1 tbd Sun Microsystems, Inc.
4 2.6.16.60-0.39.3 (SLES 10),
5 2.6.18-128.1.14.el5 (RHEL 5),
6 2.6.22.14 vanilla (kernel.org).
7 * Client support for unpatched kernels:
8 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
9 2.6.16 - 2.6.21 vanilla (kernel.org)
10 * Recommended e2fsprogs version: 1.41.6.sun1
11 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
12 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
13 removed cwd "./" (refer to Bugzilla 14399).
14 * File join has been disabled in this release, refer to Bugzilla 16929.
16 Severity : enhancement
18 Description: Enhancement for lfs(1) command to use numeric uid/gid.
22 Description: Improved cleanup procedure for conf-sanity test_46a.
26 Description: Do not start bulk transfer, if client tried to reconnect.
27 Details : Do not start bulk transfer, if client tried to reconnect,
28 but failed due to active rpcs. Check for exp_abort_active_req
33 Description: Prevent inconsistences between linux and lustre mount structures.
34 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
35 Make the sleep interruptible.
37 Severity : enhancement
39 Description: provide server to client comms path
40 Details : server to client communications path via new LDLM_SET_INFO rpc
44 Description: truncate starts GFP_FS allocation under transaction causing deadlock
45 Details : ldiskfs_truncate calls grab_cache_page which may start page
46 allocation under an open transaction. This may lead to
47 calling prune_icache with consequent lustre reentrance.
51 Description: send correctly lsm on open replay
52 Details : MDS is trust to LSM size on replay open, but client can set wrong size
57 Description: enable client interface failover
58 Details : When a child reconnects from another NID, properly update export
59 nid hash position and ldlm reverse import.
61 Severity : enhancement
63 Description: Communicate OST degraded/readonly state via statfs to MDS
64 Details : Flags in the statfs returned from OSTs indicate whether the
65 OST is in a degraded RAID state, or if the filesystem has
66 turned read-only after a filesystem error is detected.
71 Description: don't panic if EPROTO was hit when reading symlink
72 Details : correctly handling request reference in error cases.
74 Severity : enhancement
76 Description: Add LustreNetLink, a kernel-userspace communcation path.
78 Severity : enhancement
80 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
85 Description: access to llog context before init.
86 Details : move handling CATALOGS file at osc layer and forbid access to llog
90 Frequency : rare, only when client is writing to OST on local node
92 Description: Avoid deadlock for local client writes
93 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
94 memory freeing context. This allows OST threads to set the
95 PF_MEMALLOC flag on task structures in order to allocate memory
96 from reserved pools and complete IO.
97 Use GFP_HIGHUSER for OST allocations for non-local client writes,
98 so that the OST threads generate memory pressure and allow
99 inactive pages to be reclaimed.
101 Severity : enhancement
103 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
108 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
109 Details : move ldlm namespace creation in setup phase to avoid grab
110 _lprocfs_lock with cli_sem held.
114 Description: Temporarily disable grant shrink.
115 Details : Disable the feature for debugging.
118 Frequency : only when formatting test filesystems
120 Description: Unable to run several mkfs.lustre on loop devices at the same time
121 Details : mkfs.lustre returns error 256 on the concurrent loop devices
122 formatting. The solution is to proper handle the error.
124 Severity : enhancement
126 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
128 Severity : enhancement
130 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
135 Description: resolve race between obd_disconnect and class_disconnect_exports
136 Details : if obd_disconnect will be called to already disconnected export he
137 forget release one reference and osc module can't unloaded.
139 Severity : enhancement
141 Description: Allow tuning service thread via /proc
142 Details : For each service a new
143 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
144 created that can be used to set min/max thread counts, and get the
145 current number of running threads.
150 Descriptoin: don't return error if have partially created objects for file.
151 Details : lov_update_create_set uses set->set_success as index for created
152 objects, so if some requests will be failed, they will have hole
153 at end of array and qos_shrink_lsm for allocate correct lsm.
155 Severity : enhancement
157 Description: Update OFED support to 1.4.1
162 Description: Reduce small size read RPC
163 Details : Set read-ahead limit for every file and only do read-ahead when
164 available read-ahead pages are bigger than 1M to avoid small size
167 Severity : enhancement
169 Description: move AT tunable parameters for more consistent usage
170 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
172 Severity : enhancement
174 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
175 Details : allow skip disconnected ost for send statfs request and hide error
181 Description: fix racy locking of mballoc block bitmaps causing BUG
182 Details : The locking of the mballoc buddy bitmap and the in-memory
183 block bitmap was using two different spin locks in some
184 cases. This made it possible to incorrectly access the
185 mballoc bitmap while another process was modifying it,
186 causing a sanity assertion to fail. While no on-disk corruption
187 was reported, there was some risk of this happening.
190 Frequency : rare, on llog test 6
192 Descriptoin: don't allow connect to already connected import
193 Details : allowing connect to already connected import is hide connecting problem.
198 Descriptoin: allow kill process which wait statahead result
199 Details : for some reasons 'ls' can stick in waiting result from statahead,
200 in this case need way for kill this process.
202 Severity : enhancement
204 Description: Add state history info file, enhance import info file
205 Details : Track import connection state changes in a new osc/mdc proc file;
206 add overview-type data to the osc/mdc import proc file.
208 Severity : enhancement
210 Description: MDS create should not wait for statfs RPC while holding DLM lock.
212 Severity : enhancement
214 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
219 Descriptoin: OST grant too much space to client even there are not enough space.
220 Details : Client will shrink its grant cache to OST if there are no write
221 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
222 this grant cache if there are already not enough avaible space
223 (left_space < total_clients * 32M).
226 Frequency : start MDS on uncleanly shutdowned MDS device
228 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
229 Details : stay in waiting mds<>ost recovery finished produce random bugs
230 due race between two ll_sync thread for one lov target. send
231 ACTIVATE event only if connect realy finished and import have
235 Frequency : rare, connect and disconnect target at same time
237 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
238 Details : don't call obd_disconnect under lov_lock. this long time
239 operation and can block ptlrpcd which answer to connect request.
244 Descriptoin: don't lose wakeup for imp_recovery_waitq
245 Details : recover_import_no_retry or invalidate_import and import_close can
246 both sleep on imp_recovery_waitq, but we was send only one wakeup
250 Frequency : always with long access acl
252 Descriptoin: mds can't pack reply with long acl.
253 Details : mds don't control size of acl but they limited by reint/getattr
256 Severity : enhancement
258 Description: Update to SLES10 kernel-2.6.16.60-0.33.
260 Severity : enhancement
262 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
265 Frequency : start MDS on uncleanly shutdowned MDS device
267 Descriptoin: aborting recovery hang on MDS
268 Details : don't throttle destroy RPCs for the MDT.
271 Frequency : on remount
273 Description: external journal device not working after the remount
274 Details : clear dev_rdonly flag for external journal devices in
280 Description: shutdown vs evict race
281 Details : client_disconnect_export vs connect request race.
282 if client will evicted at this time - we start invalidate
283 thread without referece to import and import can be freed
287 Frequency : rare, need acl's on inode.
289 Description: client can't handle ost additional correctly
290 Details : if ost was added after client connected to mds client can have
291 hit lnet_try_match_md ... to big messages to wide striped files.
292 in this case need teach client to handle config events about add
293 lov target and update client max ea size at that event.
