is not the requested
365 start OST, then it can be skipped, otherwise skip it only
366 if it is inactive/recovering/out-of-space.
370 Description: more cleanup in mds_lov
371 Details : not send LOV EA under replay, we can't know about they size at this
372 time. Don't allow client connect to mds before any ost connected,
373 for avoid problems with LOV EA size and returning EIO to client.
375 Severity : enhancement
377 Description: Interoperability at server side (Disk interoperability)
379 Severity : enhancement
381 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
383 Severity : enhancement
385 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
387 Severity : enhancement
389 Description: New client IO stack (CLIO).
391 Severity : enhancement
393 Description: Commit on sharing. Eliminate inter-client dependencies between
394 uncommitted transactions by doing transaction commits.
395 Thereby clients may recovery independently.
398 Frequency : Create a symlink file with a very long name
400 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
401 Details : If there is no extra space in the request for early cancels,
402 ldlm_req_handles_avail() returns 0 instead of a negative value.
404 Severity : enhancement
406 Description: Add /proc entry for import status
407 Details : The mdc, osc, and mgc import directories now have
408 an import directory that contains useful import data for debugging
411 Severity : enhancement
413 Description: Re-disable certain /proc logging
414 Details : Enable and disable client's offset_stats, extents_stats and
415 extents_stats_per_process stats logging on the fly.
418 Frequency : Only on FC kernels 2.6.22+
420 Description: oops in statahead
421 Details : Do not drop reference count for the dentry from VFS when lookup,
422 VFS will do that by itself.
424 Severity : enhancement
426 Description: Generic /proc file permissions
427 Details : Set /Proc file permissions in a more generic way to enable non-
428 root users operate on some /proc files.
432 Description: Hitting mdc_commit_close() ASSERTION
433 Details : Properly handle request reference release in
434 ll_release_openhandle().
438 Description: quota recovery deadlock during mds failover
439 Details : This patch includes att18982, att18236, att18237 in bz14840.
441 1. fix osts hang when mds does failover with quotaon
442 2. prevent watchdog storm when osts threads wait for the
447 Frequency : only patchless client
448 Description: add workaround for race between add/remove dentry from hash
450 Severity : enhancement
452 Description: Allow OST glimpses to return PW locks
456 Description: LBUG when llog conf file is full
457 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
461 Description: Prevent import from entering FULL state when server in recovery
465 Description: service mount cannot take device name with ":"
466 Details : Only when device name contains ":/" will mount treat it as
472 Description: replace ptlrpcd with the statahead thread to interpret the async
473 statahead RPC callback
477 Frequency : on recovery
478 Description: I/O failures after umount during fail back
479 Details : if client reconnected to restarted server we need join to recovery
480 instead of find server handler is changed and process self
481 eviction with cancel all locks.
483 Severity : enhancement
485 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
487 Severity : enhancement
489 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
491 Severity : enhancement
493 Description: Upcall on Lustre log has been dumped
494 Details : Allow for a user mode script to be called once a Lustre log has
495 been dumped. It passes the filename of the dumped log to the
496 script, the location of the script can be specified via
497 /proc/sys/lnet/debug_log_upcall.
502 Description: avoid idr_remove called for id which is not allocated.
503 Details : Move assigment s_dev for clustered nfs to end of initialization,
504 to avoid problem with error handling.
509 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
510 Details : When connection is reused this not moved from CONN_UNUSED_HASH
511 into CONN_USED_HASH and this prodice warning when put connection
512 again in unused hash.
515 Severity : enhancement
517 Description: File striping can now be set to use an arbitrary pool of OSTs.
519 Severity : enhancement
521 Description: Export bytes_read/bytes_write count on OSC/OST.
525 Description: Early reply size mismatch, MGC loses connection
526 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
527 the connect flags are properly negotiated.
532 Description: quotas are not honored with O_DIRECT
533 Details : all writes with the flag O_DIRECT will use grants which leads to
534 this problem. Now using OBD_BRW_SYNC to guard this.
538 Description: add quota statistics
539 Details : 1. sort out quota proc entries and proc code.
540 2. add quota statistics
542 Severity : enhancement
544 Description: enable quota support for HEAD.
548 Description: Properly propagate oinfo flags from lov to osc for statfs
549 Details : restore missing copy oi_flags to lov requests.
551 Severity : enhancement
553 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
554 : l_getgroups(8), lst(8), routerstat(8)
555 Details : included man pages for llobdstat(8), llstat(8),
556 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
558 Severity : enhancement
560 Description: configure's --enable-quota should check the
561 : kernel .config for CONFIG_QUOTA
562 Details : configure is terminated if --enable-quota is passed but
563 : no quota support is in kernel
567 Description: Remove portals compatibility
568 Details : Remove portals compatibility, not interoperable with releases
573 Description: Resolve device initialization race
574 Details : Prevent proc handler from accessing devices added to the
575 obd_devs array but yet be intialized.
577 Severity : enhancement
579 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
581 Severity : enhancement
583 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
588 Description: Using wrong pointer in osc_brw_prep_request
589 Details : Access to array[-1] can produce panic if kernel compiled with
590 CONFIG_PAGE_ALLOC enabled
594 Description: Client runs out of low memory
595 Details : Consider only lowmem when counting initial number of llap pages
599 Description: *optional* service tags registration
600 Details : if the "service tags" package is installed on a Lustre node
601 When the filesystem is mounted, a local-node service tag will
602 be created. See http://inventory.sun.com/ for more information
603 about the Service Tags asset management system.
607 Description: Kernel BUG tries to release flock
608 Details : Lustre does not destroy flock lock before last reference goes
609 away. So always drop flock locks when client is evicted and
610 perform unlock regardless of successfulness of speaking to MDS.
614 Description: add refcount protection for osc callbacks, avoid panic on shutdown
618 Description: sanity test 65a fails if stripecount of -1 is set
619 Details : handle -1 striping on filesystem in ll_dirstripe_verify
624 Description: ASSERTION(CheckWriteback(page,cmd)) failed
625 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
628 Severity : enhancement
630 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
634 Description: do not process already freed flock
635 Details : flock can possibly be freed by another thread before it reaches
636 to ldlm_flock_completion_ast.
640 Description: LBUG during stress test
641 Details : Need properly lock accesses the flock deadlock detection list.
645 Description: oops in page fault handler
646 Details : kernel page fault handler can return two special 'pages' in error
647 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
651 Description: timeout with invalidate import.
652 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
653 should be handled by ptlrpcd. This produce long age waiting and
654 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
656 Severity : enhancement
658 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
662 Description: enable MGS and MDT services start separately
663 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
664 a co-located MGS without starting the MGS, which is a complement
665 to 'nosvc' mount option.
669 Frequency : after recovery
670 Description: precreate to many object's after del orphan.
671 Details : del orphan st in oscc last_id == next_id and this triger growing
672 count of precreated objects. Set flag LOW to skip increase count
673 of precreated objects.
677 Frequency : rare, on clear nid stats
678 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
679 Details : when clean nid stats sometimes try destroy live entry,
680 and this produce panic in free.
684 Description: Stack overflow during MDS log replay
685 ease stack pressure by using a thread dealing llog_process.
689 Description: wait until IO finished before start new when do lock cancel.
690 Details : VM protocol want old IO finished before start new, in this case
691 need wait until PG_writeback is cleared until check dirty flag and
692 call writepages in lock cancel callback.
694 Severity : enhancement
696 Description: using special macro for print time and cleanup in includes.
700 Description: mds_mfd_close() ASSERTION(rc == 0)
701 Details : In mds_mfd_close(), we need protect inode's writecount change
702 within its orphan write semaphore to prevent possible races.
706 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
710 Frequency : rare, on shutdown ost
711 Description: don't hit live lock with umount ost.
712 Details : shrink_dcache_parent can be in long loop with destroy dentries,
713 use shrink_dcache_sb instead.
717 Description: don't panic with use echo client
718 Details : echo client pass NULL as client nid pointer and this produce null
723 Description: fix build on ppc32
724 Details : compile code with -m64 flag produce wrong object file for ppc32.
728 Description: add message levels for liblustreapi
732 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
733 Details : Move the 'good_osts' check before the 'total_bavail' check. This
734 will result in an -EAGAIN and in the exit call path we call
735 alloc_rr() which will with increasing aggressiveness attempt to
736 aquire precreated objects on the minimum number of required OSCs.
740 Description: Use old size assignment to avoid deadlock
741 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
742 the scheduling while holding a spinlock. We do not need locking
743 for size in ll_update_inode() because size is only updated from
744 the MDS for directories or files without objects, so there is no
745 other place to do the update, and concurrent access to such inodes
746 are protected by the inode lock.
750 Description: resolve "_IOWR redefined" build error on SLES10
754 Description: dump the memory debugging after all modules are unloaded to
755 suppress false negative in conf_sanity test 39
757 Severity : enhancement
759 Description: build kernel-ib packages for OFED 1.3 in our release cycle
764 Description: fix SLES kernel versioning
765 Details : the kernel version for our SLES 10 kernel did not include a "-"
766 before the "smp" at the end. while this was not a problem in
767 general, it did mean that software trying to use the kernel
768 version to try to detect a vendor specific kernel would fail.
769 this was most evident by the OFED build scripts.
773 Description: Don't update lov_desc members until making sure they are valid
774 Details : When updating lov_desc members via proc fs, need fix their
775 validities before doing the real update.
779 Description: don't put request into delay list while invalidate in flight.
780 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
781 import in flight. this produce timeout for invalidate and sometimes
782 can cause stale data.
786 Frequency : on ppc only
787 Description: not convert ost objects for directory because it's not exist.
788 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
792 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
794 Severity : enhancement
796 Description: Update to support 2.6.22.14 vanilla kernel.
800 Frequency : rare, on recovery
801 Description: read procfs can produce deadlock in some situation
802 Details : Holding lprocfs lock which send rpc can produce block for destroy
803 obd objects and this also block reconnect with -EALREADY. This
804 isn't fix all lprocfs bugs - but make it rare.
806 Severity : enhancement
808 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
811 Frequency : frequent on X2 node
813 Description: mdc_set_open_replay_data LBUG
814 Details : Set replay data for requests that are eligible for replay.
818 Description: lustre_mgs: operation 101 on unconnected MGS
819 Details : When MGC is disconnected from MGS long enough, MGS will evict the
820 MGC, and late on MGC cannot successfully connect to MGS and a lot
821 of the error messages complaining that MGS is not connected.
825 Frequency : on network error
826 Description: panic with double free request if network error
827 Details : mdc_finish_enqueue is finish request if any network error ocuring,
828 but it's true only for synchronus enqueue, for async enqueue
829 (via ptlrpcd) this incorrect and ptlrpcd want finish request
832 Severity : enhancement
834 Description: client-side metadata stat-ahead during readdir(directory readahead)
835 Details : perform client-side metadata stat-ahead when the client detects
836 readdir and sequential stat of dir entries therein
839 Frequency : on start mds
841 Description: Implement get_info(last_id) in obdfilter.
844 Frequency : occasional
846 Description: Correctly check stale fid, not start epoch if ost not support SOM
847 Details : open with flag O_CREATE need set old fid in op_fid3 because
848 op_fid2 was overwritten with new generated fid, but mds can answer
849 with one of these two fids and both is not stale. Setattr
850 incorrectly started an epoch and assume will be called
851 done_writing, but without SOM done_writing ever being called.
854 Frequency : rare, depends on device drivers and load
856 Description: MDS or OSS nodes crash due to stack overflow
857 Details : Code changes in 1.8.0 increased the stack usage of some functions.
858 In some cases, in conjunction with device drivers that use a lot
859 of stack the MDS (or possibly OSS) service threads could overflow
860 the stack. One change which was identified to consume additional
861 stack has been reworked to avoid the extra stack usage.
864 Frequency : occasional
866 Description: Do not fail import if osc_interpret_create gets -EAGAIN
867 Details : If osc_interpret_create got -EAGAIN it immediately exits and
868 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
869 oscc_has_objects and see OSC has no objests and call
870 oscc_internal_create to resend create request.
872 Severity : enhancement
874 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
876 Severity : enhancement
878 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
881 Frequency : very rare
883 Description: Fix replay if there is an un-replied request and open
884 Details : In some cases, older replay request will revert the
885 mcd->mcd_last_xid on MDS which is used to record the client's
888 Severity : enhancement
890 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
892 Severity : enhancement
894 Description: Add rhel5 support to HEAD.
896 Serverity : enhancement
898 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
903 Description: Don't allow skipping OSTs if index has been specified.
904 Details : Don't allow skipping OSTs if index has been specified, make
905 locking in internal create lots better.
909 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
910 Details : access to bitfield in structure is always rounded to long
911 and this produce problem with not atomic change any bit.
915 Description: Lustre make rpms failed.
916 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
917 builds Lustre rpms from tarball.
920 Frequency : rare on shutdown ost
922 Description: If llog cancel was not send before clean_exports phase, this can
923 produce deadlock in llog code.
924 Details : If llog thread has last reference to obd and call class_import_put
925 this produce deadlock because llog_cleanup_commit_master wait when
926 last llog_commit_thread exited, but this never success because was
927 called from llog_commit_thread.
931 Description: allow userland application know is lost one of stripes.
932 Details : fill lvb_blocks with error code on ost and return it to
933 application if error flag found.
937 Description: NULL lov_tgts causing MDS oops
938 Details : more safe checks for NULL lov_tgts for avoid oops.
940 Severity : enhancement
942 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
946 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
947 Details : If an OST is down the MDS will hang indefinitely in
948 obd_statfs_rqset() waiting for the statfs data. While for
949 MDS QOS usage of statfs, it should not stuck in waiting.
951 Severity : enhancement
953 Description: remote_acl support
954 Details : Support ACL-based permission check for remote user.
955 Support setfacl/getfacl for remote user with the utils
956 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
957 the system "{s,g}etfacl" utils.
959 Severity : enhancement
961 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
963 Severity : enhancement
965 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
969 Description: make mgs_setparam() handle fsname containing dash
970 Details : fsname containing a dash does not work with lctl conf_param
972 Severity : enhancement
974 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
976 Severity : enhancement
978 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
982 Frequency : rare, at shutdown
983 Description: access already free / zero obd_namespace.
984 Details : if client_disconnect_export was called without force flag set,
985 and exist connect request in flight, this can produce access to
986 NULL pointer (or already free pointer) when connect_interpret
987 store ocd flags in obd_namespace.
991 Frequency : only at startup
992 Description: not alloc memory with spinlock held.
993 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
994 if any spinlock held.
996 Severity : enhancement
998 Description: make lustre randomly fail allocating memory
999 Details : Make lustre randomly failed allocating memory for testing purpose.
1001 Severity : enhancement
1003 Description: lost problems with lov objid file
1004 Details : Fixes some scability and access to not inited memory problems
1005 in work with lov objdid file.
1010 Description: lfs find does not continue on file error
1011 Details : Continue other files processing when a file/dir is absent.
1015 Description: Inconsistent usage of lustre_pack_reply()
1016 Details : Standardize the usage of lustre_pack_reply() such that it
1017 always generate a CERROR on failure.
1020 Frequency : occasional
1022 Description: MDS hang or stay in waiting lock
1023 Details : If client receive lock with CBPENDING flag ldlm need send lock
1024 cancel as separate rpc, to avoid situation when cancel request
1025 can't processed due all i/o threads stay in wait lock.
1029 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1033 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1037 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1041 Description: Update to RHEL4 latest kernel.
1043 Severity : enhancement
1045 Description: Build SLES10 patchless client fails
1046 Details : The configure was broken by run ./configure with
1047 --with-linux-obj=.... argument for patchless client. When the
1048 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1049 can't search header adequately. Use absolute path such as
1050 -I($LINUX)/include instead.
1054 Description: interrupt oig_wait produce painc on resend.
1055 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1056 private set, and this produce situation when rq_ptlrpcd_data not
1057 copyed to new allocated request and triggered LBUG on assert
1058 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1059 ptlrpcd set if request is changed and can be safety changed to use
1064 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1065 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1066 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1067 these macros and replace them with their difinitions instead.
1071 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1072 Details : Update which_patch & target file for SLES10 latest kernel.
1075 Description: Remove client patches from SLES 10 kernel.
1076 Details : This causes SLES 10 clients to behave as patchless clients
1077 even on a Lustre-patched (server) kernel.
1079 Severity : enhancement
1081 Description: self-adjustable client's lru lists
1082 Details : use adaptive algorithm for managing client cached locks lru
1083 lists according to current server load, other client's work
1084 pattern, memory activities, etc. Both, server and client
1085 side namespaces provide number of proc tunables for controlling
1088 Severity : enhancement
1090 Description: light-weight GSS support
1091 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1092 and reduce performance overhead.
1094 Severity : enhancement
1096 Description: Linux keyring support
1097 Details : Support using service of Linux keyring for Lustre GSS internal
1098 context refresh/cache mechanism.
1102 Description: Fix errors in lfs documentation
1103 Details : Fixes man pages
1107 Description: don't use GFP_* in generic Lustre code.
1108 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1112 Description: obdclass is limited by single OBD_ALLOC(idarray)
1113 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1117 Description: warnings with build patchless client with vanila 2.6.19 and up
1118 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1119 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1123 Frequency : Only for SLES
1124 Description: sanity_quota fail test_1
1125 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1126 file, which makes TSTID[2] inunique.
1130 Frequency : Only for SLES9
1131 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1132 Details : Update target/ChangeLog/which_patch .
1136 Frequency : Only for test_10 in sanity-quota.sh
1137 Description: a bug in quota test code
1138 Details : checking a test flag in a wrong place
1142 Frequency : Only for test_13 in sanity-quota.sh
1143 Description: a bug in quota test script
1144 Details : shouldn't check overfull allocated quota
1148 Description: jbd statistics
1149 Details : Port older jbd statistics patch for sles10
1151 Severity : enhancement
1153 Description: Add printing inode info into message about error in writepage.
1158 Description: reply_lock_interpret crash due to race with it and lock cancel.
1159 Details : Do not replay locks that are being cancelled. Do not reference
1160 locks by their address during replay, just by their handle.
1164 Frequency : When flocks are used.
1165 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1166 Details : Flock locks might destroy just granted lock if it could be merged
1167 with another existing flock, this is done in completion handler,
1168 so teach ldlm_cli_enquque_fini that this is a valid case for
1174 Description: Oops in read and write path when failing to allocate lock.
1175 Details : Check if lock allocation failed and return error back.
1179 Description: lstripe command fails for valid OST index
1180 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1181 instead of lov->desc.ld_active_tgt_count.
1185 Description: sanity.sh failed test 103
1186 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1191 Description: df doesn't work properly if diskfs blocksize != 4K
1192 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1196 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1197 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1199 Severity : enhancement
1201 Description: Adaptive timeouts
1202 Details : RPC timeouts adapt to changing server load and network
1203 conditions to reduce resend attempts and improve recovery time.
1207 Description: llapi_file_create() does not allow some changes
1208 Details : add llapi_file_open() that allows specifying the mode and
1209 open flags, and also returns an open file handle.
1213 Description: merge and cleanup kernel patches.
1214 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1218 Description: Add journal checksum support.(Kernel part)
1219 Details : The journal checksum feature adds two new flags i.e
1220 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1221 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1222 indicates that the commit block contains the checksum for
1223 the blocks described by the descriptor blocks. Now commit
1224 record can be sent to disk without waiting for descriptor
1225 blocks to be written to disk. This behavior is controlled
1226 using JBD2_FEATURE_ASYNC_COMMIT flag.
1230 Description: OSS needs mutliple precreate threads
1231 Details : Add ability to start more than one create thread per OSS.
1235 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1236 Details : execute lfs setstripe on client
1240 Description: mds_obd_create error creating tmp object
1241 Details : When the user sets quota on root, llog will be affected and can't
1242 create files and write files.
1245 Frequency : Always on ia64 patchless client, and possibly others.
1247 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1248 Details : This allows the patchless client to be loaded on architectures
1249 without this export.
1253 Description: RedHat Update kernel for RHEL5
1254 Details : Add the kernel config file for RHEL5.
1258 Description: RedHat Update kernel for RHEL5
1259 Details : Modify the kernel config file more closer RHEL5.
1263 Description: Build failure against Centos5 (RHEL5)
1264 Details : Define PAGE_SIZE when it isn't present.
1268 Description: test_7 on sanity-quota.sh can't be used on separate machines
1269 Details : add facet to handle it
1273 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1274 Details : with NFS, the anon dentry's parent was set to itself in
1275 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1276 corresponding dentry other than use rec->ur_name.
1278 Severity : enhancement
1280 Description: lfs setstripe enhancement
1281 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1285 Description: enable data checksumming by default
1286 Details : enable checksum by default, allow --disable-checksum
1287 configure option and "-o nochecksum" mount option. Checksums
1288 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1292 Description: lustre support for RHEL5
1293 Details : Add support for RHEL5.
1297 Description: Client eviction due to failover config
1298 Details : after a connection loss, the lustre client should attempt to
1299 reconnect to the last active server first before trying the
1300 other potential connections.