295 Severity : enhancement
297 Description: Changelogs
298 Details : Changelogs are a lightweight mechanism to track filesystem
299 metadata and namespace changes. The changelog is recorded
300 permanently on the MDTs, and is periodically "consumed" / purged
301 when records are no longer needed.
303 Severity : enhancement
305 Description: compact fld format with extents
306 Details : Store range of seq rather than every seq in FLD. Seq
307 controller update FLD rather than clients. In Case of CMD, mdt0
308 has FLD, all other metadata server act as non persistent proxy
309 for FLD queries and cache fld entries in fld cache.
314 Description: don't skip ost target if they assigned to file
315 Details : Drop slow OSCs if we can, but not for requested start idx.
316 This means "if OSC is slow and it is not the requested
317 start OST, then it can be skipped, otherwise skip it only
318 if it is inactive/recovering/out-of-space.
322 Description: more cleanup in mds_lov
323 Details : not send LOV EA under replay, we can't know about they size at this
324 time. Don't allow client connect to mds before any ost connected,
325 for avoid problems with LOV EA size and returning EIO to client.
327 Severity : enhancement
329 Description: Interoperability at server side (Disk interoperability)
331 Severity : enhancement
333 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
335 Severity : enhancement
337 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
339 Severity : enhancement
341 Description: New client IO stack (CLIO).
343 Severity : enhancement
345 Description: Commit on sharing. Eliminate inter-client dependencies between
346 uncommitted transactions by doing transaction commits.
347 Thereby clients may recovery independently.
350 Frequency : Create a symlink file with a very long name
352 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
353 Details : If there is no extra space in the request for early cancels,
354 ldlm_req_handles_avail() returns 0 instead of a negative value.
356 Severity : enhancement
358 Description: Add /proc entry for import status
359 Details : The mdc, osc, and mgc import directories now have
360 an import directory that contains useful import data for debugging
363 Severity : enhancement
365 Description: Re-disable certain /proc logging
366 Details : Enable and disable client's offset_stats, extents_stats and
367 extents_stats_per_process stats logging on the fly.
370 Frequency : Only on FC kernels 2.6.22+
372 Description: oops in statahead
373 Details : Do not drop reference count for the dentry from VFS when lookup,
374 VFS will do that by itself.
376 Severity : enhancement
378 Description: Generic /proc file permissions
379 Details : Set /Proc file permissions in a more generic way to enable non-
380 root users operate on some /proc files.
384 Description: Hitting mdc_commit_close() ASSERTION
385 Details : Properly handle request reference release in
386 ll_release_openhandle().
390 Description: quota recovery deadlock during mds failover
391 Details : This patch includes att18982, att18236, att18237 in bz14840.
393 1. fix osts hang when mds does failover with quotaon
394 2. prevent watchdog storm when osts threads wait for the
399 Frequency : only patchless client
400 Description: add workaround for race between add/remove dentry from hash
402 Severity : enhancement
404 Description: Allow OST glimpses to return PW locks
408 Description: LBUG when llog conf file is full
409 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
413 Description: Prevent import from entering FULL state when server in recovery
417 Description: service mount cannot take device name with ":"
418 Details : Only when device name contains ":/" will mount treat it as
424 Description: replace ptlrpcd with the statahead thread to interpret the async
425 statahead RPC callback
429 Frequency : on recovery
430 Description: I/O failures after umount during fail back
431 Details : if client reconnected to restarted server we need join to recovery
432 instead of find server handler is changed and process self
433 eviction with cancel all locks.
435 Severity : enhancement
437 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
439 Severity : enhancement
441 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
443 Severity : enhancement
445 Description: Upcall on Lustre log has been dumped
446 Details : Allow for a user mode script to be called once a Lustre log has
447 been dumped. It passes the filename of the dumped log to the
448 script, the location of the script can be specified via
449 /proc/sys/lnet/debug_log_upcall.
454 Description: avoid idr_remove called for id which is not allocated.
455 Details : Move assigment s_dev for clustered nfs to end of initialization,
456 to avoid problem with error handling.
461 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
462 Details : When connection is reused this not moved from CONN_UNUSED_HASH
463 into CONN_USED_HASH and this prodice warning when put connection
464 again in unused hash.
467 Severity : enhancement
469 Description: File striping can now be set to use an arbitrary pool of OSTs.
471 Severity : enhancement
473 Description: Export bytes_read/bytes_write count on OSC/OST.
477 Description: Early reply size mismatch, MGC loses connection
478 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
479 the connect flags are properly negotiated.
484 Description: quotas are not honored with O_DIRECT
485 Details : all writes with the flag O_DIRECT will use grants which leads to
486 this problem. Now using OBD_BRW_SYNC to guard this.
490 Description: add quota statistics
491 Details : 1. sort out quota proc entries and proc code.
492 2. add quota statistics
494 Severity : enhancement
496 Description: enable quota support for HEAD.
500 Description: Properly propagate oinfo flags from lov to osc for statfs
501 Details : restore missing copy oi_flags to lov requests.
503 Severity : enhancement
505 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
506 : l_getgroups(8), lst(8), routerstat(8)
507 Details : included man pages for llobdstat(8), llstat(8),
508 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
510 Severity : enhancement
512 Description: configure's --enable-quota should check the
513 : kernel .config for CONFIG_QUOTA
514 Details : configure is terminated if --enable-quota is passed but
515 : no quota support is in kernel
519 Description: Remove portals compatibility
520 Details : Remove portals compatibility, not interoperable with releases
525 Description: Resolve device initialization race
526 Details : Prevent proc handler from accessing devices added to the
527 obd_devs array but yet be intialized.
529 Severity : enhancement
531 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
533 Severity : enhancement
535 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
540 Description: Using wrong pointer in osc_brw_prep_request
541 Details : Access to array[-1] can produce panic if kernel compiled with
542 CONFIG_PAGE_ALLOC enabled
546 Description: Client runs out of low memory
547 Details : Consider only lowmem when counting initial number of llap pages
551 Description: *optional* service tags registration
552 Details : if the "service tags" package is installed on a Lustre node
553 When the filesystem is mounted, a local-node service tag will
554 be created. See http://inventory.sun.com/ for more information
555 about the Service Tags asset management system.
559 Description: Kernel BUG tries to release flock
560 Details : Lustre does not destroy flock lock before last reference goes
561 away. So always drop flock locks when client is evicted and
562 perform unlock regardless of successfulness of speaking to MDS.
566 Description: add refcount protection for osc callbacks, avoid panic on shutdown
570 Description: sanity test 65a fails if stripecount of -1 is set
571 Details : handle -1 striping on filesystem in ll_dirstripe_verify
576 Description: ASSERTION(CheckWriteback(page,cmd)) failed
577 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
580 Severity : enhancement
582 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
586 Description: do not process already freed flock
587 Details : flock can possibly be freed by another thread before it reaches
588 to ldlm_flock_completion_ast.
592 Description: LBUG during stress test
593 Details : Need properly lock accesses the flock deadlock detection list.
597 Description: oops in page fault handler
598 Details : kernel page fault handler can return two special 'pages' in error
599 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
603 Description: timeout with invalidate import.
604 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
605 should be handled by ptlrpcd. This produce long age waiting and
606 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
608 Severity : enhancement
610 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
614 Description: enable MGS and MDT services start separately
615 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
616 a co-located MGS without starting the MGS, which is a complement
617 to 'nosvc' mount option.