1304 Description: when mds and osts use different quota unit(32bit and 64bit),
1305 quota will be released repeatly.
1306 Details : void sending multiple quota reqs to mds, which will keep the
1307 status between the reqs.
1311 Description: osts not allocated evenly to files
1312 Details : change the condition to increase offset_idx
1315 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1317 Description: Data corruption for OSTs that are formatted larger than 2TB
1319 Details : When generating the bio request for lustre file writes the
1320 sector number would overflow a temporary variable before being
1321 used for the IO. The data reads correctly from Lustre (which
1322 will overflow in a similar manner) but other file data or
1323 filesystem metadata may be corrupted in some cases.
1327 Description: Tune the kernel for good SCSI performance.
1328 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1329 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1334 Description: rewrite ext2-derived code in obdclass/uuid.c
1335 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1336 from scratch preserving functionality.
1340 Description: "lfs find -obd UUID" prints directories
1341 Details : "lfs find -obd UUID" will return all directory names instead
1342 of just file names. It is incorrect because the directories
1343 do not reside on the OSTs.
1347 Description: use i_size_read and i_size_write in 2.6 port
1348 Details : replace inode->i_size access with i_size_read/write()
1351 Frequency : when removing large files
1353 Description: scheduling issue during removal of large Lustre files
1354 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1355 It causes scheduling issues when removing large files (17TB in the
1359 Frequency : only with liblustre clients on XT3
1361 Description: evictions taking too long
1362 Details : allow llrd to evict clients directly on OSTs
1365 Frequency : only on ppc
1367 Description: /proc/fs/lustre/devices broken on ppc
1368 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1369 We should make sure the type of 'index' is loff_t and then cast
1370 explicitly as needed below. Do not assign an explicitly cast
1374 Frequency : only for rhel5
1376 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1377 Details : Modify the target file & which_kernel.
1379 Severity : enhancement
1381 Description: omit set fsid for export NFS
1382 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1383 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1387 Frequency : Always, for kernels after 2.6.16
1388 Description: Fix warning idr_remove called for id=.. which is not allocated.
1389 Details : Last kernels save old s_dev before kill super and not allow
1390 to restore from callback - restore it before call kill_anon_super.
1394 Description: buffer overruns could theoretically occur
1395 Details : llapi_semantic_traverse() modifies the "path" argument by
1396 appending values to the end of the origin string, and a
1397 overrun may occur. Adding buffer overrun check in liblustreapi.
1401 Description: Fix error on 'ls .' at the top of the Lustre mount.
1402 Details : Don't revalidate dentry if it is a root dentry.
1406 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1407 Details : Modify vm-tunables-rhel4.patch.
1411 Description: Kernel config for 2.6.18-vanilla.
1412 Details : Modify targets/2.6-vanilla.target.in.
1413 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1414 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1415 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1416 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1420 Description: improve handling recoverable errors
1421 Details : if request processig with error which can be recoverable on server
1422 request should be resend, otherwise page released from cache and
1427 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1428 Details : Modify target file & which_patch.
1429 A flaw was found in the IA32 system call emulation provided
1430 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1431 value could be stored in the %RAX register, which could trigger an
1432 out-of-bounds system call table access. An untrusted local user
1433 could exploit this flaw to run code in the kernel
1434 (ie a root privilege escalation). (CVE-2007-4573).
1438 Description: O_DIRECT bypasses client statistics.
1439 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1440 read_ahead_stats were updated. copied the stats section from
1441 osc_send_oap_rpc() into async_internal().
1445 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1449 Description: change order of libsysio includes
1450 Details : '#include sysio.h' should always come before '#include xtio.h'
1454 Description: handle "serious error: objid * already exists" more gracefully
1455 Details : If LAST_ID value on disk is smaller than the objects existing in
1456 the O/0/d* directories, it indicates disk corruption and causes an
1457 LBUG(). If the object is 0-length, then we should use the existing
1458 object. This will help to avoid a full fsck in most cases.
1460 Severity : enhancement
1462 Description: adapt the lustre_config script to support the upgrade case
1463 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1464 server targets to 1.6 in parallel.
1468 Description: To avoid grant space > avaible space when the disk is almost
1469 full. Without this patch you might see the error "grant XXXX >
1470 available" or some LBUG about grant, when the disk is almost
1472 Details : In filter_check_grant, for non_grant cache write, we should
1473 check the left space by if (*left > ungranted + bytes), instead
1474 of (*left > ungranted), because only we are sure the left space
1475 is enough for another "bytes", then the ungrant space should be
1476 increase. In client, we should update cl_avail_grant only there
1477 is OBD_MD_FLGRANT in the reply.
1481 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1482 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1483 described in CVE-2007-4573. This problem could allow untrusted
1484 local users to gain root access.
1487 Frequency : when using O_DIRECT and quotas
1489 Description: Incorrect file ownership on O_DIRECT output files
1490 Details : block usage reported by 'lfs quota' does not take into account
1491 files that have been written with O_DIRECT.
1496 Description: touch file failed when fs is not full
1497 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1498 otherwise we can get ENOSP while fs is not full.
1502 Description: parallel lock callbacks
1503 Details : Instead of sending blocking and completion callbacks as separated
1504 requests, adding them to a set and sending in parallel.
1507 Frequency : only for Cray XT3
1508 Bugzilla : 12829/13455
1509 Description: Changing primary group doesn't change the group lustre assigns to
1511 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1512 overridden with the primary group provided by the group upcall,
1513 whereas the supplied fsgid can be trusted if it is in the list of
1514 supplementary groups returned by the group upcall.
1516 Severity : enhancement
1518 Description: Allow masking D_WARNING, D_ERROR messages from console
1519 Details : Console messages can now be disabled via lnet.printk.
1523 Description: User code with malformed file open parameter crashes client node
1524 Details : Before packing join_file req, all the related reference should be
1525 checked carefully in case some malformed flags cause fake
1526 join_file req on client.
1530 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1531 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1532 mode is equal to the granted mode. However, it should also check
1533 that LDLM_FL_CP_REQD is not set.
1537 Description: Heavy nfs access might result in deadlocks
1538 Details : After ELC code landed, it is now improper to enqueue any mds
1539 locks under och_sem, because enqueue might want to decide to
1540 cancel open locks for same inode we are holding och_sem for.
1544 Description: Client eviction while running blogbench
1545 Details : A lot of unlink operations with concurrent I/O can lead to a
1546 deadlock causing evictions. To address the problem, the number of
1547 oustanding OST_DESTROY requests is now throttled to
1548 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1549 callbacks are processed in priority.
1553 Description: enable ACLs on MDS by default
1554 Details : ACLs must be enabled on MDS by default.
1557 Frequency : PPC/PPC64 only
1559 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1560 Details : fix duplicated definitions between asm-ppc64/types.h and
1561 lustre_types.h on PPC.
1564 Frequency : PPC/PPC64 only
1566 Description: asm-ppc/segment.h does not exist
1567 Details : fix compile issue on PPC.
1571 Description: better handle error messages in extents code
1574 Frequency : RHEL4 only
1576 Description: mkfs is very slow on IA64/RHEL4
1577 Details : A performance regression has been discovered in the MPT Fusion
1578 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1579 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1580 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1582 Severity : enhancement
1584 Description: SNMP support enhancement
1585 Details : Adding total number of sampled request for an MDS node in snmp
1588 Severity : enhancement
1590 Description: Optimize ldlm waiting list processing for PR extent locks
1591 Details : When processing waiting list for read extent lock and meeting read
1592 lock that is same or wider to it that is not contended, skip
1593 processing rest of the list and immediatelly return current
1594 status of conflictness, since we are guaranteed there are no
1595 conflicting locks in the rest of the list.
1599 Description: Time out and refuse to reconnect
1600 Details : When the failover node is the primary node, it is possible
1601 to have two identical connections in imp_conn_list. We must
1602 compare not conn's pointers but NIDs, otherwise we can defeat
1603 connection throttling.
1607 Description: port llog fixes from b1_6 into HEAD
1608 Details : Port llog reference couting and some llog cleanups from b1_6
1609 (bug 10800) into HEAD, for protect from panic and access to already
1610 free llog structures.
1614 Description: Detect stride IO mode in read-ahead
1615 Details : When a client does stride read, read-ahead should detect that and
1616 read-ahead pages according to the detected stride pattern.
1620 Description: data checksumming impacts single node performance
1621 Details : add support for several checksum algorithm. Currently, only CRC32
1622 and Adler-32 are supported. The checksum type can be changed on
1623 the fly via /proc/fs/lustre/osc/*/checksum_type.
1627 Description: use adler32 for page checksums
1628 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1633 Description: build for x2 fails
1634 Details : fix compile issue on Cray systems.
1638 Description: Properly match for duplicate locks
1639 Details : Due to different lock order from skiplists code, we need to
1640 traverse entire list for now
1643 Frequency : only on PPC/SLES10
1645 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1646 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1647 be defined. Add a hack in mkfs_lustre.c to work around this compile
1652 Description: LASSERT on MDS when client holding flock lock dies
1653 Details : ldlm pool logic depends on number of granted locks equal to
1654 number of released locks which is not true for flock locks, so
1655 just exclude such locks from consideration.
1659 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1660 Details : Use fsfilt_sync() for both the whole filesystem sync and
1661 individual file sync to eliminate dangerous inode locking
1662 with I_LOCK that can lead to a deadlock.
1666 Description: performance in 1.6.3
1667 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1668 sure that 1MB requests are not fragmented by the block layer.
1670 Severity : enhancement
1672 Description: organize the server-side client stats on per-nid basis
1673 Details : Change the structure of stats under obdfilter and mds to
1681 The "uuid"s file would list the uuids of _active_ exports.
1682 And the clear entry is to clear all stats and stale nids.
1684 Severity : enhancement
1686 Description: eliminate client locks in face of contention
1687 Details : file contention detection and lockless i/o implementation
1688 for contended files.
1692 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1695 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1697 Description: MDS may lose file striping (and hence file data) in some cases
1698 Details : If there are additional extended attributes stored on the MDS,
1699 in particular ACLs, SELinux, or user attributes (if user_xattr
1700 is specified for the client mount options) then there is a risk
1701 of attribute loss. Additionally, the Lustre file striping
1702 needs to be larger than default (e.g. striped over all OSTs),
1703 and an additional attribute must be stored initially in the
1704 inode and then increase in size enough to be moved to the
1705 external attribute block (e.g. ACL growing in size) for file
1710 Description: skiplist implementation simplification
1711 Details : skiplists are used to group compatible locks on granted list
1712 that was implemented as tracking first and last lock of each
1713 lock group the patch changes that to using doubly linked lists
1717 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1718 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1719 because LASSERT on both the data supplied by a client, and the
1720 data on disk is dangerous and incorrect.
1722 Severity : enhancement
1724 Description: Slow truncate/writes to huge files at high offsets.
1725 Details : Directly associate cached pages to lock that protect those pages,
1726 this allows us to quickly find what pages to write and remove
1727 once lock callback is received.
1731 Description: more ldlm soft lockups
1732 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1733 starve other threads from the resource lock for a long time in
1734 case of long waiting queue, so change the debug level from
1735 D_OTHER to the less frequently used D_INFO.
1737 Severity : enhancement
1739 Description: add -gid, -group, -uid, -user options to lfs find
1743 Description: Hung threads in invalidate_inode_pages2_range
1744 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1745 once one is completed. As a result, some RPCs are stuck in the
1746 queue and are never sent.
1750 Description: filter threads hungs on waiting journal commit
1751 Details : Cleanup filter group llog code, then only filter group llog will
1752 be only created in the MDS/OST syncing process.
1756 Description: Procfs and llog threads access destoryed import sometimes.
1757 Details : Sync the import destoryed process with procfs and llog threads by
1758 the import refcount and semaphore.
1760 Severity : enhancement
1762 Description: openlock cache of b1_6 port to HEAD
1767 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1768 Details : If insertion of an extent fails, then discard the inode
1769 preallocation and free data blocks else it can lead to duplicate
1774 Description: don't always update ctime in ext3_xattr_set_handle()
1775 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1776 In some cases the ctime should not be updated, for example for
1777 2.0->1.8 compatibility it is necessary to delete an xattr and it
1778 should not update the ctime.
1782 Bugzilla : 15713/16362
1783 Description: Assertion in iopen_connect_dentry in 1.6.3
1784 Details : looking up an inode via iopen with the wrong generation number can
1785 populate the dcache with a disconneced dentry while the inode
1786 number is in the process of being reallocated. This causes an
1787 assertion failure in iopen since the inode's dentry list contains
1788 both a connected and disconnected dentry.
1792 Description: assertion failure in ldlm_handle2lock()
1793 Details : fix a race between class_handle_unhash() and class_handle2object()
1794 introduced in lustre 1.6.5 by bug 13622.
1799 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1800 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1805 Description: oops in sd_iostats_seq_show()
1806 Details : unloading/reloading the scsi low level driver triggers a kernel
1807 bug when trying to access the sd iostat file.
1812 Description: Kernel panics during QLogic driver reload
1813 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1814 patch, causing memory corruption.
1819 Description: journal_dev option does not work in b1_6
1820 Details : pass mount option during pre-mount.
1822 Severity : enhancement
1824 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1825 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1826 extent information of a file. It can be used to map logical blocks
1827 in a file to physical blocks in the block device.
1831 Description: LDLM soft lockups - improvement
1832 Details : It is be possible to send the lock handle along with each read
1833 or write request because the client is already doing a lock match
1834 itself so there isn't any reason the OST should have to re-do that
1838 Frequency : only on Cray X2
1840 Description: X2 build failures
1841 Details : fix build failures on Cray X2.
1845 Description: xid & resent requests
1846 Details : Initialize RPC XID from clock at startup (randomly if clock is
1849 Severity : enhancement
1851 Description: Add lustre_start utility to start or stop multiple Lustre servers
1856 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1857 Details : In case of memory pressure, list_del() can be called twice on
1858 req->rq_history_list, causing a kernel oops.
1862 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1863 Details : fix stack overflow in the distributed lock manager by defering
1864 export eviction after a failed AST to the elt thread instead of
1865 handling it in the dlm interpret routine.
1869 Description: Convert some comments to new format.
1870 Details : Update documenting comments to match doxygen conventions.
1874 Description: Grammar fixes.
1875 Details : A couple of trivial spelling fixes.
1879 Description: OSD_COUNTERS-mandatory
1880 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1881 to simplify the code.
1885 Description: simplify cmm_device freeing logic.
1886 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1890 Description: Add lockdep support to dt_object_operations locking interface.
1891 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1892 parameter indicating lock ordering. Update mdd code to use new
1897 Description: Introduce failloc constants for lockless IO tests.
1898 Details : Add two new failloc constants to test lockless IO. Only one of
1899 them in implemented---another is checked in yet to be landed
1904 Description: Add lockdep support for inode mutex.
1905 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1906 in the situations where Lustre has to lock more than one inode
1911 Description: Add optional invariants checking support.
1912 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1913 configure switch. This macro is to be used for consistency and
1914 sanity checks that are too expensive to be left in `production'
1919 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1920 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1925 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1926 Details : Use spin_lock_nested() in (the only) situation where more than
1927 one ldlm_lock is locked simultaneously. Also, fix possible
1928 dead-lock in ldlm_lock_change_resource() by enforcing particular
1933 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1934 Details : Instead of specifying each ldlm_lock call-back through separate
1935 parameter, wrap them into struct ldlm_callback_suite.
1939 Description: Kill join_lru obd method and its callers.
1940 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1941 in memory---a requirement implemented through ->o_join_lru() obd
1942 method in HEAD. Remove this method and its users.
1946 Description: Add asynchronous ldlm ENQUEUE completion handler.
1947 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1948 case that stock ldlm_completion_ast() cannot handle as it waits
1949 until lock is granted. Introduce new ldlm_completion_ast_async()
1950 for this. Also comment ldlm_completion_ast().
1954 Description: ldlm_error <-> errno conversion.
1955 Details : Add functions to map (rather arbitrary) between LDLM error codes
1956 and standard errno values. CLIO needs this to prevent LDLM specific
1957 constants from escaping ldlm and osc.
1961 Description: Kill unused ldlm_handle2lock_ns() function.
1962 Details : Kill unused ldlm_handle2lock_ns() function.
1966 Description: Add lu_ref support to ldlm_lock
1967 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1968 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1969 and ldlm_resource. LDLM interface has to be changed, because code that
1970 releases a reference on a lock, has to "know" what reference this is.
1971 In the most frequent case
1973 lock = ldlm_handle2lock(handle);
1975 LDLM_LOCK_PUT(lock);
1977 no changes are required. When any other reference (received _not_ from
1978 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1979 instead of LDLM_LOCK_PUT().
1981 Arguably, changes are pervasive, and interface requires some discipline
1982 for proper use. On the other hand, it was very instrumental in finding
1983 a few leaked lock references.
1987 Description: Add ldlm_lock_addref_try().
1988 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1989 attempts to addref a lock that might be being canceled
1994 Description: Add ldlm_weigh_callback().
1995 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1996 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1997 hard-coded `number of pages' logic.
2001 Description: Add lockdep annotations to llog code.
2002 Details : Use appropriately tagged _nested() locking calls in the places
2003 where llog takes more than one ->lgh_lock lock.
2007 Description: Add loi_kms_set().
2008 Details : Wrap kms updates into a helper function.
2012 Description: Constify instances of struct lsm_operations.
2013 Details : Constify instances of struct lsm_operations.
2017 Description: lu_conf support.
2018 Details : On a server, a file system object is uniquely identified
2019 by a fid, which is sufficient to locate and load all object
2020 state (inode). On a client, on the other hand, more data are
2021 necessary instantiate an object. Change lu_object_find() and
2022 friends to take additional `lu_conf' argument describing object.
2023 Typically this includes layout information.
2027 Description: lu_context fixes.
2028 Details : Introduce new lu_context functions that are needed on the client
2029 side, where some system threads (ptlrpcd) are shared by multiple
2030 modules, and so cannot be stopped during module shutdown.
2034 Description: Add start and stop methods to lu_device_type_operations.
2035 Details : Introduce two new methods in lu_device_type_operations, that are
2036 invoked when first instance of a given type is created and last one
2037 is destroyed respectively. This is need by CLIO.
2041 Description: Add lu_ref support to struct lu_device.
2042 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2043 track leaked references.
2047 Description: Introduce lu_kmem_descr.
2048 Details : lu_kmem_descr and its companion interface allow to create
2049 and destroy a number of kmem caches at once.
2053 Description: Fix lu_object finalization race.
2054 Details : Fix a race between lu_object_find() finding an object and its
2055 concurrent finalization. This race is (most likely) not possible
2056 on the server, but might happen on the client.
2060 Description: Introduce lu_ref interface.
2061 Details : lu_ref is a debugging module allowing to track references to
2062 a given object. It is quite cpu expensive, and has to be
2063 explicitly enabled with --enable-lu_ref. See usage description
2068 Description: Factor lu_site procfs stats into a separate function.
2069 Details : Separate lu_site stats printing code into a separate function
2070 that can be reused on a client.
2074 Description: Constify instances of struct {lu,dt,md}_device_operations.
2075 Details : Constify instances of struct {lu,dt,md}_device_operations.
2079 Description: Introduce struct md_site and move meta-data specific parts of
2080 struct lu_site here.
2081 Details : Move md-specific fields out of struct lu_site into special struct
2082 md_site, so that lu_site can be used on a client.
2086 Description: Kill mdd_lov_destroy().
2087 Details : Remove unused mdd code.
2091 Description: Add st_block checking to multistat.c.
2092 Details : Add st_block checking to multistat.c.
2096 Description: Add lu_ref support to struct obd_device.
2097 Details : Add lu_ref tracking to obd_device.
2101 Description: Kill obd_set_fail_loc().
2102 Details : Remove unused code.
2106 Description: Add special type for ptlrpc_request interpret functions.
2107 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2108 there. Actual usage will be in CLIO.
2112 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2113 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2114 doesn't work with lockdep.
2118 Description: Add rwv.c test program.
2119 Details : New testing program exercising readv(2) and writev(2) (Qian).
2123 Description: Add sendfile.c test program.
2124 Details : New testing program exercising sendfile(2) (Jay).
2128 Description: Ratelimit a message that can be very frequent.
2129 Details : Ratelimit a memory allocation failure message that can
2134 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2135 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2139 Description: Liblustre build fixes.
2140 Details : Liblustre build fixes.
2144 Description: libcfs: add cfs_{need,cond}_resched() interface.
2145 Details : libcfs: add cfs_{need,cond}_resched() definition and
2146 implementations for Linux, NT, and liblustre.