621 Frequency : after recovery
622 Description: precreate to many object's after del orphan.
623 Details : del orphan st in oscc last_id == next_id and this triger growing
624 count of precreated objects. Set flag LOW to skip increase count
625 of precreated objects.
629 Frequency : rare, on clear nid stats
630 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
631 Details : when clean nid stats sometimes try destroy live entry,
632 and this produce panic in free.
636 Description: Stack overflow during MDS log replay
637 ease stack pressure by using a thread dealing llog_process.
641 Description: wait until IO finished before start new when do lock cancel.
642 Details : VM protocol want old IO finished before start new, in this case
643 need wait until PG_writeback is cleared until check dirty flag and
644 call writepages in lock cancel callback.
646 Severity : enhancement
648 Description: using special macro for print time and cleanup in includes.
652 Description: mds_mfd_close() ASSERTION(rc == 0)
653 Details : In mds_mfd_close(), we need protect inode's writecount change
654 within its orphan write semaphore to prevent possible races.
658 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
662 Frequency : rare, on shutdown ost
663 Description: don't hit live lock with umount ost.
664 Details : shrink_dcache_parent can be in long loop with destroy dentries,
665 use shrink_dcache_sb instead.
669 Description: don't panic with use echo client
670 Details : echo client pass NULL as client nid pointer and this produce null
675 Description: fix build on ppc32
676 Details : compile code with -m64 flag produce wrong object file for ppc32.
680 Description: add message levels for liblustreapi
684 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
685 Details : Move the 'good_osts' check before the 'total_bavail' check. This
686 will result in an -EAGAIN and in the exit call path we call
687 alloc_rr() which will with increasing aggressiveness attempt to
688 aquire precreated objects on the minimum number of required OSCs.
692 Description: Use old size assignment to avoid deadlock
693 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
694 the scheduling while holding a spinlock. We do not need locking
695 for size in ll_update_inode() because size is only updated from
696 the MDS for directories or files without objects, so there is no
697 other place to do the update, and concurrent access to such inodes
698 are protected by the inode lock.
702 Description: resolve "_IOWR redefined" build error on SLES10
706 Description: dump the memory debugging after all modules are unloaded to
707 suppress false negative in conf_sanity test 39
709 Severity : enhancement
711 Description: build kernel-ib packages for OFED 1.3 in our release cycle
716 Description: fix SLES kernel versioning
717 Details : the kernel version for our SLES 10 kernel did not include a "-"
718 before the "smp" at the end. while this was not a problem in
719 general, it did mean that software trying to use the kernel
720 version to try to detect a vendor specific kernel would fail.
721 this was most evident by the OFED build scripts.
725 Description: Don't update lov_desc members until making sure they are valid
726 Details : When updating lov_desc members via proc fs, need fix their
727 validities before doing the real update.
731 Description: don't put request into delay list while invalidate in flight.
732 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
733 import in flight. this produce timeout for invalidate and sometimes
734 can cause stale data.
738 Frequency : on ppc only
739 Description: not convert ost objects for directory because it's not exist.
740 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
744 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
746 Severity : enhancement
748 Description: Update to support 2.6.22.14 vanilla kernel.
752 Frequency : rare, on recovery
753 Description: read procfs can produce deadlock in some situation
754 Details : Holding lprocfs lock which send rpc can produce block for destroy
755 obd objects and this also block reconnect with -EALREADY. This
756 isn't fix all lprocfs bugs - but make it rare.
758 Severity : enhancement
760 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
763 Frequency : frequent on X2 node
765 Description: mdc_set_open_replay_data LBUG
766 Details : Set replay data for requests that are eligible for replay.
770 Description: lustre_mgs: operation 101 on unconnected MGS
771 Details : When MGC is disconnected from MGS long enough, MGS will evict the
772 MGC, and late on MGC cannot successfully connect to MGS and a lot
773 of the error messages complaining that MGS is not connected.
777 Frequency : on network error
778 Description: panic with double free request if network error
779 Details : mdc_finish_enqueue is finish request if any network error ocuring,
780 but it's true only for synchronus enqueue, for async enqueue
781 (via ptlrpcd) this incorrect and ptlrpcd want finish request
784 Severity : enhancement
786 Description: client-side metadata stat-ahead during readdir(directory readahead)
787 Details : perform client-side metadata stat-ahead when the client detects
788 readdir and sequential stat of dir entries therein
791 Frequency : on start mds
793 Description: Implement get_info(last_id) in obdfilter.
796 Frequency : occasional
798 Description: Correctly check stale fid, not start epoch if ost not support SOM
799 Details : open with flag O_CREATE need set old fid in op_fid3 because
800 op_fid2 was overwritten with new generated fid, but mds can answer
801 with one of these two fids and both is not stale. Setattr
802 incorrectly started an epoch and assume will be called
803 done_writing, but without SOM done_writing ever being called.
806 Frequency : rare, depends on device drivers and load
808 Description: MDS or OSS nodes crash due to stack overflow
809 Details : Code changes in 1.8.0 increased the stack usage of some functions.
810 In some cases, in conjunction with device drivers that use a lot
811 of stack the MDS (or possibly OSS) service threads could overflow
812 the stack. One change which was identified to consume additional
813 stack has been reworked to avoid the extra stack usage.
816 Frequency : occasional
818 Description: Do not fail import if osc_interpret_create gets -EAGAIN
819 Details : If osc_interpret_create got -EAGAIN it immediately exits and
820 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
821 oscc_has_objects and see OSC has no objests and call
822 oscc_internal_create to resend create request.
824 Severity : enhancement
826 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
828 Severity : enhancement
830 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
833 Frequency : very rare
835 Description: Fix replay if there is an un-replied request and open
836 Details : In some cases, older replay request will revert the
837 mcd->mcd_last_xid on MDS which is used to record the client's
840 Severity : enhancement
842 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
844 Severity : enhancement
846 Description: Add rhel5 support to HEAD.
848 Serverity : enhancement
850 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
855 Description: Don't allow skipping OSTs if index has been specified.
856 Details : Don't allow skipping OSTs if index has been specified, make
857 locking in internal create lots better.
861 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
862 Details : access to bitfield in structure is always rounded to long
863 and this produce problem with not atomic change any bit.
867 Description: Lustre make rpms failed.
868 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
869 builds Lustre rpms from tarball.
872 Frequency : rare on shutdown ost
874 Description: If llog cancel was not send before clean_exports phase, this can
875 produce deadlock in llog code.
876 Details : If llog thread has last reference to obd and call class_import_put
877 this produce deadlock because llog_cleanup_commit_master wait when
878 last llog_commit_thread exited, but this never success because was
879 called from llog_commit_thread.
883 Description: allow userland application know is lost one of stripes.
884 Details : fill lvb_blocks with error code on ost and return it to
885 application if error flag found.
889 Description: NULL lov_tgts causing MDS oops
890 Details : more safe checks for NULL lov_tgts for avoid oops.
892 Severity : enhancement
894 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
898 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
899 Details : If an OST is down the MDS will hang indefinitely in
900 obd_statfs_rqset() waiting for the statfs data. While for
901 MDS QOS usage of statfs, it should not stuck in waiting.
903 Severity : enhancement
905 Description: remote_acl support
906 Details : Support ACL-based permission check for remote user.
907 Support setfacl/getfacl for remote user with the utils
908 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
909 the system "{s,g}etfacl" utils.