2148 Severity : enhancement
2150 Description: More exported tunables for mballoc
2151 Details : Add support for tunable preallocation window and new tunables for
2152 large/small requests
2156 Description: Detect corruption of block bitmap and checking for preallocations
2157 Details : Checks validity of on-disk block bitmap. Also it does better
2158 checking of number of applied preallocations. When corruption is
2159 found, it turns filesystem readonly to prevent further corruptions.
2163 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2164 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2165 window to 0. If it does hit cache miss, and read-pattern is still
2166 stride-io mode, does not reset the stride window, but also does not
2167 increase the stride window length in this case.
2171 Frequency : only for big-endian servers
2172 Description: Check if system is big-endian while mounting fs with extents feature
2173 Details : Mounting a filesystem with extents feature will fail on big-endian
2174 systems since ext3-based ldiskfs is not supported on big-endian
2175 systems. This can be over-riden with "bigendian_extents" mount option.
2177 Severity : enhancement
2179 Description: The root squash functionality
2180 Details : A security feature, which is to prevent users from being able
2181 to mount lustre on their desktop, run as root, and delete
2182 all of the files in the filesystem. The goal is accomplished by
2183 remapping user id (UID) and group id (GID) of the root user to
2184 a UID and GID specified by the system administartor via Lustre
2185 configuration management server (MGS). The functionality also
2186 allows to specify sets of clients for which the remapping does
2191 Description: Excessive recovery window
2192 Details : With AT enabled, the recovery window can be excessively long (6000+
2193 seconds). To address this problem, we no longer use
2194 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2195 timeout no longer depends on the service time, it is set to
2196 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2197 time via pb_service_time.
2201 Description: Watchdog triggered on MDS failover
2202 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2203 the OSTs know that the MDS "UUID" can be reused for the same export
2204 from a different NID, so we do not need to wait for the export to be
2208 Frequency : rare, only if using MMP with Linux RAID
2210 Description: MMP doesn't work with Linux RAID
2211 Details : While using HA for Lustre servers with Linux RAID, it is possible
2212 that MMP will not detect multiple mounts. To make this work we
2213 need to unplug the device queue in RAID when the MMP block is being
2214 written. Also while reading the MMP block, we should read it from
2215 disk and not the cached one.
2217 Severity : enhancement
2219 Description: open file using fid
2220 Details : A file can be opened using just its fid, like
2221 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2224 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2227 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2228 Details : add proper locking for creation of jbd_4k slab cache
2232 Description: MMP check in ext3_remount() fails without displaying any error
2233 Details : When multiple mount protection fails during remount, proper error
2236 Severity : enhancement
2238 Description: Allow stripe size to be up to 4G-64k
2239 Details : Fix math logic to allow large stripe sizes.
2243 Description: add check for >8TB ldiskfs filesystems
2244 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2245 Don't allow >8TB ldiskfs filesystems to be mounted without
2246 force_over_8tb mount option
2251 Descriptoin: OSS complains "too many blocks" and crashes
2252 Details : In filter_commitrw_write():
2253 1. lquota_chkquota();
2254 2. obdo_from_inode();
2255 3. lquota_pending_commit();
2256 "2." has changed uid and gid in this case, then caused
2257 lquota_chkquota() and lquota_pending_commit() to use
2258 different uid and gid. That is the root of the bug.
2262 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2263 Details : Detect and print out the 'stripe count = -1' case properly.
2265 --------------------------------------------------------------------------------
2267 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2269 * Support for kernels:
2270 2.6.5-7.283 (SLES 9),
2271 2.6.9-55.EL (RHEL 4),
2272 2.6.16.46-0.14 (SLES 10),
2273 2.6.18.8 vanilla (kernel.org)
2274 * Client support for unpatched kernels:
2275 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2276 2.6.16 - 2.6.22 vanilla (kernel.org)
2277 * Due to recently discovered recovery problems, we do not recommend
2278 using patchless RHEL 4 clients with this or any earlier release.
2279 * Recommended e2fsprogs version: 1.39.cfs8
2280 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2281 * Starting with this release, the ldiskfs backing filesystem required
2282 by Lustre is now in its own package, lustre-ldiskfs. This package
2283 should be installed. It is versioned separately from Lustre and
2284 may be released separately in future.
2288 Description: block reactivating mgc import until all deactivates complete
2289 Details : Fix race when failing back MDT/MGS to itself (testing)
2292 Frequency : at statup only
2294 Description: mds_lov_synchronize race leads to various problems
2295 Details : simultaneous MDT->OST connections at startup can cause the
2296 sync to abort, leaving the OSC in a bad state.
2298 Severity : enhancement
2300 Description: add optional extra BUILD_VERSION info
2301 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2302 to override the lustre version.
2305 Frequency : 2.6.18 servers only
2307 Description: ll_kern_mount() doesn't release the module reference
2308 Details : The ldiskfs module reference count never drops down to 0
2309 because ll_kern_mount() doesn't release the module reference.
2314 Description: server LBUG when using old ost_num_threads parameter
2315 Details : Accept the old ost_num_threads parameter but warn that it
2316 is deprecated, and fix an off-by-one error that caused an LBUG.
2321 Description: Transient SCSI error results in persistent IO issue
2322 Details : iobuf->dr_error is not reinitialized to 0 between two
2326 Frequency : sometimes when underlying device returns I/O errors
2328 Description: OSTs not going read-only during write failures
2329 Details : OSTs are not remounted read-only when the journal commit threads
2330 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2331 instead of the ext3 wrappers.
2335 Description: poor connect scaling with increasing client count
2336 Details : Don't run filter_grant_sanity_check for more than 100 exports
2337 to improve scaling for large numbers of clients.
2340 Frequency : SLES10 only
2342 Description: sanity-quota.sh quotacheck failed: rc = -22
2343 Details : Quotas cannot be enabled on SLES10.
2346 Frequency : liblustre clients only
2348 Description: getdirentries does not give error when run on compute nodes
2349 Details : getdirentries does not fail when the size specified as an argument
2350 is too small to contain at least one entry
2352 Severity : enhancement
2354 Description: Add LNET router traceability for debug purposes
2355 Details : If a checksum failure occurs with a router as part of the
2356 IO path, the NID of the last router that forwarded the bulk data
2357 is printed so it can be identified.
2362 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2363 Details : Due to a race condition, liblustre clients were occasionally
2364 evicted incorrectly.
2366 Severity : enhancement
2368 Description: lfs setstripe use optional parameters instead of postional
2371 Severity : enhancement
2373 Description: Nanosecond timestamp support for ldiskfs
2374 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2375 resolution timestamps. There is not yet support for this at
2376 the Lustre filesystem level.
2379 Frequency : during server recovery
2381 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2382 Details : request with rq_no_resend flag not awake l_wait_event if they get
2386 Frequency : nfs export on patchless client
2388 Description: connectathon hang when test nfs export over patchless client
2389 Details : Disconnected dentry cannot be found with lookup, so we do not need
2390 to unhash it or make it invalid
2393 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2394 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2395 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2400 Description: open req refcounting wrong on reconnect
2401 Details : If reconnect happened between getting open reply from server and
2402 call to mdc_set_replay_data in ll_file_open, we will schedule
2403 replay for unreferenced request that we are about to free.
2404 Subsequent close will crash in variety of ways.
2405 Check that request is still eligible for replay in
2406 mdc_set_replay_data().
2411 Description: disable writes to filesystem when reading health_check file
2412 Details : the default for reading the health_check proc file has changed
2413 to NOT do a journal transaction and write to disk, because this
2414 can cause reads of the /proc file to hang and block HA state
2415 checking on a healthy but otherwise heavily loaded system. It
2416 is possible to return to the previous behaviour during configure
2417 with --enable-health-write.
2419 Severity : enhancement
2421 Description: 64-bit inode version
2422 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2423 to the inode. This will be required for version-based recovery.
2428 Description: MDS fails to start if a duplicate client export is detected
2429 Details : in some rare cases it was possible for a client to connect to
2430 an MDS multiple times. Upon recovery the MDS would detect this
2431 and fail during startup. Handle this more gracefully.
2433 Severity : enhancement
2435 Description: Add -o localflock option to simulate old noflock
2437 Details : This will achieve local-only flock/fcntl locks
2443 Description: log_commit_thread vs filter_destroy race leads to crash
2444 Details : Take import reference before releasing llog record semaphore
2449 Description: Wrong request locking in request set processing
2450 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2451 imp_delayed_list, in this place should be used imp_lock.
2454 Frequency : when reconnection
2456 Description: Grant Leak when osc reconnect to OST
2457 Details : When osc reconnect ost, OST(filter) should check whether it
2458 should grant more space to client by comparing fed_grant and
2459 cl_avail_grant, and return the granted space to client instead
2460 of "new granted" space, because client will call osc_init_grant
2461 to update the client grant space info.
2464 Frequency : when client reconnect to OST
2466 Description: Grant Leak when osc do resend and replay bulk write
2467 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2468 bulk write request, because the grant info while be sync between
2469 OSC and OST when reconnect, and we should ignore the grant info
2470 these of resend/replay write req.
2475 Description: Grant space more than avaiable left space sometimes.
2476 Details : When then OST is about to be full, if two bulk writing from
2477 different clients came to OST. Accord the avaliable space of the
2478 OST, the first req should be permitted, and the second one
2479 should be denied by ENOSPC. But if the seconde arrived before
2480 the first one is commited. The OST might wrongly permit second
2481 writing, which will cause grant space > avaiable space.
2484 Frequency : when client is evicted
2486 Description: Grant might be wrongly erased when osc is evicted by OST
2487 Details : when the import is evicted by server, it will fork another
2488 thread ptlrpc_invalidate_import_thread to invalidate the
2489 import, where the grant will be set to 0. While the original
2490 thread will update the grant it got when connecting. So if
2491 the former happened latter, the grant will be wrongly errased
2492 because of this race.
2497 Description: Checking Stale with correct fid
2498 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2499 to check whether it is stale, because sometimes, we want the
2500 enqueue happened anyway, and op_data.fid2 will not be initialized.
2502 Severity : enhancement
2504 Description: update patchless client
2505 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2508 Frequency : only with 2.4 kernel
2510 Description: random memory corruption
2511 Details : size of struct ll_inode_info is to big for union inode.u and this
2512 can be cause of random memory corruption.
2517 Description: Memory leak in recovery
2518 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2519 It should also check obd_fail before fsfilt_start, otherwise if
2520 fsfilt_start return -EROFS,(failover mds during mds recovery).
2521 then the req will return with repmsg->transno = 0 and rc = EROFS.
2522 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2523 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2524 be freed no matter whether fsfilt_commit success or not.
2527 Frequency : only with huge count clients
2529 Description: Prevents from taking the superblock lock in llap_from_page for
2531 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2532 for llap_from_page call in ll_removepage prevents from taking the
2533 superblock lock for a soon died page.
2538 Description: Not check open intent error before release open handle
2539 Details : in some rare cases, the open intent error is not checked before
2540 release open handle, which may cause
2541 ASSERTION(open_req->rq_transno != 0), because it tries to release
2542 the failed open handle.
2547 Description: Set cat log bitmap only after create log success.
2548 Details : in some rare cases, the cat log bitmap is set too early. and it
2549 should be set only after create log success.
2553 Description: Accessing a block bevice can re-enable I/O when Lustre is
2554 tearing down a device.
2555 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2560 Description: service threads may hog cpus when there are a lot of requests
2561 Details : Insert cond_resched to give other threads a chance to use some CPU
2566 Description: the cat log was not initialized in recovery
2567 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2568 unlink log on mds will not be initialized until mds post
2569 recovery. And also in mds post recovery, the unlink log will
2570 initialization will be done asynchronausly, so there will be race
2571 between add unlink log and unlink log initialization.
2575 Description: brw_stats were being printed incorrectly
2576 Details : brw_stats were being printed as log2 but all of them were not
2577 recorded as log2. Also remove some code duplication arising from
2578 filter_tally_{read,write}.
2582 Frequency : rare, only in recovery.
2583 Description: ASSERTION(req->rq_type != LI_POISON) failed
2584 Details : imp_lock should be held while iterating over imp_sending_list for
2585 prevent destroy request after get timeout in ptlrpc_queue_wait.
2589 Description: replay-single.sh test 52 fails
2590 Details : A lock's skiplist need to be cleanup when it being unlinked
2591 from its resource list.
2595 Description: Short directio read returns full requested size rather than
2597 Details : Direct I/O operations should return actual amount of bytes
2598 transferred rather than requested size.
2600 Severity : enhancement
2602 Description: metadata RPC reduction (e.g. for rm performance)
2603 Details : decrease the amount of synchronous RPC between clients and servers
2604 by canceling conflicing lock before the operation on the client
2605 and packing thier handles into the main operation RPC to server.
2607 Severity : enhancement
2609 Description: Async OSC create to avoid the blocking unnecessarily.
2610 Details : If a OST has no remain object, system will block on the creating
2611 when need to create a new object on this OST. Now, ways use
2612 pre-created objects when available, instead of blocking on an
2613 empty osc while others are not empty. If we must block, we block
2614 for the shortest possible period of time.
2618 Description: improve handling recoverable errors
2619 Details : if request processig with error which can be recoverable on server
2620 request should be resend, otherwise page released from cache and
2623 Severity : enhancement
2625 Description: refine locking for avoid write wrong info into lov_objid file
2626 Details : fix possible races with add new target and write/update data in
2629 --------------------------------------------------------------------------------
2631 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2636 Frequency : on some architectures
2638 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2639 Details : Uninitialized flags sometimes cause configuration commands to
2643 Frequency : patchless clients only
2645 Description: missing __iget() symbol export
2646 Details : The __iget() symbol export is missing. To avoid the need for
2647 this on patchless clients the deathrow inode reaper is turned
2648 off, and we depend on the VM to clean up old inodes. This
2649 dependency was during via the fix for bug 12181.
2651 --------------------------------------------------------------------------------
2653 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2655 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2656 INTEROPERATE with older versions automatically. In many cases a
2657 special upgrade step is needed. Please read the
2658 user documentation before upgrading any part of a live system.
2659 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2660 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2661 * WARNING: Lustre configuration and startup changes are required with
2662 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2666 Severity : enhancement
2668 Description: MountConf
2669 Details : Lustre configuration is now managed via mkfs and mount
2670 commands instead of lmc and lconf. New obd types (MGS, MGC)
2671 are added for dynamic configuration management. See
2672 https://mail.clusterfs.com/wikis/lustre/MountConf for
2675 Severity : enhancement
2677 Description: dynamic OST addition
2678 Details : OSTs can now be added to a live filesystem
2680 Severity : enhancement
2682 Description: startup order invariance
2683 Details : MDTs and OSTs can be started in any order. Clients only
2684 require the MDT to complete startup.
2686 Severity : enhancement
2688 Description: parallel, asynchronous orphan cleanup
2689 Details : orphan cleanup is now performed in separate threads for each
2690 OST, allowing parallel non-blocking operation.
2692 Severity : enhancement
2694 Description: optimized stripe assignment
2695 Details : stripe assignments are now made based on ost space available,
2696 ost previous usage, and OSS previous usage, in order to try
2697 to optimize storage space and networking resources.
2699 Severity : enhancement
2701 Description: Permanently set tunables
2702 Details : All writable /proc/fs/lustre tunables can now be permanently
2703 set on a per-server basis, at mkfs time or on a live system.
2705 Severity : enhancement
2707 Description: Lustre message v2
2708 Details : Add lustre message format v2.
2710 Severity : enhancement
2712 Description: client OST exclusion list
2713 Details : Clients can be started with a list of OSTs that should be
2714 declared "inactive" for known non-responsive OSTs.
2718 Description: SPEC SFS validation failure on NFS v2 over lustre.
2719 Details : Changes the blocksize for regular files to be 2x RPC size,
2720 and not depend on stripe size.
2722 Severity : enhancement
2724 Description: Multiple MD RPCs in flight.
2725 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2726 To avoid overly-overloading MDS, introduce a limit on number of
2727 MDS RPCs in flight for a single client and add /proc controls
2728 to adjust this limit.
2730 Severity : enhancement
2732 Description: client read/write statistics
2733 Details : Add client read/write call usage stats for performance
2734 analysis of user processes.
2735 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2736 file access. extents_stats shows chunk size distribution.
2737 extents_stats_per_process show chunk size distribution per
2740 Severity : enhancement
2742 Description: mds statistics
2743 Details : Add detailed mds operations statistics in
2744 /proc/fs/lustre/mds/*/stats.
2748 Description: Failure of copying files with lustre special EAs.
2749 Details : Client side always return success for setxattr call for lustre
2750 special xattr (currently only "trusted.lov").
2755 Description: Refcount LNET uuids
2756 Details : The global LNET uuid list grew linearly with every startup;
2757 refcount repeated list entries instead of always adding to
2760 Severity : enhancement
2762 Description: Dynamic service threads
2763 Details : Within a small range, start extra service threads
2764 automatically when the request queue builds up.
2767 Frequency : mixed-endian client/server environments
2769 Description: mixed-endian crashes
2770 Details : The new msg_v2 system had some failures in mixed-endian
2773 Severity : enhancement
2775 Description: Easy OST removal
2776 Details : OSTs can be permanently deactivated with e.g. 'lctl
2777 conf_param lustre-OST0001.osc.active=0'
2779 Severity : enhancement
2781 Description: MGS proc entries
2782 Details : Added basic proc entries for the MGS showing what filesystems
2785 Severity : enhancement
2787 Description: provide MGS failover
2788 Details : Added config lock reacquisition after MGS server failover.
2790 Severity : enhancement
2792 Description: add Linux 2.4 support
2793 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2797 Description: a large application tries to do I/O to the same resource and dies
2798 in the middle of it.
2799 Details : Check the req->rq_arrival time after the call to
2800 ost_brw_lock_get(), but before we do anything about
2801 processing it & sending the BULK transfer request. This
2802 should help move old stale pending locks off the queue as
2803 quickly as obd_timeout.
2806 Frequency : when an incorrect nid is specified during startup
2808 Description: ptlrpc connect to non-existant node causes kernel crash
2809 Details : LNET can't be re-entered from an event callback, which
2810 happened when we expire a message after the export has been
2811 cleaned up. Instead, hand the zombie cleanup off to another
2814 Severity : enhancement
2816 Description: plain/inodebits lock performance improvement
2817 Details : Grouping plain/inodebits in granted list by their request modes
2818 and bits policy, thus improving the performance of search through
2822 Frequency : only if OST filesystem is corrupted
2824 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2825 Details : for a short time RPCs with bulk IO are in the replay list,
2826 but replay of bulk IOs is unimplemented. If the OST filesystem
2827 is corrupted due to disk cache incoherency and then replay is
2828 started it is possible to trip an assertion. Avoid putting
2829 committed RPCs into the replay list at all to avoid this issue.
2832 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2834 Description: System hang on startup
2835 Details : This bug allowed the liblustre (e.g. catamount) client to
2836 return to the app before handling all startup RPCs. This
2837 could leave the node unresponsive to lustre network traffic
2838 and manifested as a server ptllnd timeout.
2840 Severity : enhancement
2842 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2843 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2844 boolean to control whether to print peer debug info when a
2845 client's RPC times out.
2848 Frequency : only for kernels with patches from Lustre below 1.4.3
2850 Description: Remove old rdonly API
2851 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2854 Frequency : only for devices with external journals
2856 Description: Set external device read-only also
2857 Details : During a commanded failover stop, we set the disk device
2858 read-only while the server shuts down. We now also set any
2859 external journal device read-only at the same time.
2862 Frequency : when upgrading from 1.4 while trying to change parameters
2864 Description: The wrong (new) MDC name was used when setting parameters for
2865 upgraded MDT's. Also allows changing of OSC (and MDC)
2866 parameters if --writeconf is specified at tunefs upgrade time.
2869 Frequency : when setting specific ost indicies
2871 Description: QOS code breaks on skipped indicies
2872 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2873 created with --index need not be sequential.
2875 Severity : enhancement
2877 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2878 Details : The uninit_groups feature works in conjunction with the kernel
2879 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2880 up the pass1 processing of e2fsck. This is a read-only feature
2881 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2882 mount filesystems that have had this feature enabled.
2884 Severity : enhancement
2886 Description: Improve multi-block allocation algorithm to avoid fragmentation
2887 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2888 allocation locality and avoid filesystem fragmentation.
2890 ------------------------------------------------------------------------------
2892 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2894 * Support for kernels:
2895 2.4.21-47.0.1.EL (RHEL 3)
2896 2.6.5-7.283 (SLES 9)
2897 2.6.9-42.0.10.EL (RHEL 4)
2898 2.6.12.6 vanilla (kernel.org)
2899 2.6.16.27-0.9 (SLES 10)
2900 * Recommended e2fsprogs version: 1.39.cfs5
2902 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2906 Frequency : occasional, depends on client load and configuration
2907 Bugzilla : 12181, 12203
2908 Description: data loss for recently-modified files
2910 Details : In some cases it is possible that recently written or created
2911 files may not be written to disk in a timely manner (this should
2912 normally be within 30s unless client IO load is very high).