911 Severity : enhancement
913 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
915 Severity : enhancement
917 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
921 Description: make mgs_setparam() handle fsname containing dash
922 Details : fsname containing a dash does not work with lctl conf_param
924 Severity : enhancement
926 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
928 Severity : enhancement
930 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
934 Frequency : rare, at shutdown
935 Description: access already free / zero obd_namespace.
936 Details : if client_disconnect_export was called without force flag set,
937 and exist connect request in flight, this can produce access to
938 NULL pointer (or already free pointer) when connect_interpret
939 store ocd flags in obd_namespace.
943 Frequency : only at startup
944 Description: not alloc memory with spinlock held.
945 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
946 if any spinlock held.
948 Severity : enhancement
950 Description: make lustre randomly fail allocating memory
951 Details : Make lustre randomly failed allocating memory for testing purpose.
953 Severity : enhancement
955 Description: lost problems with lov objid file
956 Details : Fixes some scability and access to not inited memory problems
957 in work with lov objdid file.
962 Description: lfs find does not continue on file error
963 Details : Continue other files processing when a file/dir is absent.
967 Description: Inconsistent usage of lustre_pack_reply()
968 Details : Standardize the usage of lustre_pack_reply() such that it
969 always generate a CERROR on failure.
972 Frequency : occasional
974 Description: MDS hang or stay in waiting lock
975 Details : If client receive lock with CBPENDING flag ldlm need send lock
976 cancel as separate rpc, to avoid situation when cancel request
977 can't processed due all i/o threads stay in wait lock.
981 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
985 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
989 Description: Update to SLES9 kernel-2.6.5-7.287.3.
993 Description: Update to RHEL4 latest kernel.
995 Severity : enhancement
997 Description: Build SLES10 patchless client fails
998 Details : The configure was broken by run ./configure with
999 --with-linux-obj=.... argument for patchless client. When the
1000 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1001 can't search header adequately. Use absolute path such as
1002 -I($LINUX)/include instead.
1006 Description: interrupt oig_wait produce painc on resend.
1007 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1008 private set, and this produce situation when rq_ptlrpcd_data not
1009 copyed to new allocated request and triggered LBUG on assert
1010 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1011 ptlrpcd set if request is changed and can be safety changed to use
1016 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1017 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1018 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1019 these macros and replace them with their difinitions instead.
1023 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1024 Details : Update which_patch & target file for SLES10 latest kernel.
1027 Description: Remove client patches from SLES 10 kernel.
1028 Details : This causes SLES 10 clients to behave as patchless clients
1029 even on a Lustre-patched (server) kernel.
1031 Severity : enhancement
1033 Description: self-adjustable client's lru lists
1034 Details : use adaptive algorithm for managing client cached locks lru
1035 lists according to current server load, other client's work
1036 pattern, memory activities, etc. Both, server and client
1037 side namespaces provide number of proc tunables for controlling
1040 Severity : enhancement
1042 Description: light-weight GSS support
1043 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1044 and reduce performance overhead.
1046 Severity : enhancement
1048 Description: Linux keyring support
1049 Details : Support using service of Linux keyring for Lustre GSS internal
1050 context refresh/cache mechanism.
1054 Description: Fix errors in lfs documentation
1055 Details : Fixes man pages
1059 Description: don't use GFP_* in generic Lustre code.
1060 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1064 Description: obdclass is limited by single OBD_ALLOC(idarray)
1065 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1069 Description: warnings with build patchless client with vanila 2.6.19 and up
1070 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1071 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1075 Frequency : Only for SLES
1076 Description: sanity_quota fail test_1
1077 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1078 file, which makes TSTID[2] inunique.
1082 Frequency : Only for SLES9
1083 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1084 Details : Update target/ChangeLog/which_patch .
1088 Frequency : Only for test_10 in sanity-quota.sh
1089 Description: a bug in quota test code
1090 Details : checking a test flag in a wrong place
1094 Frequency : Only for test_13 in sanity-quota.sh
1095 Description: a bug in quota test script
1096 Details : shouldn't check overfull allocated quota
1100 Description: jbd statistics
1101 Details : Port older jbd statistics patch for sles10
1103 Severity : enhancement
1105 Description: Add printing inode info into message about error in writepage.
1110 Description: reply_lock_interpret crash due to race with it and lock cancel.
1111 Details : Do not replay locks that are being cancelled. Do not reference
1112 locks by their address during replay, just by their handle.
1116 Frequency : When flocks are used.
1117 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1118 Details : Flock locks might destroy just granted lock if it could be merged
1119 with another existing flock, this is done in completion handler,
1120 so teach ldlm_cli_enquque_fini that this is a valid case for
1126 Description: Oops in read and write path when failing to allocate lock.
1127 Details : Check if lock allocation failed and return error back.
1131 Description: lstripe command fails for valid OST index
1132 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1133 instead of lov->desc.ld_active_tgt_count.
1137 Description: sanity.sh failed test 103
1138 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1143 Description: df doesn't work properly if diskfs blocksize != 4K
1144 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1148 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1149 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1151 Severity : enhancement
1153 Description: Adaptive timeouts
1154 Details : RPC timeouts adapt to changing server load and network
1155 conditions to reduce resend attempts and improve recovery time.
1159 Description: llapi_file_create() does not allow some changes
1160 Details : add llapi_file_open() that allows specifying the mode and
1161 open flags, and also returns an open file handle.
1165 Description: merge and cleanup kernel patches.
1166 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1170 Description: Add journal checksum support.(Kernel part)
1171 Details : The journal checksum feature adds two new flags i.e
1172 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1173 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1174 indicates that the commit block contains the checksum for
1175 the blocks described by the descriptor blocks. Now commit
1176 record can be sent to disk without waiting for descriptor
1177 blocks to be written to disk. This behavior is controlled
1178 using JBD2_FEATURE_ASYNC_COMMIT flag.
1182 Description: OSS needs mutliple precreate threads
1183 Details : Add ability to start more than one create thread per OSS.
1187 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1188 Details : execute lfs setstripe on client
1192 Description: mds_obd_create error creating tmp object
1193 Details : When the user sets quota on root, llog will be affected and can't
1194 create files and write files.
1197 Frequency : Always on ia64 patchless client, and possibly others.
1199 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1200 Details : This allows the patchless client to be loaded on architectures
1201 without this export.
1205 Description: RedHat Update kernel for RHEL5
1206 Details : Add the kernel config file for RHEL5.
1210 Description: RedHat Update kernel for RHEL5
1211 Details : Modify the kernel config file more closer RHEL5.
1215 Description: Build failure against Centos5 (RHEL5)
1216 Details : Define PAGE_SIZE when it isn't present.
1220 Description: test_7 on sanity-quota.sh can't be used on separate machines
1221 Details : add facet to handle it
1225 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1226 Details : with NFS, the anon dentry's parent was set to itself in
1227 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1228 corresponding dentry other than use rec->ur_name.
1230 Severity : enhancement
1232 Description: lfs setstripe enhancement
1233 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
1237 Description: enable data checksumming by default
1238 Details : enable checksum by default, allow --disable-checksum
1239 configure option and "-o nochecksum" mount option. Checksums
1240 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1244 Description: lustre support for RHEL5
1245 Details : Add support for RHEL5.
1249 Description: Client eviction due to failover config
1250 Details : after a connection loss, the lustre client should attempt to
1251 reconnect to the last active server first before trying the
1252 other potential connections.