2913 The problem appears as zero-length files or files that are a
2914 multiple of 1MB in size after a client crash or client eviction
2915 that are missing data at the end of the file.
2917 This problem is more likely to be hit on clients where files are
2918 repeatedly created and unlinked in the same directory, clients
2919 have a large amount of RAM, have many CPUs, the filesystem has
2920 many OSTs, the clients are rebooted frequently, and/or the files
2921 are not accessed by other nodes after being written.
2923 The presence of the problem can be detected by looking at
2924 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2925 smaller than the second (nr_unused) then dirty files will not
2926 be flushed automatically to disk. "sync; sleep 10" should be
2927 run several times on the node before unmounting it to update
2928 Lustre (this is also safe to run on nodes without this problem).
2930 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2931 that can cause this same problem, so customers using that kernel
2932 also need to update the kernel in addition to Lustre. In order
2933 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2935 It is normal that files written just before a client crash (less
2936 than 30s) may not yet have been flushed to disk, even for local
2940 Frequency : frequent on thin XT3 nodes
2942 Description: UUID collision on thin XT3 Linux nodes
2943 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2944 randomly, since we relied on an insufficiently-seeded PRNG.
2949 Description: OSS hangs after "All ost request buffers busy"
2950 Details : A deadlock between quota and journal operations caused OSS
2951 hangs after printing "All ost request buffers busy."
2954 Frequency : always on liblustre builds
2956 Description: Cleanup compiler warnings on liblustre
2959 Frequency : always on liblustre builds on XT3
2961 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2965 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2966 > 32000 subdirectories
2967 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2968 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2969 subdirectory count crosses 32000. This will aid e2fsck to
2970 correctly handle more than 32000 subdirectories.
2973 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2975 Description: System hang on startup
2976 Details : This bug allowed the liblustre (e.g. catamount) client to
2977 return to the app before handling all startup RPCs. This
2978 could leave the node unresponsive to lustre network traffic
2979 and manifested as a server ptllnd timeout.
2981 Severity : enhancement
2983 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2984 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2985 boolean to control whether to print peer debug info when a
2986 client's RPC times out.
2991 Description: make O_SYNC working on 2.6 kernels
2992 Details : 2.6 kernels use different method for mark pages for write,
2993 so need add a code to lustre for O_SYNC work.
2998 Description: Failure to close file and release space on NFS
2999 Details : Put inode details into lock acquired in ll_intent_file_open.
3000 Use mdc_intent_lock in ll_intent_open to properly
3001 detect all kind of errors unhandled by mdc_enqueue.
3006 Description: proc file read during shutdown sometimes raced obd removal,
3008 Details : Add lock to prevent obd access after proc file removal.
3011 Frequency : Only for files larger than 4GB on 32-bit clients.
3013 Description: improperly doing page alignment of locks
3014 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3015 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3020 Description: RPCs being resent when they shouldn't be
3021 Details : Some RPCs that should not be resent are being resent. This
3022 can cause inconsistencies in the RPC state machine. Do not
3023 resend such requests.
3026 Frequency : rare, only with NFS export
3028 Description: Crash on NFS re-export node
3029 Details : under very unusual load conditions an assertion is hit in
3030 ll_intent_file_open()
3033 Frequency : only if OST filesystem is corrupted
3035 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3036 Details : for a short time RPCs with bulk IO are in the replay list,
3037 but replay of bulk IOs is unimplemented. If the OST filesystem
3038 is corrupted due to disk cache incoherency and then replay is
3039 started it is possible to trip an assertion. Avoid putting
3040 committed RPCs into the replay list at all to avoid this issue.
3045 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3046 Details : Large single O_DIRECT read and write calls can fail to allocate
3047 a sufficiently large buffer to process the request. In case of
3048 allocation failure the allocation is retried with a smaller
3049 buffer and broken into smaller requests.
3051 Severity : enhancement
3053 Description: Add -o localflock option to simulate old noflock behaviour.
3054 Details : This will achieve local-only flock/fcntl locks coherentness.
3059 Description: versioning check is incomplete
3060 Details : Checking the version difference of client vs. server, report
3061 error if the gap is too big.
3066 Description: add support PG_writeback bit
3067 Details : add support for PG_writeback bit for Lustre, for more carefull
3068 work with page cache in 2.6 kernel. This also fix some deadlocks
3069 and remove hack for work O_SYNC with 2.6 kernel.
3071 Severity : enhancement
3073 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3074 Details : The uninit_groups feature works in conjunction with the kernel
3075 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3076 up the pass1 processing of e2fsck. This is a read-only feature
3077 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3078 mount filesystems that have had this feature enabled.
3080 Severity : enhancement
3082 Description: Improve multi-block allocation algorithm to avoid fragmentation
3083 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3084 allocation locality and avoid filesystem fragmentation.
3086 ------------------------------------------------------------------------------
3088 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3090 * Support for kernels:
3091 2.6.9-42.0.3.EL (RHEL 4)
3092 2.6.5-7.276 (SLES 9)
3093 2.4.21-47.0.1.EL (RHEL 3)
3094 2.6.12.6 vanilla (kernel.org)
3095 2.6.16.21-0.8 (SLES10)
3096 * Recommended e2fsprogs version: 1.39.cfs2-0
3098 * The backwards-compatible /proc/sys/portals symlink has been removed
3099 in this release. Before upgrading, please ensure that you change
3100 any configuration scripts or /etc/sysctl.conf files that access
3101 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3102 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3103 in advance of the upgrade on any system running Lustre 1.4.6 or
3104 newer, since /proc/sys/lnet was added in that version.
3105 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3109 Frequency : only when quota is used
3111 Description: avoid scanning export list for quota master
3112 Details : Change the algorithms to avoid scanning export list in order
3113 to improve the efficiency.
3116 Frequency : MDS failover only, very rarely
3118 Description: "went back in time" messages on mds failover
3119 Details : The greatest transno may be lost when the current operation
3120 finishes with an error (transno==0) and the client's last_rcvd
3121 record is over-written. Save the greatest transno in the
3122 mds_last_transno for this case.
3125 Frequency : always for specific kernels and striping counts
3127 Description: client may get "Matching packet too big" without ACL support
3128 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3129 when trying to access files in certain configurations. The
3130 clients should in fact be denied when mounting because they do
3131 not understand ACLs.
3134 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3136 Description: many clients connecting with IO in progress causes connect timeouts
3137 Details : Avoid synchronous journal commits to avoid delays caused by many
3138 clients connecting/disconnecting when bulk IO is in progress.
3139 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3140 OST_IO_PORTAL to avoid delays behind potentially many pending
3144 Frequency : occasionally with multiple writers to a single file
3146 Description: shared writes to file may result in wrong size reported by stat()
3147 Details : Allow growing of kms when extent lock is cancelled
3150 Frequency : always with random mmap IO to multi-striped file
3152 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3153 Details : Only if the hole is at the end of OST object so that kms is too
3154 small. Fix is to increase kms accordingly in ll_nopage.
3157 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3159 Description: writes to a missing object would leak memory on the OST
3160 Details : If there is an inconsistency between the MDS and OST filesystems,
3161 such that the MDS references an object that doesn't exist, writes
3162 to that object will leak memory due to incorrect cleanup in the
3163 error handling path, eventually running out of memory on the OST.
3168 Description: Creating too long symlink causes lustre errors
3169 Details : Check symlink and name lengths before sending requests to MDS.
3172 Frequency : only if flock is enabled (not on by default)
3174 Description: posix locks not released on fd closure on 2.6.9+
3175 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3176 kernels, this caused such locks not to be released on fd close and
3177 then assertions on fs unmount about still used locks.
3180 Frequency : MDS failover only, very rarely
3182 Description: clients may get ASSERTION(granted_lock != NULL)
3183 Details : When request was taking a long time, and a client was resending
3184 a getattr by name lock request. The were multiple lock requests
3185 with the same client lock handle and
3186 mds_getattr_name->fixup_handle_for_resent_request found one of the
3187 lock handles but later failed with ASSERTION(granted_lock != NULL).
3192 Description: handle->h_buffer_credits > 0, assertion failure
3193 Details : h_buffer_credits is zero after truncate, causing assertion
3194 failure. This patch extends the transaction or creates a new
3198 Frequency : NFS re-export or patchless client
3199 Bugzilla : 11179, 10796
3200 Description: Crash on NFS re-export node (__d_move)
3201 Details : We do not want to hash the dentry if we don't have a lock.
3202 But if this dentry is later used in d_move, we'd hit uninitialised
3203 list head d_hash, so we just do this to init d_hash field but
3204 leave dentry unhashed.
3207 Frequency : NFS re-export or patchless client
3209 Description: NFS exports has problem with symbolic link
3210 Details : lustre client didn't properly install dentry when re-exported
3211 to NFS or running patchless client.
3214 Frequency : NFS re-export or patchless client
3216 Description: Various nfs/patchless fixes.
3217 Details : fixes reuse disconected alias for lookup process - this fixes
3218 warning "find_exported_dentry: npd != pd",
3219 fix permission error with open files at nfs.
3220 fix apply umask when do revalidate.
3223 Frequency : occasional
3225 Description: Crash on NFS re-export node
3226 Details : calling clear_page() on the wrong pointer triggered oops in
3227 generic_mapping_read().
3230 Frequency : rarely, using O_DIRECT IO
3232 Description: unaligned directio crashes client with LASSERT
3233 Details : check for unaligned buffers before trying any requests.
3236 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3238 Description: stale data returned from RAID cache
3239 Details : If only a small amount of IO is done to the RAID device before
3240 reading it again it is possible to get stale data from the RAID
3241 cache instead of reading it from disk.
3244 Frequency : always for sles10 kernel
3246 Description: sles10 support
3247 Details : ll_follow_link: compile fixes and using of nd_set_link
3248 under newer kernels.
3251 Frequency : depends on arch, kernel and compiler version, always on sles10
3254 Description: recursive or deep enough symlinks cause stack overflow
3255 Details : getting rid of large stack-allocated variable in
3259 Frequency : depends on hardware
3261 Description: lustre write performance loss in the SLES10 kernel
3262 Details : the performance loss is caused by using of write barriers in the
3263 ext3 code. The SLES10 kernel turns barrier support on by
3264 default. The fix is to undo that change for ldiskfs.
3266 ------------------------------------------------------------------------------
3268 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3270 * Support for kernels:
3271 2.6.9-42.0.3EL (RHEL 4)
3272 2.6.5-7.276 (SLES 9)
3273 2.4.21-47.0.1.EL (RHEL 3)
3274 2.6.12.6 vanilla (kernel.org)
3278 Frequency : quota enabled and large files being deleted
3280 Description: releasing more than 4GB of quota at once hangs OST
3281 Details : If a user deletes more than 4GB of files on a single OST it
3282 will cause the OST to spin in an infinite loop. Release
3283 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3288 Description: statfs data retrieved from /proc may be stale or zero
3289 Details : When reading per-device statfs data from /proc, in the
3290 {kbytes,files}_{total,free,avail} files, it may appear
3291 as zero or be out of date.
3294 Frequency : systems with MD RAID1 external journal devices
3296 Description: lconf's call to blkid is confused by RAID1 journal devices
3297 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3298 of returning all block devices that match the journal UUID.
3301 Frequency : always, for aggregate stripe size over 4GB
3303 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3304 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3305 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3308 Frequency : always if "lfs find" used on a local file/directory
3310 Description: "lfs find" segfaults if used on a local file/directory
3311 Details : The case where a directory component was not specified wasn't
3312 handled correctly. Handle this properly.
3315 Frequency : always on ppc64
3317 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3318 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3319 perform endianness conversion.
3322 Frequency : rarely (truncate to non-zero file size after write under load)
3323 Bugzilla : 10730, 10687
3324 Description: Files padded with zeros to next 4K multiple
3325 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3326 it is possible that files that are truncated to a non-zero size
3327 immediately after being written are filled with zero bytes beyond
3328 the truncated size. No file data is lost.
3330 Severity : enhancement
3332 Description: Allow recovery/failover for liblustre clients.
3333 Details : liblustre clients were unaware of failover configurations until
3336 Severity : enhancement
3338 Description: user file locks should fail when not mounting with flock option
3339 Details : Set up an error-returning stub in ll_file_operations.lock field
3340 to prevent incorrect behaviour when client is mounted without
3341 flock option. Also, set up properly f_op->flock field for
3345 Frequency : always on ia64
3347 Description: "lfs df" loops on printing out MDS statfs information
3348 Details : The obd_ioctl_data was not initialized and in some systems
3349 this caused a failure during the ioctl that did not return
3350 an error. Initialize the struct and return an error on failure.
3353 Frequency : SLES 9 only
3355 Description: Error of copying files with lustre special EAs as root
3356 Details : Client side always return success for setxattr call for lustre
3357 special xattr (currently only "trusted.lov").
3360 Frequency : rarely on clusters with both ia64+i386 clients
3362 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3363 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3364 clients are doing concurrent writes to the same file it is
3365 possible that the ia64 clients may LASSERT because the OST
3366 extent locks are not PAGE_SIZE aligned. Ensure that grown
3367 locks are always aligned on the request boundary.
3370 Frequency : specific use, occasional
3372 Description: Overwriting in use executable truncates on-disk binary image
3373 Details : If one node attempts to overwrite an executable in use by
3374 another node, we now correctly return ETXTBSY instead of
3375 truncating the file.
3377 Severity : enhancement
3379 Description: Async OSC create to avoid the blocking unnecessarily.
3380 Details : If a OST has no remain object, system will block on the creating
3381 when need to create a new object on this OST. Now, ways use
3382 pre-created objects when available, instead of blocking on an
3383 empty osc while others are not empty. If we must block, we block
3384 for the shortest possible period of time.
3389 Description: chmod on Lustre root is propagated to other clients
3390 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3391 invalid by the follow_mount time.
3394 Frequency : liblustre clients only
3396 Description: Race in 'instant cancel' lock handling could lead to such locks
3397 never to be granted in case of SMP MDS
3398 Details : Do not destroy not yet granted but cbpending locks in
3402 Frequency : replay/resend of open
3404 Description: non null lock assetion failure in mds_intent_policy
3405 Details : Trying to replay/resend lockless open requests resulted in
3406 mds_open() returning 0 with no lock. Now it sets a flag if
3407 a lock is going to be returned.
3409 Severity : enhancement
3411 Description: Checksum enhancements
3412 Details : New checksum enhancements allow for resending RPCs that failed
3415 Severity : enhancement
3417 Description: Tunables on number of dirty pages in cacche
3418 Details : Allow to set limit on number of dirty pages cached.
3423 Description: client crash on unmount - lock still has references
3424 Details : In some error handling cases it was possible to leak a lock
3425 reference on a client while accessing a file. This was not
3426 harmful to the client during operation, but would cause the
3427 client to crash when the filesystem is unmounted.
3430 Frequency : specific case, rare
3432 Description: ETXTBSY on mds though file not in use by client
3433 Details : ETXTBSY is no longer incorrectly returned when attempting to
3434 chmod or chown a directory that the user previously tried to
3435 execute or a currently-executing binary.
3438 Frequency : extremely rare except on liblustre-based clients
3440 Description: Lustre space not freed when files are deleted
3441 Details : Clean up open-unlinked files after client eviction. Previously
3442 the unlink was skipped and the files remained as orphans.
3447 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3448 Details : In some cases it was possible to send a blocking callback to
3449 a client doing a glimpse, even though that client didn't get
3450 a lock granted. When the glimpse lock is cancelled on the OST
3451 the freed lock is left on the waiting list and corrupted the list.
3454 Frequency : all core dumps
3456 Description: Broke core dumps to lustre
3457 Details : Negative dentry may be unhashed if parent does not have UPDATE
3458 lock, but some callers, e.g. do_coredump, expect dentry to be
3459 hashed after successful create, hash it in ll_create_it.
3461 ------------------------------------------------------------------------------
3463 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3465 * Support for kernels:
3466 2.6.9-42.0.2.EL (RHEL 4)
3467 2.6.5-7.276 (SLES 9)
3468 2.4.21-40.EL (RHEL 3)
3469 2.6.12.6 vanilla (kernel.org)
3473 Frequency : always on RHEL 3
3475 Description: Number of open files grows over time
3476 Details : The number of open files grows over time, whether or not
3477 Lustre is started. This was due to a filp leak introduced
3478 by one of our kernel patches.
3480 ------------------------------------------------------------------------------
3482 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3484 * Support for kernels:
3485 2.6.9-42.EL (RHEL 4)
3486 2.6.5-7.276 (SLES 9)
3487 2.4.21-40.EL (RHEL 3)
3488 2.6.12.6 vanilla (kernel.org)
3493 Bugzilla : 5719, 9635, 9792, 9684
3494 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3495 Details : If a server is under heavy load and cannot reply to new
3496 connection requests before the client resends the (re)connect,
3497 the connection handling code can behave badly if two service
3498 threads are concurrently handing separate (re)connections from
3499 the same client. Add better locking to the connection handling
3500 code, and ensure that only a single connection will be processed
3501 for a given client UUID, even if the lock is dropped.
3503 Severity : enhancement
3505 Description: add TCP zero-copy support to kernel
3506 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3507 sends if the hardware supports scatter-gather and checksumming.
3508 This allows socklnd to do client-write and server-read more
3509 efficiently and reduce CPU utilization from skbuf copying.
3512 Frequency : only if NFS exporting from client
3514 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3515 Details : If setting attributes on a file created under NFS that had
3516 never been opened it would be possible to oops the client
3517 if the file had no objects.
3520 Frequency : always for liblustre
3522 Description: liblustre client does MDS+OSTs setattr RPC for each write
3523 Details : When doing a write from a liblustre client, the client
3524 incorrectly issued an RPC to the MDS and each OST the file was
3525 striped over in order to update the timestamps. When writing
3526 with small chunks and many clients this could overwhelm the MDS
3527 with RPCs. In all cases it would slow down the write because
3528 these RPCs are unnecessary.
3530 Severity : enhancement
3532 Description: allow number of MDS service threads to be changed at module load
3533 Details : It is now possible to change the number of MDS service threads
3534 running. Adding "options mds mds_num_threads={N}" to the MDS's
3535 /etc/modprobe.conf will set the number of threads for the next
3536 time Lustre is restarted (assuming the "mds" module is also
3537 reloaded at that time). The default number of threads will
3538 stay the same, 32 for most systems.
3543 Description: OST crash if filesystem is unformatted or corrupt
3544 Details : If an OST is started on a device that has never been formatted
3545 or if the filesystem is corrupt and cannot even mount then the
3546 error handling cleanup routines would dereference a NULL pointer.
3551 Description: NULL pointer deref in llap_from_page.
3552 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3553 incorrect) mapping if the page was truncated/reclaimed while it was
3554 searched for. Check for this condition and skip such pages when
3555 doing readahead. Introduce extra check to llap_from_page() to
3556 verify page->mapping->host is non-NULL (so page is not anonymous).
3559 Frequency : Sometimes when using sys_sendfile
3561 Description: "page not covered by a lock" warnings from ll_readpage
3562 Details : sendfile called ll_readpage without right page locks present.
3563 Now we introduced ll_file_sendfile that does necessary locking
3564 around call to generic_file_sendfile() much like we do in
3568 Frequency : with certain MDS communication failures at client mount time
3570 Description: NULL pointer deref after failed client mount
3571 Details : a client connection request may delayed by the network layer
3572 and not be sent until after the PTLRPC layer has timed out the
3573 request. If the client fails the mount immediately it will try
3574 to clean up before the network times out the request. Add a
3575 reference from the request import to the obd device and delay
3576 the cleanup until the network drops the request.
3579 Frequency : occasionally during client (re)connect
3581 Description: assertion failure during client (re)connect
3582 Details : processing a client connection request may be delayed by the
3583 client or server longer than the client connect timeout. This
3584 causes the client to resend the connection request. If the
3585 original connection request is replied in this interval, the
3586 client may trip an assertion failure in ptlrpc_connect_interpret()
3587 which thought it would be the only running connect process.
3590 Frequency : only with obd_echo servers and clients that are rebooted
3592 Description: kernel BUG accessing uninitialized data structure
3593 Details : When running an obd_echo server it did not start the ping_evictor
3594 thread, and when a client was evicted an uninitialized data
3595 structure was accessed. Start the ping_evictor in the RPC
3596 service startup instead of the OBD startup.
3598 Severity : enhancement
3599 Bugzilla : 10193 (patchless)
3600 Description: Remove dependency on various unexported kernel interfaces.
3601 Details : No longer need reparent_to_init, exit_mm, exit_files,
3602 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3605 Frequency : rare (only users of deprecated and unsupported LDAP config)
3607 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3608 Details : LDAP apparently contains 'lustreName' attributes instead of
3609 'name'. A simple remapping of the name is sufficient.