1256 Description: when mds and osts use different quota unit(32bit and 64bit),
1257 quota will be released repeatly.
1258 Details : void sending multiple quota reqs to mds, which will keep the
1259 status between the reqs.
1263 Description: osts not allocated evenly to files
1264 Details : change the condition to increase offset_idx
1267 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1269 Description: Data corruption for OSTs that are formatted larger than 2TB
1271 Details : When generating the bio request for lustre file writes the
1272 sector number would overflow a temporary variable before being
1273 used for the IO. The data reads correctly from Lustre (which
1274 will overflow in a similar manner) but other file data or
1275 filesystem metadata may be corrupted in some cases.
1279 Description: Tune the kernel for good SCSI performance.
1280 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1281 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1286 Description: rewrite ext2-derived code in obdclass/uuid.c
1287 Details : rewrite inherited code (uuid parsing code from ext2 utils)
1288 from scratch preserving functionality.
1292 Description: "lfs find -obd UUID" prints directories
1293 Details : "lfs find -obd UUID" will return all directory names instead
1294 of just file names. It is incorrect because the directories
1295 do not reside on the OSTs.
1299 Description: use i_size_read and i_size_write in 2.6 port
1300 Details : replace inode->i_size access with i_size_read/write()
1303 Frequency : when removing large files
1305 Description: scheduling issue during removal of large Lustre files
1306 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1307 It causes scheduling issues when removing large files (17TB in the
1311 Frequency : only with liblustre clients on XT3
1313 Description: evictions taking too long
1314 Details : allow llrd to evict clients directly on OSTs
1317 Frequency : only on ppc
1319 Description: /proc/fs/lustre/devices broken on ppc
1320 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1321 We should make sure the type of 'index' is loff_t and then cast
1322 explicitly as needed below. Do not assign an explicitly cast
1326 Frequency : only for rhel5
1328 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1329 Details : Modify the target file & which_kernel.
1331 Severity : enhancement
1333 Description: omit set fsid for export NFS
1334 Details : fix set/restore device id for avoid EMFILE error and mark lustre
1335 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
1339 Frequency : Always, for kernels after 2.6.16
1340 Description: Fix warning idr_remove called for id=.. which is not allocated.
1341 Details : Last kernels save old s_dev before kill super and not allow
1342 to restore from callback - restore it before call kill_anon_super.
1346 Description: buffer overruns could theoretically occur
1347 Details : llapi_semantic_traverse() modifies the "path" argument by
1348 appending values to the end of the origin string, and a
1349 overrun may occur. Adding buffer overrun check in liblustreapi.
1353 Description: Fix error on 'ls .' at the top of the Lustre mount.
1354 Details : Don't revalidate dentry if it is a root dentry.
1358 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1359 Details : Modify vm-tunables-rhel4.patch.
1363 Description: Kernel config for 2.6.18-vanilla.
1364 Details : Modify targets/2.6-vanilla.target.in.
1365 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1366 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1367 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1368 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1372 Description: improve handling recoverable errors
1373 Details : if request processig with error which can be recoverable on server
1374 request should be resend, otherwise page released from cache and
1379 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1380 Details : Modify target file & which_patch.
1381 A flaw was found in the IA32 system call emulation provided
1382 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1383 value could be stored in the %RAX register, which could trigger an
1384 out-of-bounds system call table access. An untrusted local user
1385 could exploit this flaw to run code in the kernel
1386 (ie a root privilege escalation). (CVE-2007-4573).
1390 Description: O_DIRECT bypasses client statistics.
1391 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
1392 read_ahead_stats were updated. copied the stats section from
1393 osc_send_oap_rpc() into async_internal().
1397 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
1401 Description: change order of libsysio includes
1402 Details : '#include sysio.h' should always come before '#include xtio.h'
1406 Description: handle "serious error: objid * already exists" more gracefully
1407 Details : If LAST_ID value on disk is smaller than the objects existing in
1408 the O/0/d* directories, it indicates disk corruption and causes an
1409 LBUG(). If the object is 0-length, then we should use the existing
1410 object. This will help to avoid a full fsck in most cases.
1412 Severity : enhancement
1414 Description: adapt the lustre_config script to support the upgrade case
1415 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1416 server targets to 1.6 in parallel.
1420 Description: To avoid grant space > avaible space when the disk is almost
1421 full. Without this patch you might see the error "grant XXXX >
1422 available" or some LBUG about grant, when the disk is almost
1424 Details : In filter_check_grant, for non_grant cache write, we should
1425 check the left space by if (*left > ungranted + bytes), instead
1426 of (*left > ungranted), because only we are sure the left space
1427 is enough for another "bytes", then the ungrant space should be
1428 increase. In client, we should update cl_avail_grant only there
1429 is OBD_MD_FLGRANT in the reply.
1433 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1434 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1435 described in CVE-2007-4573. This problem could allow untrusted
1436 local users to gain root access.
1439 Frequency : when using O_DIRECT and quotas
1441 Description: Incorrect file ownership on O_DIRECT output files
1442 Details : block usage reported by 'lfs quota' does not take into account
1443 files that have been written with O_DIRECT.
1448 Description: touch file failed when fs is not full
1449 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1450 otherwise we can get ENOSP while fs is not full.
1454 Description: parallel lock callbacks
1455 Details : Instead of sending blocking and completion callbacks as separated
1456 requests, adding them to a set and sending in parallel.
1459 Frequency : only for Cray XT3
1460 Bugzilla : 12829/13455
1461 Description: Changing primary group doesn't change the group lustre assigns to
1463 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1464 overridden with the primary group provided by the group upcall,
1465 whereas the supplied fsgid can be trusted if it is in the list of
1466 supplementary groups returned by the group upcall.
1468 Severity : enhancement
1470 Description: Allow masking D_WARNING, D_ERROR messages from console
1471 Details : Console messages can now be disabled via lnet.printk.
1475 Description: User code with malformed file open parameter crashes client node
1476 Details : Before packing join_file req, all the related reference should be
1477 checked carefully in case some malformed flags cause fake
1478 join_file req on client.
1482 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1483 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1484 mode is equal to the granted mode. However, it should also check
1485 that LDLM_FL_CP_REQD is not set.
1489 Description: Heavy nfs access might result in deadlocks
1490 Details : After ELC code landed, it is now improper to enqueue any mds
1491 locks under och_sem, because enqueue might want to decide to
1492 cancel open locks for same inode we are holding och_sem for.
1496 Description: Client eviction while running blogbench
1497 Details : A lot of unlink operations with concurrent I/O can lead to a
1498 deadlock causing evictions. To address the problem, the number of
1499 oustanding OST_DESTROY requests is now throttled to
1500 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1501 callbacks are processed in priority.
1505 Description: enable ACLs on MDS by default
1506 Details : ACLs must be enabled on MDS by default.
1509 Frequency : PPC/PPC64 only
1511 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1512 Details : fix duplicated definitions between asm-ppc64/types.h and
1513 lustre_types.h on PPC.