3612 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3614 Description: waiting_locks_callback trips kernel BUG if client is evicted
3615 Details : Running with the dump_on_timeout debug flag turned on makes
3616 it possible that the waiting_locks_callback() can try to dump
3617 the Lustre kernel debug logs from an interrupt handler. Defer
3618 this log dumping to the expired_lock_main() thread.
3620 Severity : enhancement
3622 Description: Support NFS exporting on 2.6 kernels.
3623 Details : Implement non-rawops metadata methods for NFS server to use without
3624 changing NFS server code.
3627 Frequency : very rare (synthetic metadata workload only)
3629 Description: two racing renames might cause an MDS thread to deadlock
3630 Details : Running the "racer" program may cause one MDS thread to rename
3631 a file from being the source of a rename to being the target of
3632 a rename at exactly the same time that another thread is doing
3633 so, and the second thread has already enqueued these locks after
3634 doing a lookup of the target and is trying to relock them in
3635 order. Ensure that we don't try to re-lock the same resource.
3638 Frequency : only very large systems with liblustre clients
3640 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3641 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3642 all OSTs in parallel. This allows the eviction of stale liblustre
3643 clients to proceed much faster than if they were done in series,
3644 and also offers similar improvements for other set_info RPCs.
3649 Description: excessive CPU usage during initial read phase on client
3650 Details : During the initial read phase on a client, it would agressively
3651 retry readahead on the file, consuming too much CPU and impacting
3652 performance (since 1.4.5.8). Improve the readahead algorithm
3653 to avoid this, and also improve some other common cases (read
3654 of small files in particular, where "small" is files smaller than
3655 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3660 Description: MDS crash when receiving packet with unknown intent.
3661 Details : Do not LBUG in unknown intent case, just return -EFAULT
3663 Severity : enhancement
3664 Bugzilla : 9293, 9385
3665 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3666 Details : Do not serialize getattr (non-intent version) and statfs.
3669 Frequency : occasional, when OST network is overloaded/intermittent
3671 Description: client evicted by OST after bulk IO timeout
3672 Details : If a client sends a bulk IO request (read or write) the OST
3673 may evict the client if it is unresposive to its data GET/PUT
3674 request. This is incorrect if the network is overloaded (takes
3675 too long to transfer the RPC data) or dropped the OST GET/PUT
3676 request. There is no need to evict the client at all, since
3677 the pinger and/or lock callbacks will handle this, and the
3678 client can restart the bulk request.
3681 Frequency : Always when mmapping file with no objects
3683 Description: client crashes when mmapping file with no objects
3684 Details : Check that we actually have objects in a file before doing any
3685 operations on objects in ll_vm_open, ll_vm_close and
3691 Description: Request leak when working with deleted CWD
3692 Details : Introduce advanced request refcount tracking for requests
3693 referenced from lustre intent.
3695 Severity : Enhancement
3697 Description: Cache open file handles on client.
3698 Details : MDS now will return special lock along with openhandle, if
3699 requested and client is allowed to hold openhandle, even if unused,
3700 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3701 closing files for every read/write openration.
3703 Severity : Enhancement
3705 Description: Cache open negative dentries on client when possible.
3706 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3707 negative dentries on lock revocation.
3712 Description: Remounting a client read-only wasn't possible with a zconf mount
3713 Details : It wasn't possible to remount a client read-only with llmount.
3715 Severity : enhancement
3716 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3717 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3718 MPICH is included to add Lustre-specific ADIO interfaces.
3719 This is based closely on the UFS ADIO layer and only differs
3720 in file creation, in order to allow the OST striping to be set.
3721 This is user-contributed code and not supported by CFS.
3726 Description: extended inode attributes (immutable, append-only) work improperly
3727 when 2.4 and 2.6 kernels are used on client/server or vice versa
3728 Details : Introduce kernel-independent values for these flags.
3730 Severity : enhancement
3733 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3734 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3735 granularity. Now, Fractional MB granularity is be supported,
3736 this is very useful for low memory system.
3738 Severity : enhancement
3740 Description: Getattr by fid
3741 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3742 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3743 have locks and avoids getattr rpc requests for it on every stat.
3746 Frequency : Always, for filesystems larger than 2TB
3748 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3749 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3750 filesystems larger than 2TB to be mounted with mballoc enabled.
3753 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3755 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3756 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3757 filesystems larger than 2TB is mounted then the kernel will
3758 silently corrupt the start of the filesystem. CONFIG_LBD is
3759 enabled for all CFS-supported kernels, but the possibility of
3760 this happening with a modified kernel config exists.
3762 Severity : enhancement
3764 Description: add client O_DIRECT support for 2.6 kernels
3765 Details : It is now possible to do O_DIRECT reads and writes to files
3766 in the Lustre client mountpoint on 2.6 kernel clients.
3768 Severity : enhancement
3770 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3771 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3772 parallel, not waiting for response from every OST before sending
3773 a rpc to the next OST.
3778 Description: setattr vs write race when updating file timestamps
3779 Details : Client processes that update a file timestamp into the past
3780 right after writing to the file (e.g. tar) it is possible that
3781 the updated file modification time can be reset to the current
3782 time due to a race between processing the setattr and write RPC.
3784 Severity : enhancement
3786 Description: Bring 'lfs find' closer in line with regular Linux find.
3787 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3788 -print0 options and obtains all the needed info through the lustre
3791 Severity : enhancement
3793 Description: support up to 1024 configured devices on one node
3794 Details : change obd_dev array from statically allocated to dynamically
3795 allocated structs as they are first used to reduce memory usage
3800 Description: Flush dirty partially truncated pages during truncate
3801 Details : Immediatelly flush partially truncated pages in filter_setattr,
3802 this way we completely avoid having any pages in page cache on OST
3803 and can retire ugly workarounds during writes to flush such pages.
3808 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3809 Details : protect inode from truncation within vfs_unlink() context
3810 just take a reference before calling vfs_unlink() and release it
3811 when parent's i_sem is free.
3816 Description: last_id value checked outside lock on OST caused LASSERT failure
3817 Details : If there were multiple MDS->OST object precreate requests in
3818 flight, it was possible that the OST's last object id was checked
3819 outside a lock and incorrectly tripped an assertion. Move checks
3820 inside locks, and discard old precreate requests.
3823 Frequency : always, if extents are used on OSTs
3825 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3826 Details : OSTs using the extents format would not zero the high 16 bits of
3827 the index physical block number. This is not a problem for any
3828 OST filesystems smaller than 16TB, and no kernels support ext3
3829 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3830 new/modified files) and can be fixed for existing filesystems
3831 with e2fsprogs-1.39-cfs1.
3836 Description: import connection selection may be incorrect if timer wraps
3837 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3838 import connections to be ignored if the 32-bit jiffies counter
3839 wraps. Use a 64-bit jiffies counter.
3842 Frequency : very large clusters immediately after boot
3844 Description: LNET request buffers exhausted under heavy short-term load
3845 Details : If a large number of client requests are generated on a service
3846 that has previously never seen so many requests it is possible
3847 that the request buffer growth cannot keep up with the spike in
3848 demand. Instead of dropping incoming requests, they are held in
3849 the LND until the RPC service can accept more requests.
3852 Frequency : Sometimes during replay
3854 Description: Assertion failure in ll_local_open after replay.
3855 Details : If replay happened on an open request reply before we were able
3856 to set replay handler, reply will become not swabbed tripping the
3857 assertion in ll_local_open. Now we set the handler right after
3858 recognising of open request
3861 Frequency : very rare
3863 Description: kernel reports "badness in vsnprintf"
3864 Details : Reading from the "recovery_status" /proc file in small chunks
3865 may cause a negative length in lprocfs_obd_rd_recovery_status()
3866 call to vsnprintf() (which is otherwise harmless). Exit early
3867 if there is no more space in the output buffer.
3869 Severity : enhancement
3871 Description: clear OBD RPC statistics by writing to them
3872 Details : It is now possible to clear the OBD RPC statistics by writing
3873 to the "stats" file.
3878 Description: Client mtime is not the same on different clients after utimes
3879 Details : In some cases, the client was using the utimes() syscall on
3880 a file cached on another node. The clients now validate the
3881 ctime from the MDS + OSTs to determine which one is right.
3886 Description: Inability to activate failout mode
3887 Details : lconf script incorrectly assumed that in pythong string's numeric
3888 value is used in comparisons.
3891 Frequency : always with multiple stripes per file
3893 Description: Inefficient object allocation for mutli-stripe files
3894 Details : When selecting which OSTs to stripe files over, for files with
3895 a stripe count that divides evenly into the number of OSTs,
3896 the MDS is always picking the same starting OST for each file.
3897 Return the OST selection heuristic to the original design.
3902 Description: mount failures may take full timeout to return an error
3903 Details : Under some heavy load conditions it is possible that a
3904 failed mount can wait for the full obd_timeout interval,
3905 possibly several minutes, before reporting an error.
3906 Instead return an error as soon as the status is known.
3908 Frequency : quota enabled and large files being deleted
3910 Description: releasing more than 4GB of quota at once hangs OST
3911 Details : If a user deletes more than 4GB of files on a single OST it
3912 will cause the OST to spin in an infinite loop. Release
3913 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3918 Description: statfs data retrieved from /proc may be stale or zero
3919 Details : When reading per-device statfs data from /proc, in the
3920 {kbytes,files}_{total,free,avail} files, it may appear
3921 as zero or be out of date.
3924 Frequency : systems with MD RAID1 external journal devices
3926 Description: lconf's call to blkid is confused by RAID1 journal devices
3927 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3928 of returning all block devices that match the journal UUID.
3931 Frequency : always, for aggregate stripe size over 4GB
3933 Description: assertion fails when trying to use 4GB stripe size
3934 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3935 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3938 Frequency : always on ppc64
3940 Description: the first write on an ext3 filesystem with mballoc got stuck
3941 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3942 perform endianness conversion.
3944 ------------------------------------------------------------------------------
3946 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3948 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3949 INTEROPERATE with older versions automatically. Please read the
3950 user documentation before upgrading any part of a live system.
3951 * WARNING: Lustre networking configuration changes are required with
3952 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3955 * Support for kernels:
3956 2.6.9-22.0.2.EL (RHEL 4)
3957 2.6.5-7.244 (SLES 9)
3958 2.6.12.6 vanilla (kernel.org)
3961 Severity : enhancement
3962 Bugzilla : 7981/8208
3963 Description: Introduced Lustre Networking (LNET)
3964 Details : LNET is new networking infrastructure for Lustre, it includes
3965 a reorganized network configuration mode (see the user
3966 documentation for full details) as well as support for routing
3967 between different network fabrics. Lustre Networking Devices
3968 (LNDs) for the supported network fabrics have also been
3969 created for this new infrastructure.
3971 Severity : enhancement
3972 Description: Introduced Access control lists
3973 Details : clients can set ACLs on files and directories in order to have
3974 more fine-grained permissions than the standard Unix UGO+RWX.
3975 The MDS must be started with the "-o acl" mount option.
3977 Severity : enhancement
3978 Description: Introduced filesystem quotas
3979 Details : Administrators may now establish per-user quotas on the
3982 Severity : enhancement
3984 Description: Configuration change for the XT3
3985 The PTLLND is now used to run Lustre over Portals on the XT3
3986 The configure option(s) --with-cray-portals are no longer used.
3987 Rather --with-portals=<path-to-portals-includes> is used to
3988 enable building on the XT3. In addition to enable XT3 specific
3989 features the option --enable-cray-xt3 must be used.
3994 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3995 Details : A race between allocating a new llcd and re-getting the llcd_lock
3996 allowed another thread to grab newly-allocated llcd.
3998 Severity : enhancement
4000 Description: 2.6 OST async journal commit and locking fix to improve performance
4001 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4002 2.6 kernels are now async as they already were in 2.4 kernels so
4003 that they can commit concurrently with the network bulk transfer.
4004 For block-allocated files the filter allocation semaphore is held
4005 to avoid filesystem fragmentation during allocation. BKL lock
4006 removed for 2.6 xattr operations where it is no longer needed.
4011 Description: lconf incorrectly determined whether two IP networks could talk
4012 Details : In some more complicated routing and multiple-network
4013 configurations, lconf will avoid trying to make a network
4014 connection to a disjoint part of the IP space. It was doing the
4015 math incorrectly for one set of cases.
4020 Description: Fix for potential infinite loop processing records in an llog.
4021 Details : If an llog record is corrupted/zeroed, it is possible to loop
4022 forever in llog_process(). Validate the llog record length
4023 and skip the remainder of the block on error.
4026 Frequency : occasional (liblustre only)
4028 Description: liblustre could not open files whose last component is a symlink
4029 Details : sysio_path_walk() would incorrectly pass the open intent to
4030 intermediate path components.
4033 Frequency : rare (liblustre only with non-standard tuning)
4034 Bugzilla : 7201 (7350)
4035 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4036 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4037 after it has been released, only do this for referenced locks
4039 Severity : enhancement
4041 Description: specifying an (invalid) directory default stripe_size of -1
4042 would reset the directory default striping
4043 Details : stripe_size -1 was used internally to signal directory stripe
4044 removal, now use "all default" to signal dir stripe removal
4045 as a directory striping of "all default" is not useful
4048 Frequency : common for large clusters running liblustre clients
4050 Description: doing an ls when liblustre clients are running is slow
4051 Details : sending a glimpse AST to a liblustre client waits for every AST
4052 to time out, as liblustre clients will not respond. Since they
4053 cannot cache data we refresh the OST lock LVB from disk instead.
4055 Severity : enhancement
4057 Description: doing an ls at the same time as file IO can be slow
4058 Details : enqueue and other "small" requests can be blocked behind many
4059 large IO requests. Create a new OST IO portal for non-IO
4060 requests so they can be processed faster.
4063 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4065 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4066 created file with mode 000
4067 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4068 allow an ftruncate() as a non-root user to a file with mode 000.
4069 Lustre now respects this flag to disable mode checks when
4070 truncating a file owned by the user
4073 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4076 Description: Revoking locks from clients that went dead or catatonic might take
4078 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4079 cancellation of such locks instant on servers without waiting for
4080 any reply from clients. Clients drops these locks when cancel
4081 notification from server is received without replying.
4084 Frequency : liblustre-only, when liblustre client dies or becomes busy
4086 Description: Doing ls on Linux clients can take a long time with active
4088 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4089 granting such locks to it in the first place if possible. Locks
4090 are taken by proxy on the OST during the read or write and
4091 dropped immediately afterward. Add connect flags handling, do
4092 not grant locks to liblustre clients for glimpse ASTs.
4094 Severity : enhancement
4096 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4097 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4098 sequential reads and behaves badly with applications doing random
4099 reads. Improve it by reading ahead at least read region, and
4100 avoiding excessive large RPC for small reads.
4102 Severity : enhancement
4104 Description: Creating more than 1000 files for a single job may cause a load
4105 imbalance on the OSTs if there are also a large number of OSTs.
4106 Details : qos_prep_create() uses an OST index reseed value that is an
4107 even multiple of the number of available OSTs so that if the
4108 reseed happens in the middle of the object allocation it will
4109 still utilize the OSTs as uniformly as possible.
4114 Description: OST or MDS may oops in ping_evictor_main()
4115 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4116 but doesn't restart at beginning of obd_exports_timed list
4119 Severity : enhancement
4121 Description: improve by-nid export eviction on the MDS and OST
4122 Details : allow multiple exports with the same NID to be evicted at one
4123 time without re-searching the exports list.
4126 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4128 Description: MDS may oops in groups_free()
4129 Details : in rare race conditions a newly allocated group_info struct is
4130 freed again, and this can be NULL. The 2.4 compatibility code
4131 for groups_free() checked for a NULL pointer, but 2.6 did not.
4134 Frequency : common for liblustre clients doing little filesystem IO
4135 Bugzilla : 9352, 7313
4136 Description: server may evict liblustre clients accessing contended locks
4137 Details : if a client is granted a lock or receives a completion AST
4138 with a blocking AST already set it would not reply to the AST
4139 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4142 Frequency : lfs setstripe, only systems with more than 160 OSTs
4144 Description: unable to set striping with a starting offset beyond OST 160
4145 Details : llapi_create_file() incorrectly limited the starting stripe
4146 index to the maximum single-file stripe count.
4149 Frequency : LDAP users only
4151 Description: lconf did not handle in-kernel recovery with LDAP properly
4152 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4154 Severity : enhancement
4156 Description: bind OST threads to NUMA nodes to improve performance
4157 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4158 node and do their allocations there to localize memory access
4160 Severity : enhancement
4162 Description: llmount can determine client NID directly from Myrinet (GM)
4163 Details : the client NID code from gmnalnid was moved directly into
4164 llmount, removing the need to use this or specifying the
4165 client NID explicitly when mounting GM clients with zeroconf
4168 Frequency : if client is started with down MDS
4170 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4171 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4172 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4173 ptlrpc_import_delayed_req() because we didn't check if the
4174 request was interrupted, and we also didn't break out of the
4175 event loop if there was no timeout
4180 Description: data loss during non-page-aligned writes to a single file from
4181 both multiple nodes and multiple threads on one node at same time
4182 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4183 inconsistency led to false short-reads, that were cached and later
4184 used by ->prepare_write() to fill in partially written page,
4185 leading to data loss.
4188 Frequency : always, if lconf --abort_recovery used
4190 Description: lconf --abort_recovery fails with 'Operation not supported'
4191 Details : lconf was attempting to abort recovery on the MDT device and not
4194 Severity : enhancement
4196 Description: remove cleanup logs
4197 Details : replace lconf-generated cleanup logs with lustre internal
4198 cleanup routines. Eliminates the need for client-cleanup and
4201 Severity : enhancement
4203 Description: add support for EAs (user and system) on lustre filesystems
4204 Details : it is now possible to store extended attributes in the Lustre
4205 client filesystem, and with the user_xattr mount option it
4206 is possible to allow users to store EAs on their files also
4208 Severity : enhancement
4210 Description: Add possibility (config option) to show minimal available OST free
4212 Details : When compiled with --enable-mindf configure option, statfs(2)
4213 (and so, df) will return least minimal free space available from
4214 all OSTs as amount of free space on FS, instead of summary of
4215 free spaces of all OSTs.
4217 Severity : enhancement
4219 Description: do not expand extent locks acquired on OST-side
4220 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4221 by server: they are not cached anyway.
4224 Frequency : when mmap is used/binaries executed from Lustre
4226 Description: Unmmap pages before throwing them away from read cache.
4227 Details : llap_shrink cache now attempts to unmap pages before discarding
4228 them (if unmapping failed - do not discard). SLES9 kernel has
4229 extra checks that trigger if this unmapping is not done first.
4234 Description: lconf didn't resolve symlinks before checking to see whether a
4235 given mountpoint was already in use
4238 Frequency : when migrating failover services
4239 Bugzilla : 6395, 9514
4240 Description: When migrating a subset of services from a node (e.g. failback
4241 from a failover service node) the remaining services would
4242 time out and evict clients.
4243 Details : lconf --force (implied by --failover) sets the global obd_timeout
4244 to 5 seconds in order to quickly disconnect, but this caused
4245 other RPCs to time out too quickly. Do not change the global
4246 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4248 Severity : enhancement
4249 Frequency : if MDS is started with down OST
4250 Bugzilla : 9439,5706
4251 Description: Allow startup/shutdown of an MDS without depending on the
4252 availability of the OSTs.
4253 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4254 Add appropriate locking and lov-osc refcounts for safe
4255 cleaning. Add osc abort_inflight calls in case the
4256 synchronize never started.
4259 Frequency : occasional (Cray XT3 only)
4261 Description: root not authorized to access files in CRAY_PORTALS environment
4262 Details : The client process capabilities were not honoured on the MDS in
4263 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4264 been accessed by an authorized user then root was able to access
4265 the file on the local client also. The root user capabilities
4266 are now allowed on the MDS, as this environment has secure UID.
4269 Frequency : occasional
4271 Description: ldiskfs "too long searching" message happens too often
4272 Details : A debugging message (otherwise harmless) prints too often on
4273 the OST console. This has been reduced to only happen when
4274 there are fragmentation problems on the filesystem.
4279 Description: Division by zero in statfs when all OSCs are inactive
4280 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4281 lov_statfs divides by value returned by lov_get_stripecnt().
4285 Bugzilla : 9489, 3273
4286 Description: First write from each client to each OST was only 4kB in size,
4287 to initialize client writeback cache, which caused sub-optimal
4288 RPCs and poor layout on disk for the first writen file.
4289 Details : Clients now request an initial cache grant at (re)connect time
4290 and so that they can start streaming writes to the cache right
4291 away and always do full-sized RPCs if there is enough data.
4292 If the OST is rebooted the client also re-establishes its grant
4293 so that client cached writes will be honoured under the grant.
4298 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4299 Details : Now I/O RPCs go to different portal number and (presumably) fast
4300 lock requests (and glimses) and other RPCs get their own service
4301 threads pool that should be able to service those RPCs
4304 Severity : enhancement
4306 Description: Ability to exchange lustre version between client and servers and
4307 issue warnings at client side if client is too old. Also for
4308 liblustre clients there is ability to refuse connection of too old
4310 Details : New 'version' field is added to connect data structure that is
4311 filled with version info. That info is later checked by server and
4315 Frequency : rare, liblustre only.
4316 Bugzilla : 9296, 9581
4317 Description: Two simultaneous writes from liblustre at offset within same page
4318 might proceed at the same time overwriting eachother with stale
4320 Details : I/O lock withing llu_file_prwv was released too early, before data
4321 actually was hitting the wire. Extended lock-holding time until
4322 server acknowledges receiving data.
4325 Frequency : extremely rare. Never observed in practice.
4327 Description: avoid generating lustre_handle cookie of 0.
4328 Details : class_handle_hash() generates handle cookies by incrementing
4329 global counter, and can hit 0 occasionaly (this is unlikely, but
4330 not impossible, because initial value of cookie counter is
4331 selected randonly). Value of 0 is used as a sentinel meaning
4332 "unassigned handle" --- avoid it. Also coalesce two critical
4333 sections in this function into one.
4335 Severity : enhancement
4337 Description: allow liblustre clients to delegate truncate locking to OST
4338 Details : To avoid overhead of locking, liblustre client instructs OST to
4339 take extent lock in ost_punch() on client's behalf. New connection
4340 flag is added to handle backward compatibility.
4342 Severity : enhancement
4343 Bugzilla : 4928, 7341, 9758
4344 Description: allow number of OST service threads to be specified
4345 Details : a module parameter allows the number of OST service threads
4346 to be specified via "options ost ost_num_threads={N}" in the
4347 OSS's /etc/modules.conf or /etc/modprobe.conf.
4351 Bugzilla : 6146, 9635, 9895
4352 Description: servers crash with bad pointer in target_handle_connect()
4353 Details : In rare cases when a client is reconnecting it was possible that
4354 the connection request was the last reference for that export.
4355 We would temporarily drop the export reference and get a new
4356 one, but this may have been the last reference and the export
4357 was just destroyed. Get new reference before dropping old one.
4359 Severity : enhancement
4360 Frequency : if client is started with failover MDS
4362 Description: Allow multiple MDS hostnames in the mount command
4363 Details : Try to read the configuration from all specified MDS
4364 hostnames during a client mount in case the "primary"
4367 Severity : enhancement
4369 Description: Stop sending data to evicted clients as soon as possible.
4370 Details : Check if the client we are about to send or are sending data to
4371 was evicted already. (Check is done every second of waiting,
4372 for which l_wait_event interface was extended to allow checking
4373 of exit condition at specified intervals).
4376 Frequency : rare, normally only when NFS exporting is done from client
4378 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4380 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4383 Frequency : always, if srand() is called before liblustre initialization
4385 Description: Liblustre uses system PRNG disturbing its usage by user application
4386 Details : Introduce internal to lustre fast and high-quality PRNG for
4387 lustre usage and make liblustre and some other places in generic
4388 lustre code to use it.
4390 Severity : enhancement
4391 Bugzilla : 9477, 9557, 9870
4392 Description: Verify that the MDS configuration logs are updated when xml is
4393 Details : Check if the .xml configuration logs are newer than the config
4394 logs stored on the MDS and report an error if this is the case.
4395 Request --write-conf, or allow starting with --old_conf.
4397 Severity : enhancement
4399 Description: Handle symlinks in the path when checking if Lustre is mounted.
4400 Details : Resolve intermediate symlinks when checking if a client has
4401 mounted a filesystem to avoid duplicate client mounts.
4406 Description: lconf can hit an error exception but still return success.
4407 Details : The lconf command catches the Command error exception at the top
4408 level script context and will exit with the associated exit
4409 status, but doesn't ensure that this exit status is non-zero.
4414 Description: failure of ptlrpc thread startup can cause oops
4415 Details : Starting a ptlrpc service thread can fail if there are a large
4416 number of threads or the server memory is very fragmented.
4417 Handle this without oopsing.
4420 Frequency : always, only if liblustre and non-default acceptor port was used
4422 Description: liblustre cannot connect to servers with non-default acceptor port
4423 Details : tcpnal_set_default_params() was not called and was therefore
4424 ignoring the environment varaible TCPNAL_PORT, as well as other
4425 TCPNAL_ environment variables
4430 Description: two objects could be created on the same OST for a single file
4431 Details : If an OST is down, in some cases it was possible to create two
4432 objects on a single OST for a single file. No problems other
4433 than potential performance impact and spurious error messages.
4437 Bugzilla : 5681, 9562
4438 Description: Client may oops in ll_unhash_aliases
4439 Details : Client dcache may become inconsistent in race condition.
4440 In some cases "getcwd" can fail if the current directory is
4446 Description: Inode refcounting problems in NFS export code
4447 Details : link_raw functions used to call d_instantiate without obtaining
4448 extra inode reference first.
4452 Bugzilla : 9942, 9903
4453 Description: Referencing freed requests leading to crash, memleaks with NFS.
4454 Details : We used to require that call to ll_revalidate_it was always
4455 followed by ll_lookup_it. Also with revalidate_special() it is
4456 possible to call ll_revalidate_it() twice for the same dentry
4457 even if first occurence returned success. This fix changes semantic
4458 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4459 reference on a request referred from the intent.
4460 ll_intent_release() then releases such a request.
4463 Frequency : rare, normally benchmark loads only
4465 Description: unlinked inodes were kept in memory on the client
4466 Details : If a client is repeatedly creating and unlinking files it
4467 can accumulate a lot of stale inodes in the inode slab cache.
4468 If there is no other client load running this can cause the
4469 client node to run out of memory. Instead flush old inodes
4470 from client cache that have the same inode number as a new inode.
4473 Frequency : SLES9 2.6.5 kernel and long filenames only
4474 Bugzilla : 9969, 10379
4475 Description: utime reports stale NFS file handle
4476 Details : SLES9 uses out-of-dentry names in some cases, which confused
4477 the lustre dentry revalidation. Change it to always use the
4481 Frequency : rare, unless heavy write-truncate concurrency is continuous
4482 Bugzilla : 4180, 6984, 7171, 9963, 9331
4483 Description: OST becomes very slow and/or deadlocked during object unlink
4484 Details : filter_destroy() was holding onto the parent directory lock
4485 while truncating+unlinking objects. For very large objects this
4486 may block other threads for a long time and slow overall OST
4487 responsiveness. It may also be possible to get a lock ordering
4488 deadlock in this case, or run out of journal credits because of
4489 the combined truncate+unlink. Solution is to do object truncate
4490 first in one transaction without parent lock, and then do the
4491 final unlink in a new transaction with the parent lock. This
4492 reduces the lock hold time dramatically.
4495 Frequency : rare, 2.4 kernels only
4497 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4498 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4499 before unmounting their filesystem in order to avoid deadlock.
4500 The kernel_locked() function in 2.4 kernels only checks whether
4501 the kernel lock is held, not whether it is this process that is
4502 holding it as 2.6 kernels do.
4507 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4508 Details : The client ptlrpc code may be trying to reconnect to a down
4509 server before a previous connection attempt has timed out.
4510 Increase the reconnect interval to be longer than the connection
4511 timeout interval to avoid sending duplicate connections to
4515 Frequency : echo_client brw_test command
4517 Description: fix echo_client to work with OST preallocated code
4518 Details : OST preallocation code (5137) didn't take echo_client IO path
4519 into account: echo_client calls filter methods outside of any
4520 OST thread and, hence, there is no per-thread preallocated
4521 pages and buffers to use. Solution: hijack pga pages for IO. As
4522 a byproduct, this avoids unnecessary data copying.
4526 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4527 Description: Client can oops in mdc_commit_close() after open replay
4528 Details : It was possible for the MDS to return an open request with no
4529 transaction number in mds_finish_transno() if the client was
4530 evicted, but without actually returning an error. Clients
4531 would later try to replay that open and may trip an assertion
4532 Simplify the client close codepath, and always return an error
4533 from the MDS in case the open is not successful.
4536 Frequency : rare, 2.6 OSTs only
4538 Description: OST may deadlock under high load on fragmented files
4539 Details : If there was a heavy load and highly-fragmented OST filesystems
4540 it was possible to have all the OST threads deadlock waiting on
4541 allocation of biovecs, because the biovecs were not released
4542 until the entire RPC IO was completed. Instead, release biovecs
4543 as soon as they are complete to ensure forward IO progress.
4545 Severity : enhancement
4547 Description: Support for specifying external journal device at mount
4548 Details : If an OST or MDS device is formatted with an external journal
4549 device, this device major/minor is stored in the ext3 superblock
4550 and may not be valid for failover. Allow detecting and
4551 specifying the external journal at mount time.
4556 Description: Mounting an MDS with pending unlinked files may cause oops
4557 Details : target_finish_recovery() calls mds_postrecov() which returned
4558 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4559 considers this an error and immediately begins cleaning up the
4560 lov, just after starting the mds_lov process
4562 Severity : enhancement
4564 Description: Implement 'lfs df' to report actual free space on per-OST basis
4565 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4566 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4567 usage of objects; '-h' to report in human readable format.
4569 ------------------------------------------------------------------------------
4571 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4578 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4579 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4580 at startup during orphan unlinking. Instead, skip buddy update
4581 before setup, as it will be regenerated after recovery is complete.
4586 Description: If an OST is inactive, its locks might reference stale inodes.
4587 Details : lov_change_cbdata() must iterate over all namespaces, even if
4588 they are inactive to clear inode references from the lock.
4590 Severity : enhancement
4591 Frequency : occasional, if non-standard max_dirty_mb used
4593 Description: Client will block write RPCs if not enough grant
4594 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4595 then the client will block writes while waiting for another RPC
4596 to complete instead of consuming its dirty limit. With change
4597 we get improved performance when max_dirty_mb is small.
4599 Severity : enhancement
4600 Bugzilla : 3389, 6253
4601 Description: Add support for supplementary groups on the MDS.
4602 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4603 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4604 lookups for user supplementary groups into a cache.
4608 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4609 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4610 for liblustre O_EXCL is enforced by MDS.
4614 Description: readdir never returns NULL in liblustre.
4615 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4616 dentry in struct dirent.
4620 Frequency : liblustre only, depends on application IO pattern
4621 Description: liblustre clients evicted if not contacting servers
4622 Details : Don't put liblustre clients into the ping_evictor list, so
4623 they will not be evicted by the pinger ever.
4625 Severity : enhancement
4627 Description: Add ability to evict clients by NID from MDS.
4628 Details : By echoing "nid:$NID" string into
4629 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4630 $NID would be instantly evicted from this MDS and from all active
4631 OSTs connected to it.
4635 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4636 Details : lookup_it_finish() used to query file size from OSTs that was not
4641 Description: service threads change working directory to that of init
4642 Details : Starting lustre service threads may pin the working directory
4643 of the parent thread, making that filesystem busy. Threads
4644 now change to the working directory of init to avoid this.
4648 Frequency : during shutdown only
4649 Description: shutdown with a failed MDS or OST can cause unmount to hang
4650 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4655 Frequency : 2.6 only
4656 Description: chmod/chown may include an extra supplementary group
4657 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4658 supplementary group and if none is specified this is used.
4661 Bugzilla : 5479 (6816)
4663 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4664 Details : If lookup is in progress on a file that is unlinked we might try
4665 to revalidate the inode and fail in revalidate after lookup is
4666 complete and ll_file_open() enqueues the open again but
4667 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4670 Frequency : always, if lconf --abort_recovery used
4672 Description: lconf --abort_recovery fails with 'Operation not supported'
4673 Details : lconf was attempting to abort recovery on the MDT device and not
4676 ------------------------------------------------------------------------------
4678 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4683 Frequency : rare (only unsupported configurations with a node running as an
4685 Bugzilla : 6514, 5137
4686 Description: Mounting a Lustre file system on a node running as an OST could
4688 Details : OSTs now preallocates memory needed to write out data at
4689 startup, instead of when needed, to avoid having to
4690 allocate memory in possibly low memory situations.
4691 Specifically, if the file system is mounted on on OST,
4692 memory pressure could force it to try to write out data,
4693 which it needed to allocate memory to do. Due to the low
4694 memory, it would be unable to do so and the node would
4695 become unresponsive.
4697 Severity : enhancement
4699 Description: Addition of lconf --service command line option
4700 Details : lconf now accepts a '--service <arg>' option, which is
4701 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4703 Severity : enhancement
4705 Description: Failover mode is now the default for OSTs.
4706 Details : By default, OSTs will now run in failover mode. To return to
4707 the old behaviour, add '--failout' to the lmc line for OSTs.
4709 Severity : enhancement
4711 Description: Health checks are now provided for MDS and OSTs
4712 Details : Additional detailed health check information on MSD and OSTs
4713 is now provided through the procfs health_check value.
4716 Frequency : occasional, depends on IO load
4718 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4719 after numerous create/delete cycles
4720 Details : The ext3 inode allocation policy would not allocate new inodes
4721 very well on the OSTs because there are no new directories
4722 being created. Instead we look for groups with free space if
4723 the parent directories are nearly full.
4728 Description: Network or server problems during mount may cause partially
4729 mounted clients instead of returning an error.
4730 Details : The config llog parsing code may overwrite the error return
4731 code during mount error handling, returning success instead
4737 Description: MDS can fail to allocate large reply buffers
4738 Details : After long uptimes the MDS can fail to allocate large reply
4739 buffers (e.g. zconf client mount config records) due to memory
4740 fragmentation or consumption by the buffer cache. Preallocate
4741 some large reply buffers so that these replies can be sent even
4742 under memory pressure.
4746 Frequency : rare (liblustre)
4747 Description: fsx running with liblustre complained that using truncate() to
4748 extend the file doesn't work. This patch corrects that issue.
4749 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4750 ATTR_SIZE and lsm use in llu_setattr_raw.
4754 Frequency : rare, only 2.6 kernels
4755 Description: Unusual file access patterns on the MDS may result in inode
4756 data being lost in very rare circumstances.
4757 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4758 code in the 2.6 kernel caused the inode and/or EA data not to
4759 be read from disk, causing single-file corruption.
4763 Frequency : rare, only 2.6 filesystems using extents
4764 Description: Heavy concurrent write and delete load may cause data corruption.
4765 Details : It was possible under high-load situations to have an extent
4766 metadata block in the block device cache from a just-unlinked
4767 file overwrite a newly-allocated data block. We now unmap any
4768 metadata buffers that alias just-allocated data blocks.
4772 Frequency : filesystems with default stripe_count larger than 77
4773 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4774 Details : lconf specifies an inode size of 4096 bytes when the default
4775 stripe_count is larger than 77. This conflicts with the default
4776 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4777 this case to avoid pinning too much memory for large EAs.
4779 ------------------------------------------------------------------------------
4781 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4786 Frequency : rare (extremely heavy IO load with hundreds of clients)
4788 Description: Client is evicted, gets IO error writing to file
4789 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4790 caused clients to be evicted for AST timeouts. The fixes in
4791 bug 5192 mean we no longer need to have such short AST timeouts
4792 so ldlm_timeout has been increased.
4795 Frequency : occasional during --force or --failover shutdown under load
4796 Bugzilla : 5949, 4834
4797 Description: Server oops/LBUG if stopped with --force or --failover under load
4798 Details : a collection of import/export refcount and cleanup ordering
4799 issues fixed for safer force cleanup
4802 Frequency : only filesystems larger than 120 OSTs
4803 Bugzilla : 5990, 6223
4804 Description: lfs getstripe would oops on a very large filesystem
4805 Details : lov_getconfig used kfree on vmalloc'd memory
4808 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4809 Bugzilla : 6242, 6243
4810 Description: reading from files that had been truncated to a non-zero size
4811 but never opened returned no data
4812 Details : ll_file_read() reads zeros from no-object files to EOF
4817 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4818 Details : There's little harm in aborting MDS/OSS recovery and letting it
4819 try again, so I removed the LASSERT and return an error instead.
4821 Severity : enhancement
4823 Description: New debugging infrastructure for tracking down data corruption
4824 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4825 (b) cover more of the client-side code path, and (c) try to narrow
4826 down where problems occurred
4830 Bugzilla : 3819, 4364, 4397, 6313
4831 Description: Racing close and eviction MDS could cause assertion in mds_close
4832 Details : It was possible to get multiple mfd references during close and
4833 client eviction, leading to one thread referencing a freed mfd.
4835 Severity: : enhancement
4836 Bugzilla : 3262, 6359
4837 Description: Attempts to reconnect to servers are now more aggressive.
4838 Details : This builds on the enhanced upcall-less recovery that was added
4839 in 1.4.2. When trying to reconnect to servers, clients will
4840 now try each server in the failover group every 10 seconds. By
4841 default, clients would previously try one server every 25 seconds.
4846 Description: After recovery, certain operations trigger a failed
4847 assertion on a client.
4848 Details : Failing over an mds, using lconf -d --failover, while a
4849 client was doing a readdir() call would cause the client to
4850 LBUG after recovery completed and the readdir() was resent.
4852 Severity : enhancement
4854 Description: Default groups are now added by lconf
4855 Details : You can now run lconf --group <servicename> without having to
4856 manually add groups with lmc.
4859 Frequency : occasional
4861 Description: Nodes with an elan id of 0 trigger a failed assertion
4864 Frequency : always when accessing e.g. tty/console device nodes
4866 Description: tty and some other devices nodes cannot be used on lustre
4867 Details : file's private_data field is used by device data and lustre
4868 values in there got lost. New field was added to struct file to
4869 store fs-specific private data.
4872 Frequency : when exporting Lustre via NFS
4874 Description: NFSD failed occasionally when looking up a path component
4875 Details : NFSD is looking up ".." which was broken in ext3 directories
4876 that had grown large enough to become hashed.
4879 Frequency : Clusters with multiple interfaces not on the same subnet
4881 Description: Nodes will repeatedly try to reconnect to an interface which it
4882 cannot reach and report an error to the log.
4883 Details : Extra peer list entries will be created by lconf with some peers
4884 unreachable. lconf now validates the peer before adding it.
4887 Frequency : Only if a default stripe is set on the filesystem root.
4889 Description: Setting a default stripe on the filesystem root prevented the
4890 filesystem from being remounted.
4891 Details : The client was sending extra request flags in the root getattr
4892 request and did not allocate a reply buffer for the dir EA.
4895 Frequency : occasional, higher if lots of files are accessed by one client
4896 Bugzilla : 6159, 6097
4897 Description: Client trips assertion regarding lsm mismatch/magic
4898 Details : While revalidating inodes the VFS looks up inodes with ifind()
4899 and in rare cases can find an inode that is being freed.
4900 The ll_test_inode() code will free the lsm during ifind()
4901 when it finds an existing inode and then the VFS later attaches
4902 this free lsm to a new inode.
4906 Bugzilla : 6422, 7030
4907 Description: MDS deadlock between mkdir and client eviction
4908 Details : Creating a new file via mkdir or mknod (starting a transaction
4909 and getting the ns lock) can deadlock with client eviction
4910 (gets ns lock and trying to finish a synchronous transaction).
4913 Frequency : occasional
4914 Description: While starting a server, the fsfilt_ext3 module could not be
4916 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4917 kernels. Previously, lconf would still use the ext3 name
4918 when trying to load modules. Now, it will correctly use
4919 ext3 on 2.4 and ldiskfs on 2.6.
4921 Severity : enhancement
4922 Description: The default stripe count has been changed to 1
4923 Details : The interpretation of the default stripe count (0, to lfs
4924 or lmc) has been changed to mean striping across a single
4925 OST, rather than all available. For general usage we have
4926 found a stripe count of 1 or 2 works best.
4928 Severity : enhancement
4929 Description: Add support for compiling against Cray portals.
4930 Details : Conditional compiling for some areas that are different
4934 Frequency : occasional
4935 Bugzilla : 6409, 6834
4936 Description: Creating files with an explicit stripe count may lead to
4937 a failed assertion on the MDS
4938 Details : If some OSTs are full or unavailable, creating files may
4939 trigger a failed assertion on the MDS. Now, Lustre will
4940 try to use other servers or return an error to the
4944 Frequency : occasional
4946 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4947 caused lock timeout during readahead (since 1.4.2).
4948 Details : Processes doing readahead might match a lock that hasn't been
4949 granted yet if there are overlapping and conflicting lock
4950 requests. The readahead process waits on ungranted lock
4951 (original lock is CBPENDING), while OST waits for that process
4952 to cancel CBPENDING read lock and eventually evicts client.
4954 Severity : enhancement
4956 Description: Initial enabling of flock support for clients
4957 Details : Implements fcntl advisory locking and file status functions.
4958 This feature is provided as an optional mount flag (default
4959 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4960 locking are implemented yet, and those that are are not guaranteed
4961 to be completely correct in production environments.