1516 Frequency : PPC/PPC64 only
1518 Description: asm-ppc/segment.h does not exist
1519 Details : fix compile issue on PPC.
1523 Description: better handle error messages in extents code
1526 Frequency : RHEL4 only
1528 Description: mkfs is very slow on IA64/RHEL4
1529 Details : A performance regression has been discovered in the MPT Fusion
1530 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1531 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1532 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1534 Severity : enhancement
1536 Description: SNMP support enhancement
1537 Details : Adding total number of sampled request for an MDS node in snmp
1540 Severity : enhancement
1542 Description: Optimize ldlm waiting list processing for PR extent locks
1543 Details : When processing waiting list for read extent lock and meeting read
1544 lock that is same or wider to it that is not contended, skip
1545 processing rest of the list and immediatelly return current
1546 status of conflictness, since we are guaranteed there are no
1547 conflicting locks in the rest of the list.
1551 Description: Time out and refuse to reconnect
1552 Details : When the failover node is the primary node, it is possible
1553 to have two identical connections in imp_conn_list. We must
1554 compare not conn's pointers but NIDs, otherwise we can defeat
1555 connection throttling.
1559 Description: port llog fixes from b1_6 into HEAD
1560 Details : Port llog reference couting and some llog cleanups from b1_6
1561 (bug 10800) into HEAD, for protect from panic and access to already
1562 free llog structures.
1566 Description: Detect stride IO mode in read-ahead
1567 Details : When a client does stride read, read-ahead should detect that and
1568 read-ahead pages according to the detected stride pattern.
1572 Description: data checksumming impacts single node performance
1573 Details : add support for several checksum algorithm. Currently, only CRC32
1574 and Adler-32 are supported. The checksum type can be changed on
1575 the fly via /proc/fs/lustre/osc/*/checksum_type.
1579 Description: use adler32 for page checksums
1580 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1585 Description: build for x2 fails
1586 Details : fix compile issue on Cray systems.
1590 Description: Properly match for duplicate locks
1591 Details : Due to different lock order from skiplists code, we need to
1592 traverse entire list for now
1595 Frequency : only on PPC/SLES10
1597 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1598 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1599 be defined. Add a hack in mkfs_lustre.c to work around this compile
1604 Description: LASSERT on MDS when client holding flock lock dies
1605 Details : ldlm pool logic depends on number of granted locks equal to
1606 number of released locks which is not true for flock locks, so
1607 just exclude such locks from consideration.
1611 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1612 Details : Use fsfilt_sync() for both the whole filesystem sync and
1613 individual file sync to eliminate dangerous inode locking
1614 with I_LOCK that can lead to a deadlock.
1618 Description: performance in 1.6.3
1619 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1620 sure that 1MB requests are not fragmented by the block layer.
1622 Severity : enhancement
1624 Description: organize the server-side client stats on per-nid basis
1625 Details : Change the structure of stats under obdfilter and mds to
1633 The "uuid"s file would list the uuids of _active_ exports.
1634 And the clear entry is to clear all stats and stale nids.
1636 Severity : enhancement
1638 Description: eliminate client locks in face of contention
1639 Details : file contention detection and lockless i/o implementation
1640 for contended files.
1644 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
1647 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
1649 Description: MDS may lose file striping (and hence file data) in some cases
1650 Details : If there are additional extended attributes stored on the MDS,
1651 in particular ACLs, SELinux, or user attributes (if user_xattr
1652 is specified for the client mount options) then there is a risk
1653 of attribute loss. Additionally, the Lustre file striping
1654 needs to be larger than default (e.g. striped over all OSTs),
1655 and an additional attribute must be stored initially in the
1656 inode and then increase in size enough to be moved to the
1657 external attribute block (e.g. ACL growing in size) for file
1662 Description: skiplist implementation simplification
1663 Details : skiplists are used to group compatible locks on granted list
1664 that was implemented as tracking first and last lock of each
1665 lock group the patch changes that to using doubly linked lists
1669 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
1670 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
1671 because LASSERT on both the data supplied by a client, and the
1672 data on disk is dangerous and incorrect.
1674 Severity : enhancement
1676 Description: Slow truncate/writes to huge files at high offsets.
1677 Details : Directly associate cached pages to lock that protect those pages,
1678 this allows us to quickly find what pages to write and remove
1679 once lock callback is received.
1683 Description: more ldlm soft lockups
1684 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
1685 starve other threads from the resource lock for a long time in
1686 case of long waiting queue, so change the debug level from
1687 D_OTHER to the less frequently used D_INFO.
1689 Severity : enhancement
1691 Description: add -gid, -group, -uid, -user options to lfs find
1695 Description: Hung threads in invalidate_inode_pages2_range
1696 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
1697 once one is completed. As a result, some RPCs are stuck in the
1698 queue and are never sent.
1702 Description: filter threads hungs on waiting journal commit
1703 Details : Cleanup filter group llog code, then only filter group llog will
1704 be only created in the MDS/OST syncing process.
1708 Description: Procfs and llog threads access destoryed import sometimes.
1709 Details : Sync the import destoryed process with procfs and llog threads by
1710 the import refcount and semaphore.
1712 Severity : enhancement
1714 Description: openlock cache of b1_6 port to HEAD
1719 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
1720 Details : If insertion of an extent fails, then discard the inode
1721 preallocation and free data blocks else it can lead to duplicate
1726 Description: don't always update ctime in ext3_xattr_set_handle()
1727 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
1728 In some cases the ctime should not be updated, for example for
1729 2.0->1.8 compatibility it is necessary to delete an xattr and it
1730 should not update the ctime.
1734 Bugzilla : 15713/16362
1735 Description: Assertion in iopen_connect_dentry in 1.6.3
1736 Details : looking up an inode via iopen with the wrong generation number can
1737 populate the dcache with a disconneced dentry while the inode
1738 number is in the process of being reallocated. This causes an
1739 assertion failure in iopen since the inode's dentry list contains
1740 both a connected and disconnected dentry.
1744 Description: assertion failure in ldlm_handle2lock()
1745 Details : fix a race between class_handle_unhash() and class_handle2object()
1746 introduced in lustre 1.6.5 by bug 13622.
1751 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
1752 Details : remove the limit of 256 scsi disks in the sd_iostat patch
1757 Description: oops in sd_iostats_seq_show()
1758 Details : unloading/reloading the scsi low level driver triggers a kernel
1759 bug when trying to access the sd iostat file.
1764 Description: Kernel panics during QLogic driver reload
1765 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
1766 patch, causing memory corruption.
1771 Description: journal_dev option does not work in b1_6
1772 Details : pass mount option during pre-mount.
1774 Severity : enhancement
1776 Description: Add a FIEMAP(FIle Extent MAP) ioctl
1777 Details : FIEMAP ioctl will allow an application to efficiently fetch the
1778 extent information of a file. It can be used to map logical blocks
1779 in a file to physical blocks in the block device.
1783 Description: LDLM soft lockups - improvement
1784 Details : It is be possible to send the lock handle along with each read
1785 or write request because the client is already doing a lock match
1786 itself so there isn't any reason the OST should have to re-do that
1790 Frequency : only on Cray X2
1792 Description: X2 build failures
1793 Details : fix build failures on Cray X2.
1797 Description: xid & resent requests
1798 Details : Initialize RPC XID from clock at startup (randomly if clock is
1801 Severity : enhancement
1803 Description: Add lustre_start utility to start or stop multiple Lustre servers
1808 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
1809 Details : In case of memory pressure, list_del() can be called twice on
1810 req->rq_history_list, causing a kernel oops.
1814 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
1815 Details : fix stack overflow in the distributed lock manager by defering
1816 export eviction after a failed AST to the elt thread instead of
1817 handling it in the dlm interpret routine.
1821 Description: Convert some comments to new format.
1822 Details : Update documenting comments to match doxygen conventions.
1826 Description: Grammar fixes.
1827 Details : A couple of trivial spelling fixes.
1831 Description: OSD_COUNTERS-mandatory
1832 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
1833 to simplify the code.
1837 Description: simplify cmm_device freeing logic.
1838 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
1842 Description: Add lockdep support to dt_object_operations locking interface.
1843 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
1844 parameter indicating lock ordering. Update mdd code to use new
1849 Description: Introduce failloc constants for lockless IO tests.
1850 Details : Add two new failloc constants to test lockless IO. Only one of
1851 them in implemented---another is checked in yet to be landed
1856 Description: Add lockdep support for inode mutex.
1857 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
1858 in the situations where Lustre has to lock more than one inode
1863 Description: Add optional invariants checking support.
1864 Details : Add new LINVRNT() macro, optional on new --enable-invariants
1865 configure switch. This macro is to be used for consistency and
1866 sanity checks that are too expensive to be left in `production'
1871 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
1872 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
1877 Description: Add lockdep support for ldlm_lock and ldlm_resource.
1878 Details : Use spin_lock_nested() in (the only) situation where more than
1879 one ldlm_lock is locked simultaneously. Also, fix possible
1880 dead-lock in ldlm_lock_change_resource() by enforcing particular
1885 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
1886 Details : Instead of specifying each ldlm_lock call-back through separate
1887 parameter, wrap them into struct ldlm_callback_suite.
1891 Description: Kill join_lru obd method and its callers.
1892 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
1893 in memory---a requirement implemented through ->o_join_lru() obd
1894 method in HEAD. Remove this method and its users.
1898 Description: Add asynchronous ldlm ENQUEUE completion handler.
1899 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
1900 case that stock ldlm_completion_ast() cannot handle as it waits
1901 until lock is granted. Introduce new ldlm_completion_ast_async()
1902 for this. Also comment ldlm_completion_ast().
1906 Description: ldlm_error <-> errno conversion.
1907 Details : Add functions to map (rather arbitrary) between LDLM error codes
1908 and standard errno values. CLIO needs this to prevent LDLM specific
1909 constants from escaping ldlm and osc.
1913 Description: Kill unused ldlm_handle2lock_ns() function.
1914 Details : Kill unused ldlm_handle2lock_ns() function.
1918 Description: Add lu_ref support to ldlm_lock
1919 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
1920 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
1921 and ldlm_resource. LDLM interface has to be changed, because code that
1922 releases a reference on a lock, has to "know" what reference this is.
1923 In the most frequent case
1925 lock = ldlm_handle2lock(handle);
1927 LDLM_LOCK_PUT(lock);
1929 no changes are required. When any other reference (received _not_ from
1930 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
1931 instead of LDLM_LOCK_PUT().
1933 Arguably, changes are pervasive, and interface requires some discipline
1934 for proper use. On the other hand, it was very instrumental in finding
1935 a few leaked lock references.
1939 Description: Add ldlm_lock_addref_try().
1940 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
1941 attempts to addref a lock that might be being canceled
1946 Description: Add ldlm_weigh_callback().
1947 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
1948 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
1949 hard-coded `number of pages' logic.
1953 Description: Add lockdep annotations to llog code.
1954 Details : Use appropriately tagged _nested() locking calls in the places
1955 where llog takes more than one ->lgh_lock lock.
1959 Description: Add loi_kms_set().
1960 Details : Wrap kms updates into a helper function.
1964 Description: Constify instances of struct lsm_operations.
1965 Details : Constify instances of struct lsm_operations.
1969 Description: lu_conf support.
1970 Details : On a server, a file system object is uniquely identified
1971 by a fid, which is sufficient to locate and load all object
1972 state (inode). On a client, on the other hand, more data are
1973 necessary instantiate an object. Change lu_object_find() and
1974 friends to take additional `lu_conf' argument describing object.
1975 Typically this includes layout information.
1979 Description: lu_context fixes.
1980 Details : Introduce new lu_context functions that are needed on the client
1981 side, where some system threads (ptlrpcd) are shared by multiple
1982 modules, and so cannot be stopped during module shutdown.
1986 Description: Add start and stop methods to lu_device_type_operations.
1987 Details : Introduce two new methods in lu_device_type_operations, that are
1988 invoked when first instance of a given type is created and last one
1989 is destroyed respectively. This is need by CLIO.
1993 Description: Add lu_ref support to struct lu_device.
1994 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
1995 track leaked references.
1999 Description: Introduce lu_kmem_descr.
2000 Details : lu_kmem_descr and its companion interface allow to create
2001 and destroy a number of kmem caches at once.
2005 Description: Fix lu_object finalization race.
2006 Details : Fix a race between lu_object_find() finding an object and its
2007 concurrent finalization. This race is (most likely) not possible
2008 on the server, but might happen on the client.
2012 Description: Introduce lu_ref interface.
2013 Details : lu_ref is a debugging module allowing to track references to
2014 a given object. It is quite cpu expensive, and has to be
2015 explicitly enabled with --enable-lu_ref. See usage description
2020 Description: Factor lu_site procfs stats into a separate function.
2021 Details : Separate lu_site stats printing code into a separate function
2022 that can be reused on a client.
2026 Description: Constify instances of struct {lu,dt,md}_device_operations.
2027 Details : Constify instances of struct {lu,dt,md}_device_operations.
2031 Description: Introduce struct md_site and move meta-data specific parts of
2032 struct lu_site here.
2033 Details : Move md-specific fields out of struct lu_site into special struct
2034 md_site, so that lu_site can be used on a client.
2038 Description: Kill mdd_lov_destroy().
2039 Details : Remove unused mdd code.
2043 Description: Add st_block checking to multistat.c.
2044 Details : Add st_block checking to multistat.c.
2048 Description: Add lu_ref support to struct obd_device.
2049 Details : Add lu_ref tracking to obd_device.
2053 Description: Kill obd_set_fail_loc().
2054 Details : Remove unused code.
2058 Description: Add special type for ptlrpc_request interpret functions.
2059 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2060 there. Actual usage will be in CLIO.
2064 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2065 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2066 doesn't work with lockdep.
2070 Description: Add rwv.c test program.
2071 Details : New testing program exercising readv(2) and writev(2) (Qian).
2075 Description: Add sendfile.c test program.
2076 Details : New testing program exercising sendfile(2) (Jay).
2080 Description: Ratelimit a message that can be very frequent.
2081 Details : Ratelimit a memory allocation failure message that can
2086 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2087 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2091 Description: Liblustre build fixes.
2092 Details : Liblustre build fixes.
2096 Description: libcfs: add cfs_{need,cond}_resched() interface.
2097 Details : libcfs: add cfs_{need,cond}_resched() definition and
2098 implementations for Linux, NT, and liblustre.
2100 Severity : enhancement
2102 Description: More exported tunables for mballoc
2103 Details : Add support for tunable preallocation window and new tunables for
2104 large/small requests
2108 Description: Detect corruption of block bitmap and checking for preallocations
2109 Details : Checks validity of on-disk block bitmap. Also it does better
2110 checking of number of applied preallocations. When corruption is
2111 found, it turns filesystem readonly to prevent further corruptions.