4962 mount -t lustre -o [flock|noflock] ...
4965 Frequency : occasional
4967 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4968 assertion in the ext3 JBD (journaling) layer.
4969 Details : The b_committed_data struct is protected by the big kernel lock
4970 in 2.4 kernels, serializing journal_commit_transaction() and
4971 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4972 access to this struct. In 2.6 kernels there is finer grained
4973 locking to improve SMP performance of the JBD layer.
4977 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4982 Description: merge and cleanup kernel patches.
4985 -----------------------------------------------------------------------------
4987 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4989 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4990 versions of Lustre. Please update all servers and clients to
4991 version 1.4.2 or later at the same time. You must also run
4992 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4993 to update the configuration logs.
4995 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4996 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4997 - swab llog records of type '0' so we get proper header size/idx (5861)
4998 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4999 - fix rename of one directory over another leaking an inode (5953)
5000 - avoid SetPageDirty on 2.6 (5981)
5001 - don't re-add just-being-destroyed locks to the waiting list (5653)
5002 - when creating new directories, inherit the parent's custom
5003 striping settings if present parent (3048)
5004 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5005 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5006 - don't hold client locks on temporary worklist from l_lru (5666)
5007 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5008 - automatically evict dead clients (5921)
5009 - Update file size properly in create+truncate+fstat case (6196)
5010 - Do not unhash mountpoint dentries, do not allow removal of
5012 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5013 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5014 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5015 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5016 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5017 - Correctly handle reads of files with no objects (6243)
5018 - lctl recover will also mark a device active if deactivate used (5933)
5020 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5021 - allow --write-conf on an MDS with different nettype than client (5619)
5022 - don't write config llogs to MDS for mounts not from that MDS (5617)
5023 - lconf should create multiple TCP connections from a client (5201)
5024 - init scripts are now turned off by default; run chkconfig --on
5025 lustre and chkconfig --on lustrefs to use them
5026 - upcalls are no longer needed for clients to recover to failover
5028 - add --abort-recovery option to lconf to abort recovery on device
5030 - add support for an arbitrary number of OSTs (3026)
5031 - Quota support protocol changes.
5032 - forward compatibility changes to wire structs (6007)
5033 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5034 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5035 - improved "lustre status" script
5036 - initialize blocksize for non-regular files (6062)
5037 - added --disable-server and --disable-client configure options (5782)
5038 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5039 - Vanilla 2.4.29 support
5040 - increase maximum number of obd devices to 520 (6242)
5041 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5042 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5043 SLES 9 2.6.5 kernels
5044 - limit stripes per file to 160 (the maximum EA size) (6093)
5046 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5049 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5050 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5051 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5052 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5053 - small ext3 extents cleanups and fixes (5733)
5054 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5055 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5056 - avoid needless client->OST connect, fix handle mismatch (5317)
5057 - fix DLM error path that led to out-of-sync client, long delays (5779)
5058 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5059 - fix several locking issues related to i_size (5492,5624,5654,5672)
5060 - don't move pending lock onto export if it is already evicted (5683)
5061 - fix kernel oops when creating .foo in unlinked directory (5548)
5062 - fix deadlock in obdfilter statistics vs. object create (5811)
5063 - use time_{before,after} to avoid timer jiffies wrap (5882)
5064 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5065 - Do not leak request if server was not able to process it (5154)
5066 - If mds_open unable to find parent dir, make that negative lookup(5154)
5067 - don't create new directories with extent-mapping (5909, 5936)
5069 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5070 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5071 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5072 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5073 - support for RHEL4 kernel on i686 (5773)
5074 - provide error messages when incompatible logs are encountered (5898)
5076 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5077 * version 1.4.0.10 (1.4.1 release candidate 1)
5079 - don't keep a lock reference when lock is not granted (4238)
5080 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5081 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5082 - drop import inflight refcount on signal_completed_replay error (5255)
5083 - unlock page after async write error during send (3677)
5084 - handle missing objects in filter_preprw_read properly (5265)
5085 - no transno return for symlink open, don't save no-trasno open (3440)
5086 - don't try to complete elan receive that already failed (4012)
5087 - free RPC server reply state on error (5406)
5088 - clean up thread from ptlrpc_start_thread() on error (5160)
5089 - readahead could read extra page into cache that wasn't ejected (5388)
5090 - prevent races in class_attach/setup/cleanup/detach (5260)
5091 - don't dereference de->d_inode after l_dput of de (5458)
5092 - use "int" for stripe value returned from lock_to_stripe (5544)
5093 - mballoc allocation and error-checking fixes in 2.6 (5504)
5094 - block device patches to fix I/O request sizes in 2.6 (5482)
5095 - look up hostnames for IB nals (5602)
5096 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5097 - don't start multiple acceptors for the same port (5277)
5098 - fix incorrect LASSERT in mds_getattr_name (5635)
5099 - export a proc file for general "ping" checking (5628)
5100 - fix "lfs check" to not block when the MDS is down (5628)
5102 - service request history (4965)
5103 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5104 - create an "evict_client" /proc entry on OSTs, like the MDS has
5105 - fix mount usage message, return errors per mount(8) (5168)
5106 - change grep [] to grep "[]" in tests so they work in more UMLs
5107 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5108 - remove ancient LOV_MAGIC_V0 EA support (5047)
5109 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5110 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5111 - allow readahead allocations to fail when low on memory (5383)
5112 - mmap locking landed again, after considerable improvement (2828)
5113 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5115 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5118 - send OST transaction number in read/write reply to free req (4966)
5119 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5120 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5121 - the watchdog thread now runs as interruptible (5246)
5122 - flock/lockf fixes (but it's still disabled, pending 5135)
5123 - don't use EXT3 constants in llite code (5094)
5124 - memory shortage at startup could cause assertion (5176)
5126 - reorganization of lov code
5127 - single portals codebase
5129 - add extents/mballoc support (5025)
5130 - direct I/O reads in the obdfilter (4048)
5131 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5133 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5136 - send OST transaction number in read/write reply to free req (4966)
5137 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5138 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5139 - the watchdog thread now runs as interruptible (5246)
5140 - handle missing objects in filter_preprw_read properly (5265)
5141 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5142 - drop import inflight refcount on signal_completed_replay error (5255)
5143 - unlock page after async write error during send (3677)
5144 - return original error code on reconstructed replies (3761)
5145 - no transno return for symlink open, don't save no-trasno open (3440)
5147 - add pid to ldlm debugging output (4922)
5148 - bump the watchdog timeouts -- we can't handle 30sec yet
5149 - extra debugging for orphan dentry/inode bug (5259)
5151 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5154 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5155 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5156 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5157 - fix NULL dereference in /proc/sys/portals/routes (4827)
5158 - allow failed mdc_close() operations to be interrupted (4561)
5159 - stop precreate on OST before MDS would time out on it (4778)
5160 - don't send partial-page writes before EOF from client (4410)
5161 - discard client grant for sub-page writes on large-page clients (4520)
5162 - don't free dentries not owned by NFS code, check generation (4806)
5163 - fix lsm leak if mds_create_objects() fails (4801)
5164 - limit debug_daemon file size, always print CERROR messages (4789)
5165 - use transno after validating reply (3892)
5166 - process timed out requests if import state changes (3754)
5167 - update mtime on OST during writes, return in glimpse (4829)
5168 - add mkfsoptions to LDAP (4679)
5169 - use ->max_readahead method instead of zapping global ra (5039)
5170 - don't interrupt __l_wait_event() during strace
5172 - add software watchdogs to catch hung threads quickly (4941)
5173 - make lustrefs init script start after nfs is mounted
5174 - fix CWARN/ERROR duplication (4930)
5175 - return async write errors to application if possible (2248)
5176 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5177 - print NAL number in %x format (4645)
5178 - update barely-supported suse-2.4.21-171 series (4842)
5179 - support for sles 9 %post scripts
5180 - support for building 2.6 kernel-source packages
5181 - support for sles km_* packages
5183 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5186 - ignore -ENOENT errors in osc_destroy (3639)
5187 - notify osc create thread that OSC is being cleaned up (4600)
5188 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5189 - reconstruct ost_handle() like mds_handle() (4657)
5190 - create a new thread to do import eviction to avoid deadlock (3969)
5191 - let lconf resolve symlinked-to devices (4629)
5192 - don't unlink "objects" from directory with default EA (4554)
5193 - hold socknal file ref over connect in case target is down (4394)
5194 - allow more than 32000 subdirectories in a single directory (3244)
5195 - fix blocks count for O_DIRECT writes (3751)
5196 - OST returns ENOSPC from object create when no space left (4539)
5197 - don't send truncate RPC if file size isn't changing (4410)
5198 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5199 - bind to privileged port in socknal and tcpnal (3689)
5201 - rate limit CERROR/CWARN console message to avoid overload (4519)
5202 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5203 - basic mmap support (3918)
5204 - kernel patch series update from b1_4 (4711)
5206 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5209 - avoid crash during MDS cleanup with OST shut down (2775)
5210 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5211 - don't use bad inodes on the MDS (3744)
5212 - dynamic object preallocation to improve recovery speed (4236)
5213 - don't hold spinlock over lock dumping or change debug flags (4401)
5214 - don't zero obd_dev when it is force cleaned (3651)
5215 - print grants to console if they go negative (4431)
5216 - "lctl deactivate" will stop automatic recovery attempts (3406)
5217 - look for existing locks in ldlm_handle_enqueue() (3764)
5218 - don't resolve lock handle twice in recovery avoiding race (4401)
5219 - revalidate should check working dir is a directory (4134)
5221 - don't always mark "slow" obdfilter messages as errors (4418)
5223 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5226 - don't close LustreDB during write_conf until it is done (3860)
5227 - fix typo in lconf for_each_profile (3821)
5228 - allow dumping logs from multiple threads at one time (3820)
5229 - don't allow multiple threads in OSC recovery (3812)
5230 - fix debug_size parameters (3864)
5231 - fix mds_postrecov to initialize import for llog ctxt (3121)
5232 - replace config semaphore with spinlock (3306)
5233 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5234 - don't allow enqueue to complete on a destroyed export (3822)
5235 - down write_lock before checking llog header bitmap (3825)
5236 - recover from lock replay timeout (3764)
5237 - up llog sem before sending rpc (3652)
5238 - reduce ns lock hold times when setting kms (3267)
5239 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5240 - fix NULL deref and obd_dev leak on setup error (3312)
5241 - replace some LBUG about llog ops with error handling (3841)
5242 - don't match INVALID dentries from d_lookup and spin (3784)
5243 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5244 - fix invalid assertion in ptlrpc_set_wait (3880)
5246 - add libwrap support for the TCP acceptor (3996)
5247 - add /proc/sys/portals/routes for non-root route listing (3994)
5248 - allow setting MDS UUID in .xml (2580)
5249 - print the stack of a process that LBUGs (4228)
5251 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5254 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5255 - make sure to unset replay flag from failed open requests (3440)
5256 - if default stripe count is 0, use OST count for inode size (3636)
5257 - update parent mtime/ctime on client for create/unlink (2611)
5258 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5259 - free recovery state on server during a forced cleanup (3571)
5260 - unregister_reply for resent reqs (3063)
5261 - loop back devices mounting and status check on 2.6 (3563)
5262 - fix resource-creation race that can provoke i_size == 0 (3513)
5263 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5264 - more debugging for page-accounting assertion (3746)
5265 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5266 - avoid deadlock after precreation failure (3758)
5267 - fix race and lock order deadlock in orphan handling (3450, 3750)
5268 - add validity checks when grabbing inodes from l_ast_data (3599)
5270 - add /proc/.../recovery_status to obdfilter (3428)
5271 - lightweight CDEBUG infrastructure, debug daemon (3668)
5272 - change default OSC RPC parameters to be better on small clusters
5273 - turn off OST read cache for files smaller than 32MB
5274 - install man pages and include them in rpms (3100)
5275 - add new init script for (un)mounting lustre filesystems (2593)
5276 - run chkconfig in %post for init scripts (3701)
5277 - drop scimac NAL (unmaintained)
5279 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5282 - clean kiobufs before and after use (3485)
5283 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5284 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5285 - change init script to fail more clearly if not run as root (1528)
5286 - allow clients to reconnect during replay (1742)
5287 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5288 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5289 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5290 - fix cond_resched (3554)
5291 - only evict unfinished clients after recovery (3515)
5292 - allow bulk resend, prevent data loss (3570)
5293 - dynamic ptlrpc request buffer allocation (2102)
5294 - don't allow unlinking open directory if it isn't empty (2904)
5295 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5296 - remove extraneous obd dereference causing LASSERT failure (3334)
5297 - don't use get_cycles() when creating temp. files on the mds (3156)
5298 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5299 - handle EEXIST for set-stripe, set proper directory name (3336)
5301 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5302 - fix ksocknal_fmb_callback() error messages (2918)
5304 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5307 - don't copy lvb into (possibly NULL) reply on error (2983)
5308 - don't deref dentry after dput, don't free lvb on error (2922)
5309 - use the kms to determine writeback rpc length (2947)
5310 - increment oti_logcookies when osc is inactive (2948)
5311 - update client's i_blocks count via lvb messages (2543)
5312 - handle intent open/close of special files properly (1557)
5313 - mount MDS with errors=remount-ro, like obdfilter (2009)
5314 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5315 - don't use cancelling-locks' kms values (2947)
5316 - use highest lock extent for kms, not last one (2925)
5317 - don't dereference ERR_PTR() dentry in error handling path (3107)
5318 - fix thread race in portals_debug_dumplog() (3122)
5319 - create lprocfs device entries at setup instead of at attach (1519)
5320 - common AST error handler, don't evict client on completion race (3145)
5321 - zero nameidata in detach_mnt in 2.6 (3118)
5322 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5323 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5324 - initialize RPC timeout timer earlier for 2.6 (3219)
5325 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5326 - print nal/nid for unknown nid (3258)
5327 - additional checks for oscc recovery before doing precreate (3284)
5328 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5329 - don't crash in mdc_close for bad permissions on open (3285)
5330 - zero i_rdev for non-device files (3147)
5331 - clear page->private before handing to FS, better assertion (3119)
5332 - tune the read pipeline (3236)
5333 - fix incorrect decref of invalidated dentry (2350)
5334 - provide read-ahead stats and refine rpc in flight stats (3328)
5335 - don't hold journal transaction open across create RPC (3313)
5336 - update atime on MDS at close time (3265)
5337 - close LDAP connection when recovering to avoid server load (3315)
5338 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5339 - don't leak open file on MDS after open resend (3325)
5340 - serialize filter_precreate and filter_destroy_precreated (3329)
5341 - loop device shouldn't call sync_dev() for nul device (3092)
5342 - clear page cache after eviction (2766)
5343 - resynchronize MDS->OST in background (2824)
5344 - refuse to mount the same filesystem twice on same mountpoint (3394)
5345 - allow llmount to create routes for mounting behind routers (3320)
5346 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5347 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5348 - fix rare problem with rename on htree directories (3417)
5350 - allow default OST striping configuration per directory (1414)
5351 - fix compilation for qswnal for 2.6 kernels (3125)
5352 - increase maximum number of MDS request buffers for large systems
5353 - change liblustreapi to be useful for external progs like lfsck (3098)
5354 - increase local configuration timeout for slow disks (3353)
5355 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5357 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5360 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5361 - don't overwrite extent policy data in reply if lock was blocked (2901)
5362 - drop filter export grants atomically with removal from device (2663)
5363 - del obd_self_export from work_list in class_disconnect_exports (2908)
5364 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5365 - swab reply message in mdc_close, other PPC fixes (2464)
5366 - fix destroying of named logs (2325)
5367 - overwrite old logs when running lconf --write_conf (2264)
5368 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5369 - fix race in target_handle_connect (2898)
5370 - mds_reint_create() should take same inode create lock (2926)
5371 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5372 - don't close files for self_export to avoid uninitialized obd (2936)
5373 - allow MDS with the same name as client node (2939)
5374 - hold dentry reference for closed log files for unlink (2325)
5375 - reserve space for all logs during transactions (2059)
5376 - don't evict page beyond end of stripe extent (2925)
5377 - don't oops on a deleted current working directory (2399)
5378 - handle hard links to targets without a parent properly (2517)
5379 - don't dereference NULL lock when racing during eviction (2867)
5380 - don't grow lock extents when lots of conflicting locks (2919)
5382 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5385 - account for cache space usage on clients to avoid data loss (974)
5386 - lfsck support in lustre kernel code (2349)
5387 - reduce journal credits needed for BRW writes (2370)
5388 - orphan handling to avoid losing space on client/server crashes
5389 - ptlrpcd can be blocked, stopping ALL progress (2477)
5390 - use lock value blocks to assist in proper KMS, faster stat (1021)
5391 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5392 - recovery for initial connections (2355)
5393 - fixes for mds_cleanup_orphans (1934)
5394 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5395 - block all file creations until orphan recovery completes (1901)
5396 - client remove rq_connection from request struct (2423)
5397 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5398 - recovery timer race (2670)
5399 - mdc_close recovey bug (2532)
5400 - ptlrpc cleanup bug (2710)
5401 - mds timeout on local locks (2588)
5402 - namespace lock held during RPCs (2431)
5403 - handle interrupted sync write properly (2503)
5404 - don't try to handle a message that hasn't been replied to (2699)
5405 - client assert failure during cleanup after abort recovery (2701)
5406 - leak mdc device after failed mount (2712)
5407 - ptlrpc_check_set allows timedout requests to complete (2714)
5408 - wait for inflight reqs when ptlrpcd finishes (2710)
5409 - make sure unregistered services are removed from the srv_list
5410 - reset bulk XID's when resending them (caught by 1138 test)
5411 - unregister_bulk after timeout
5412 - fix lconf error (2694)
5413 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5414 - readahead locks pages, leaves pending causing memory pressure (2673)
5415 - increase OST request buffers to 4096 on large machines (2729)
5416 - fix up permission of existing directories in simple_mkdir (2661)
5417 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5418 - don't assign transno to errored transactions (2742)
5419 - don't delete objects on OST if given a bogus objid from MDS (2751)
5420 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5421 - if rq_no_resend, then timeout request after recovery (2432)
5422 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5423 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5424 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5425 - MDS recovery shouldn't delete objects that it has given out (2730)
5426 - if enqueue arrives after completion, don't clobber LVB (2819)
5427 - don't unlock pages twice when trigger_group_io returns error (2814)
5428 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5429 - don't write pages to disk if there was an error (1450)
5430 - don't ping imports that have recovery disabled (2676)
5431 - take buffered bytes into account when balancing socknal conn (2817)
5432 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5433 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5434 - remove little-endian swabbing of llog records (1987)
5435 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5436 - retry reposting request buffers if they fail (1191)
5437 - grow extent at grant time to avoid granting a revoked lock (2809)
5438 - lock revoke doesn't evict page if covered by a second lock (2765)
5439 - disable VM readahead to avoid reading outside lock extents (2805)
5441 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5442 - updated kernel patches for hp-2.4.20 kernel (2681)
5444 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5447 - fix truncated write corruption (2366)
5448 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5450 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5451 - ensure len(uuid) < 37 in lmc (1171)
5452 - fix ia64 OOPS in llog_test (2255)
5453 - zero end of page at obdfilter for partial page writes (2648)
5454 - don't leave stale dentries around after renames (bug 2428)
5455 - fix timeouts when evicting a client with a single lock held (2642)
5456 - set deadline for the initial HELLO message to drain (2634)
5457 - print out dotted-quad IP addresses in the socknal (2302)
5459 - additional debugging for MDS client eviction problem (2443)
5460 - fix mkfsoptions support for osts (2603, 2604)
5462 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5465 - add series for the vanilla 2.6.0 kernel
5466 - add series for the vanilla 2.4.24 kernel
5467 - add series for a cray x86/64 UL kernel drop
5468 - fix xattr patches for the vanilla 2.4.19 series
5470 - generate true UUIDs in lmc (1171)
5471 - have portals stack dumping break in UML (2466)
5472 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5473 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5474 - wake callback waiting threads on client eviction (2460)
5475 - Add --ptldebug and --subsystem to lmc (1719)
5476 - update assertion to allow safe interrupt allocation
5477 - set rq_no_resend for cancel requests (2432)
5478 - recalculate ptlrpcd timeout after resend (2494)
5479 - call vfs_rmdir when removing pending directories (2368)
5480 - fix renaming a file to itself (2429)
5481 - lmc creates a default one-stripe lov (2454)
5482 - expand procfs space to handle large clusters (2326)
5483 - increase UML stack to avoid overflow
5484 - update lconf's list of debug and subsystem masks
5485 - fix lfs find --obd (2510)
5486 - /proc tunable for disabling filter read caching (2591)
5487 - stop rpm packages from altering slapd.conf (2301)
5488 - disable nagle in the socknal under 0conf (2578)
5489 - choose mds inode size based on stripe count (2572)
5490 - fix kernel-source rpm problems (2516)
5492 - add --disable-doc to avoid pdf generation (2421)
5493 - update documentation, tests, type-os, comments
5494 - avoid format warnings on ia64
5495 - remove the TOE NAL
5496 - tiny code cleanups by removing unused fields
5498 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5501 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5502 - remove the most blatant lies from BUILDING file (2371)
5503 - change default debug level to reasonable production setting
5504 - reduce client side cache size to reduce cache flush time
5505 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5506 - make TCP zerocopy and pinger support enabled by default (2476)
5507 - sync writes completed after process exits caused crashes (2319)
5508 - maintain correct mount count on the MDS (2356)
5509 - backout 1557, because 2316 wasn't really fixed
5510 - better file I/O statistics gathering in /proc
5511 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5512 - another kernel patch to fix zero-copy TCP function export
5513 - don't take duplicate lock when processing re-sent getattr (2420)
5514 - lctl uses obd_self_export instead of creating new conn (2353)
5515 - MDS/OST recovery case which requires object creation asserted (2425)
5516 - move lfs from /usr/sbin to /usr/bin in packages
5517 - fix race between mds_client_add and mds_client_free (2417)
5518 - use kmalloc instead of slabs in portals (2430)
5519 - don't create duplicate records when a failover MDS is present (2442)
5520 - remove unnecessary mount age check (2332)
5521 - don't remove directory inodes from locks prematurely (2451)
5522 - don't break if MDS service name is the same as hostname (2103)
5523 - fix races in client write RPC generation when cache full (2482)
5525 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5528 - remove now-unused request->rq_obd (278)
5529 - if an allocation fails, print out how much memory we've used (1933)
5530 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5531 - add the "configurable stack size" patch to most series files (1256)
5532 - ability to write large log records, for 100+ OST configs (2306)
5533 - fix NULL deref when filter_prep fails (2314)
5534 - fix operator precedence error in filter_sync
5535 - dynamic allocation of socknal TX descriptors (2315)
5536 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5537 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5538 - ensure that CWARN messages in Portals always get to the syslog
5539 - __init/__exit are not for prototype decls (ldlm_init/exit)
5540 - x86-64 compile warning fixes
5541 - fix gateway LMC keyword conflict (2318)
5542 - fix MDS lock inversions in getattr/reint paths (1844)
5543 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5544 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5545 - DLM race condition prevented some lock evictions (2328)
5546 - ENOMEM detection and retry on socknal sends (2230)
5547 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5548 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5549 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5550 - better error messages when a client is rejected during recovery (1505)
5551 - avoid cancelling locks which were never granted, after failure (2330)
5552 - fix i_sem/journal inversion in mds_client_add (2333)
5553 - fix truncate/getattr lock cycle deadlock (2334)
5554 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5555 - fix two rare exit paths which could leak an l_lock() ref (2321)
5556 - fencepost error in MDS/OST orphan recovery (2226)
5557 - make log record alignment 8 bytes (1988)
5558 - lstripe now fails when requested offset > ost_count (2237)
5559 - ensure that all kernel series have a complete list.h (1607)
5560 - fix crashes in special-file operations (2316)
5561 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5563 - allow configurable automake binary, for testing new versions
5564 - small update to the lfs documentation
5566 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5568 * fix negative export reference count in fsfilt_sync (2312)
5570 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5571 * release candidate 0.9.1
5573 - orphans are moved into the PENDING directory for possible recovery
5574 - replayed opens now open by fid for orphan/rename safety (1042)
5575 - last close of an orphan inode generates a transno (683)
5576 - chdir() and mount() now pin the directory entry (1020)
5577 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5578 - discard very old requests without processing them (1502)
5579 - remove some common, well-understood CERRORs (1505)
5580 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5581 - clear "grant" flags in OST replies until OST grant code lands (1644)
5582 - fix read performance by not clobbering i_blksize on client (1598)
5583 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5584 - make LRU size a /proc tunable, clears locks when reduced (707)
5585 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5586 - server locks take references on exports now (1558)
5587 - build fixes for 2.4.20-rh trees (1663)
5588 - return an error from lov_create if all OSCs are inactive (1751)
5589 - fix import levels when a reconnect happens without a timeout (1597)
5590 - exit early from mds_open if we get a lookup error (1749)
5591 - partial page read at EOF wouldn't wait for disk before sending (1642)
5592 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5593 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5594 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5595 - write/truncate lock inversion (1639)
5596 - Don't auto-load obdclass, portals modules during cleanup (1495)
5597 - fix timestamps from jumping to "now" (1763)
5598 - extra journal assertions (1648)
5599 - add an extra multiunlink test (1771)
5600 - fix read_record/write_record API (1776)
5601 - fix leak of offset_extent, possible incorrect i_size later (1772)
5602 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5603 - Debugging for the kqswnal_get_idle_tx problems (1820)
5604 - Allow recovery to be attempted multiple times (1536)
5605 - Write out MDS last_rcvd file after it is first created (1600)
5606 - Fix tx_descriptor leak in failed transmit situations (1827)
5607 - ext3 journaling fixes for assertion failure after IO error (1871)
5608 - class_export_put() on freed export after completion AST error (1896)
5609 - Fix revalidate looping in VFS (1322)
5610 - Don't access a freed export during MDS_REINT timeout (1521)
5611 - Add open-unlink recovery support on the MDS (1673,1764)
5612 - Return an error if no MDS data was read from last_rcvd (1946)
5613 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5614 - Don't setup a disk device that doesn't match exported UUID (317)
5615 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5616 - avoid committing NULL handle in force close
5617 - local.sh is now a one-stripe LOV configuration
5618 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5619 - fix ext3 htree duplicate directory entry corruption (1516)
5620 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5621 - update to new LOV EA format (2097)
5622 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5624 2003-06-15 Phil Schwan <phil@clusterfs.com>
5627 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5628 - per-import/export recovery handling (958, 931, 959)
5629 - multiple last-rcvd slots, for serving multiple FSes (949)
5630 - connections are again shared between multiple imp/exports (963, 964)
5631 - "umount -f" would hang if any requests needed to be sent (393, 978)
5632 - avoid pinning large req buffer by copying for queued messages (989)
5633 - add "uuid" to "lctl device" command to help upcalls (991)
5634 - "open" RPCs with transnos would confuse recovery counters (1037)
5635 - do proper endian conversion of all wire messages (288, 340, 891)
5636 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5637 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5638 - fix ptlrpc_connection leak in target_handle_connect (1174)
5639 - fix import refcounting bug in OST and MDS cleanup (1134)
5640 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5641 - fix very unlikely obd_types race condition (501)
5642 - remove osc_open hack for echo_client (1187)
5643 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5644 - a failure in read_inode2 leads to deadlock (1139)
5645 - cancel ack-locks as soon as transaction is committed (1072)
5646 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5647 - make sure to commitrw after any preprw to avoid deadlock (1162)
5648 - failing to execute a file in a lustre FS would lock inode (1203)
5649 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5650 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5651 - obd_uuid2tgt would walk off the end of the list (1255)
5652 - on IA64 the osc would give portals incorrect bulk size (1258)
5653 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5654 - fix lock inversion caused by new llite matching code (1282)
5655 - limit the number of dirty pages on a client to 10MB (1286)
5656 - timed out locks were not being corrected cancelled (1289)
5657 - fix O_DIRECT above 4GB on IA-32 (1292)
5658 * major user-visible changes
5659 - fail out/fail over policy now controlled by the upcall (993)
5661 - add OBD_PING to check server availability and failure (954)
5662 - lustre messages are now sent in sending host order (288, 340, 891)
5663 - add eadatalen to MDS getattr reply (340)
5664 - OST read replies may contain second buffer, with per-page status (593)
5666 2003-03-11 Phil Schwan <phil@clusterfs.com>
5669 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5670 - failed open()s could cause deadlock; fixed (867, 869)
5671 - stop cancelling OST locks when files are closed (481)
5672 - overlapping XID spaces caused network corruption (851, 853)
5673 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5674 - setattr_raw added, to do single-RPC, server-side setattrs
5675 - lmc/lconf syntax change for OST UUIDs
5676 - fix crashy race condition between ptlrpc_free_req and osc_close
5677 - don't use request in mdc_enqueue if we hit a timeout (889)
5678 - don't set the inode i_size for regular files from the MDS (896)
5679 - handle out of order completion AST (842)
5680 - don't LBUG if a lock request times out after receiving AST (913)
5681 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5682 - if a bad lock AST arrives, send an error instead of dropping entirely
5683 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5684 - fix leak in bulk IO when only partially completed (899, 900, 926)
5685 - fix O_DIRECT for ia64 (55)
5686 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5687 - C-z after timeout could hang a process forever; fixed (977)
5689 - client-side I/O cache (678, 924, 929, 941, 970)
5691 - READPAGE and SETATTRs which don't take server-side locks get
5694 2003-02-11 Phil Schwan <phil@clusterfs.com>
5697 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5698 - Fix fsfilt_extN_readpage() to read a full page of directory
5699 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5700 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5701 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5702 - Most intent operations were removed, in favour of a new RPC mode
5703 that does a single RPC to the server and bypasses most of the VFS
5704 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5705 - Aggressively cancel local locks on DLM servers
5706 - mds_reint_unlink sends EA to the client if it's the last nlink.
5707 client uses that EA to unlink OST objects.
5708 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5709 - recursive symlinks were fixed (439)
5710 - fixed NULL deref in DEBUG_REQ
5711 - filter_update_lastobjid no longer calls sync, which annoyed extN
5712 - fixed multi-client small-writes to a single file problem (445)
5713 - fixed mtime updates during file writes (607)
5714 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5715 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5716 - fixed Linux OST size reporting problem (444, 656)
5717 - OST now updates object mtime with writes or setattr (607, 619)
5718 - client verifies file size before zeroing page past EOF (445)
5719 - OST now writes last allocated objid to disk with allocation (108)
5720 - LOV on echo now works (409)
5722 - mds_reint_unlink sends a new buffer, with the EA included. this
5723 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5724 set if a regular file was being unlinked, and it was the last link
5725 - use PtlGet from the target for bulk writes (315)
5726 - OST now updates object mtime with writes or setattr (607, 619)
5727 - LDLM now has a grant-time callback to revalidate locked items, if
5729 - Many MDS operations were reorganized to combat race conditions
5731 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5732 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5734 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5737 - Fully reactivate OST imports after reconnection (512, others)
5738 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5739 - More graceful error handling for truncating on dead OST (515)
5740 - Don't error out unless we're actually accessing dead stripes (474)
5741 - Fix garbage sizes when stripes are missing (410)
5742 - LRU counters were broken, causing constant lock purge (433, 432)
5743 - garbage on read from stripes with failed OSTs (441)
5744 - mark OSCs as active before reconnecting during recovery (438)
5745 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5746 - lfind did not preserve OST order in output (443)
5747 - symlinks cause hung clients, incorrect data (439)
5748 - stop dereferencing request after dropping refcount (457)
5749 - don't LASSERT(spin_is_locked) on non-SMP (455)
5750 - fixes for many rename() bugs
5751 - fstat didn't correctly synchronize attributes (399)
5752 - server must handle lock cancellation during blocking AST prep (487)
5753 - bulk descriptors were free()d too soon (511)
5754 - fix paths in lconf, which would load incorrect modules (451, 507)
5755 - fix confusing lconf 'host not found' error message (386)
5756 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5757 - fix race condition in mdc_blocking_ast() for inode access (526)
5758 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5759 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5760 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5761 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5762 - fix lstripe in several ways (kernel side) (527)
5763 - fix request leak in ldlm_cli_enqueue (262)
5764 - incorrect OSC was marked inactive after OST failure
5765 - call mds_fs_cleanup before unmounting filesystem (524)
5766 - fix races between taking ns_lock and ldlm_lock_change_resource
5767 - fix races updating LOV export open file list
5768 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5769 - fix recovery NULL deref in ldlm_cli_cancel_unused
5770 - fix some DLM races by using new hash table for lock handles (419)
5771 - permit the client to specify desired inodes, at replay
5772 - duplicate requests when we queue them for replay reintegration
5773 - fix last_rcvd offset calculation
5774 - sync after each recovered transaction, so we always make progress
5775 - never, not always, ERESTART requests without transnos
5776 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5778 - skip replay if the MDS says that the client is already connected
5779 - don't check for a recovery-enabled export to match lctl's UUID
5780 - don't INC_USE_COUNT for phantom exports
5781 - don't crash when cleaning up phantom exports (567)
5782 - don't double-finish or set replay data for errored mdc_open requests
5783 - abort requests when they time out, so we don't get old replies
5784 - send/receive replies for AST messages again
5785 - if the client says that it doesn't have the lock, cancel it on the
5787 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5788 mark it as destroyed, it will all work out when decref is called
5789 - fix module use counts (22, 581)
5791 - ASTs now expect a reply (server cancels lock on error reply)
5793 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5796 - fix many simultaneous client startup (392)
5797 - fix dentry->d_it clobbering
5798 - credentials weren't being shipped for readdir/getattr operations
5799 - remove invalid assertions triggered during some concurrent MD
5801 - proper Lustre versions added (336, 389)
5802 - fix memory leak for create error case (398)
5803 - fix LOV locking bug that would get cli/srv out of sync
5804 - fix echo client over LOV (409)
5805 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5806 - fix double-O_EXCL intent crash (424)
5807 - avoid sending multiple lock CANCELs (352)
5809 - MDS can do multi-client recovery (modulo bugs in new code)
5811 - many updates, edits, cleanups
5813 2002-11-18 Phil Schwan <phil@clusterfs.com>
5816 - fix null d_it dereference (346)
5817 - fix full OST/dbench hang (333)
5818 - fix permission problem with file removal (286)
5819 - fix removal of OSCs from LOV when they fail
5820 - fix NULL deref during bulk timeout (214)
5821 - fix problems related to multiple filesystems on one MDS (241)
5822 - fixed serious subtle metadata locking bugs
5823 - free locks on clients when inodes are removed due to memory
5825 - fix inode pointer in lock data (285)
5826 - partial support for multiple MDS on a single host (241)
5827 - data locks weren't cancelled at clear_inode time (290, 311)
5828 - intent locks could lead to unbounded lock growth (205)
5829 - added a maximum lock count, an LRU list, and a flusher
5830 - fix multiple rename (365)
5831 - properly abstracted the echo client
5832 - OSC locked 1 byte too many; fixed
5833 - rewrote brw callback code:
5834 - fixed recovery bugs related to LOVs (306)
5835 - fixed too-many-pages-in-one-write crash (191)
5836 - fixed (again) crash in sync_io_timeout (214)
5837 - probably fixed callback-related race (385)
5839 - Add capability to MDS protocol
5840 - LDLM cancellations and callbacks on different portals
5842 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5845 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5846 - defer open object destruction to close time (601981, 138)
5847 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5848 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5849 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5850 - don't call obd_setattr in ll_file_release for destroyed objects
5851 * protocol change to lustre_msg: move |version| and add |flags|
5852 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5853 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5854 * added replay of create, unlink, link and rename operations during
5855 MDS failover; recovery should be much more robust now
5856 * remove failed OSCs from LOVs (only lov_create uses this so far)
5857 * the lustre-HOWTO was brought (more) up to date (582544)
5859 2002-10-23 Phil Schwan <phil@clusterfs.com>
5862 - in-use dentries weren't being reused properly (617851)
5863 - prevent multiple LDLM setup (599178)
5864 - fix LOV size calculations for truncate (617853)
5865 - fix client handling of MDS intent errors (POSIX)
5866 - fix permission bug in lovstripe.c test (624321)
5867 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5868 - truncate past end of file could corrupt data
5869 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5870 - a race in recovery could return ETIMEDOUT to apps (623947)
5871 - building outside the source directory was fixed
5872 * the lustre-HOWTO was brought (more) up to date (582544)
5873 * major progress was made on recovery functionality
5875 2002-10-10 Phil Schwan <phil@clusterfs.com>
5878 - recovery deadlock fix
5879 - rm -rf causes LBUG fix (617817)
5880 - file open by multiple tasks fix (618962)
5881 - directory permissions bugs (602707 and 620007)
5882 - journal_stop fixed with locking (611313)
5883 - O_APPEND failures resolved (618273, perhaps 614459)
5884 - lconf PATH fix (619770)
5885 - IA64 build fix (621450)
5886 - RPC buffer sizes scale with amount of memory
5888 2002-10-01 Phil Schwan <phil@clusterfs.com>
5891 - locks would be cancelled without throwing away data pages,
5892 resulting in inconsistent data (605627)
5893 - inode attributes were not always being refreshed (605627, 612449)
5894 - lconf now continues to cleanup after lctl reports an error
5895 - MDS now enforces user permissions (602707)
5896 - lprocfs cleanup fixed, but not yet enabled (614157)
5897 - fixed infinite server hang, should a client not respond to an AST
5898 - avoid going into recovery if user calls readlink() with a buffer
5899 that's too small (613941)
5900 - AST RPCs no longer require replies (614867) -- this may be changed
5901 - don't crash server if client sends an IOV that's too big (611336)
5902 - fixed lock conversion deadlock (611892)
5903 - fixed the following of symlinks (614622)
5904 * recovery: the server can remove locks from a client that dies, other
5905 clients can make progress
5906 * more extN patch fixes
5907 * compile-time configurable ptlrpc buffer allocations
5909 - collaborative read cache document
5910 - Lustre Lite Performance CDR document-in-progress
5912 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5915 - fix typo in patch-2.4.18
5917 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5920 - clear ptlrpc request each time in handle_incoming_request()
5921 - unlink of files now destroys the object on the OST
5923 2002-09-19 Peter Braam <braam@clusterfs.com>
5925 * add hard link support
5926 * change obdfile creation method
5927 * kernel patch changed
5929 2002-09-19 Peter Braam <braam@clusterfs.com>
5932 - stack overflow bug in extN fixed
5934 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5936 * documentation updates
5937 - add man pages for config tools
5938 - update tests/README to describe testing with new config tools
5939 - finish metadata API descriptions
5940 * bug fixes and cleanups
5941 - statfs workaround for 16TB limit
5942 - LOV stripe allocation improved, can stripe on subset of OSTs
5943 - LOV file size/IO offset was wrong for files > 4GB in size
5944 - object EA data was being dropped, caused files to be unreadable
5945 - memory overflow with non-LOV OST caused memory corruption
5946 - fixed regression tests to work with new config tools, obdfilter
5947 - fixed bug when directory size became larger than 1 block
5948 - fixed bug (for single client case) when PWD was deleted
5949 - invalidate local directory pages when doing intent-based ops
5950 - avoid LDLM oops when lock callback contained bad data
5952 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5954 * documentation updates
5955 * bug fixes and cleanups
5956 - configuration tools
5959 - 64-bit compile warnings
5960 - 64-bit internal statfs data
5962 * test_brw on persistent OST devices
5964 * lprocfs (disabled)
5966 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5968 * documentation updates
5969 * bug fixes and cleanups
5970 * configuration tools
5972 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5975 * many small fixes to 0_5_4
5976 * io/network handling
5977 * thinkos in MDS operations
5979 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5982 * crucial basic fixes to 0.5.3
5983 * IOR, Iozone work over Elan
5986 2002-08-07 Phil Schwan <phil@clusterfs.com>
5987 * version 0_5_3, our first alpha
5988 * we use the new Portals iovs
5989 * documentation updates
5990 * bug fixes and cleanups
5991 * small changes in the DLM wire protocol
5993 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5994 * version 0_5_1 with some initial stability,
5995 * locking on MD and file I/O.
5996 * documentation updates
5997 * several bug fixes since 0.5.0
5998 * small changes in wire protocol
6000 2002-07-18 Phil Schwan <phil@clusterfs.com>
6002 * delivered as Lustre Light Alpha
6003 * fixed a crash after handling invalid MDS requests
6004 * fixed directory pages for architectures with non-4k pages sizes
6006 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6007 * release version v0_4_4
6008 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6010 * Fixes a number of interruption problems with OST operations.
6011 * Update documentation for portals header changes
6012 * Move all wire protocol structs/defines to lustre_idl.h
6013 * Fixes symlink length bug.
6014 * Add tcpdump to repository.
6016 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6017 * release version v0_4_3
6018 * Fixes statfs for inodes on extN.
6019 * Fixes bug in runtests which would delete /etc/hosts.
6020 * Use 64-bit object IDs wherever possible (not into VFS though)
6021 Remove ost_get_info, which is unused by lustre, and out of date.
6023 2002-07-03 Peter Braam <braam@clusterfs.com>
6024 * release version v0_4_2 Fixes a lookup error (type not passed)
6025 * move forward to head of Portals
6026 * move forward to latest Lustre kernel
6028 2002-06-25 Peter Braam <braam@clusterfs.com>
6029 * release version v0_4_1. Hopefully stable on single node use.