2115 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2116 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2117 window to 0. If it does hit cache miss, and read-pattern is still
2118 stride-io mode, does not reset the stride window, but also does not
2119 increase the stride window length in this case.
2123 Frequency : only for big-endian servers
2124 Description: Check if system is big-endian while mounting fs with extents feature
2125 Details : Mounting a filesystem with extents feature will fail on big-endian
2126 systems since ext3-based ldiskfs is not supported on big-endian
2127 systems. This can be over-riden with "bigendian_extents" mount option.
2129 Severity : enhancement
2131 Description: The root squash functionality
2132 Details : A security feature, which is to prevent users from being able
2133 to mount lustre on their desktop, run as root, and delete
2134 all of the files in the filesystem. The goal is accomplished by
2135 remapping user id (UID) and group id (GID) of the root user to
2136 a UID and GID specified by the system administartor via Lustre
2137 configuration management server (MGS). The functionality also
2138 allows to specify sets of clients for which the remapping does
2143 Description: Excessive recovery window
2144 Details : With AT enabled, the recovery window can be excessively long (6000+
2145 seconds). To address this problem, we no longer use
2146 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2147 timeout no longer depends on the service time, it is set to
2148 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2149 time via pb_service_time.
2153 Description: Watchdog triggered on MDS failover
2154 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2155 the OSTs know that the MDS "UUID" can be reused for the same export
2156 from a different NID, so we do not need to wait for the export to be
2160 Frequency : rare, only if using MMP with Linux RAID
2162 Description: MMP doesn't work with Linux RAID
2163 Details : While using HA for Lustre servers with Linux RAID, it is possible
2164 that MMP will not detect multiple mounts. To make this work we
2165 need to unplug the device queue in RAID when the MMP block is being
2166 written. Also while reading the MMP block, we should read it from
2167 disk and not the cached one.
2169 Severity : enhancement
2171 Description: open file using fid
2172 Details : A file can be opened using just its fid, like
2173 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2176 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2179 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2180 Details : add proper locking for creation of jbd_4k slab cache
2184 Description: MMP check in ext3_remount() fails without displaying any error
2185 Details : When multiple mount protection fails during remount, proper error
2188 Severity : enhancement
2190 Description: Allow stripe size to be up to 4G-64k
2191 Details : Fix math logic to allow large stripe sizes.
2195 Description: add check for >8TB ldiskfs filesystems
2196 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2197 Don't allow >8TB ldiskfs filesystems to be mounted without
2198 force_over_8tb mount option
2203 Descriptoin: OSS complains "too many blocks" and crashes
2204 Details : In filter_commitrw_write():
2205 1. lquota_chkquota();
2206 2. obdo_from_inode();
2207 3. lquota_pending_commit();
2208 "2." has changed uid and gid in this case, then caused
2209 lquota_chkquota() and lquota_pending_commit() to use
2210 different uid and gid. That is the root of the bug.
2214 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2215 Details : Detect and print out the 'stripe count = -1' case properly.
2217 --------------------------------------------------------------------------------
2219 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2221 * Support for kernels:
2222 2.6.5-7.283 (SLES 9),
2223 2.6.9-55.EL (RHEL 4),
2224 2.6.16.46-0.14 (SLES 10),
2225 2.6.18.8 vanilla (kernel.org)
2226 * Client support for unpatched kernels:
2227 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2228 2.6.16 - 2.6.22 vanilla (kernel.org)
2229 * Due to recently discovered recovery problems, we do not recommend
2230 using patchless RHEL 4 clients with this or any earlier release.
2231 * Recommended e2fsprogs version: 1.39.cfs8
2232 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2233 * Starting with this release, the ldiskfs backing filesystem required
2234 by Lustre is now in its own package, lustre-ldiskfs. This package
2235 should be installed. It is versioned separately from Lustre and
2236 may be released separately in future.
2240 Description: block reactivating mgc import until all deactivates complete
2241 Details : Fix race when failing back MDT/MGS to itself (testing)
2244 Frequency : at statup only
2246 Description: mds_lov_synchronize race leads to various problems
2247 Details : simultaneous MDT->OST connections at startup can cause the
2248 sync to abort, leaving the OSC in a bad state.
2250 Severity : enhancement
2252 Description: add optional extra BUILD_VERSION info
2253 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2254 to override the lustre version.
2257 Frequency : 2.6.18 servers only
2259 Description: ll_kern_mount() doesn't release the module reference
2260 Details : The ldiskfs module reference count never drops down to 0
2261 because ll_kern_mount() doesn't release the module reference.
2266 Description: server LBUG when using old ost_num_threads parameter
2267 Details : Accept the old ost_num_threads parameter but warn that it
2268 is deprecated, and fix an off-by-one error that caused an LBUG.
2273 Description: Transient SCSI error results in persistent IO issue
2274 Details : iobuf->dr_error is not reinitialized to 0 between two
2278 Frequency : sometimes when underlying device returns I/O errors
2280 Description: OSTs not going read-only during write failures
2281 Details : OSTs are not remounted read-only when the journal commit threads
2282 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2283 instead of the ext3 wrappers.
2287 Description: poor connect scaling with increasing client count
2288 Details : Don't run filter_grant_sanity_check for more than 100 exports
2289 to improve scaling for large numbers of clients.
2292 Frequency : SLES10 only
2294 Description: sanity-quota.sh quotacheck failed: rc = -22
2295 Details : Quotas cannot be enabled on SLES10.
2298 Frequency : liblustre clients only
2300 Description: getdirentries does not give error when run on compute nodes
2301 Details : getdirentries does not fail when the size specified as an argument
2302 is too small to contain at least one entry
2304 Severity : enhancement
2306 Description: Add LNET router traceability for debug purposes
2307 Details : If a checksum failure occurs with a router as part of the
2308 IO path, the NID of the last router that forwarded the bulk data
2309 is printed so it can be identified.
2314 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2315 Details : Due to a race condition, liblustre clients were occasionally
2316 evicted incorrectly.
2318 Severity : enhancement
2320 Description: lfs setstripe use optional parameters instead of postional
2323 Severity : enhancement
2325 Description: Nanosecond timestamp support for ldiskfs
2326 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2327 resolution timestamps. There is not yet support for this at
2328 the Lustre filesystem level.
2331 Frequency : during server recovery
2333 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2334 Details : request with rq_no_resend flag not awake l_wait_event if they get
2338 Frequency : nfs export on patchless client
2340 Description: connectathon hang when test nfs export over patchless client
2341 Details : Disconnected dentry cannot be found with lookup, so we do not need
2342 to unhash it or make it invalid
2345 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2346 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2347 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2352 Description: open req refcounting wrong on reconnect
2353 Details : If reconnect happened between getting open reply from server and
2354 call to mdc_set_replay_data in ll_file_open, we will schedule
2355 replay for unreferenced request that we are about to free.
2356 Subsequent close will crash in variety of ways.
2357 Check that request is still eligible for replay in
2358 mdc_set_replay_data().
2363 Description: disable writes to filesystem when reading health_check file
2364 Details : the default for reading the health_check proc file has changed
2365 to NOT do a journal transaction and write to disk, because this
2366 can cause reads of the /proc file to hang and block HA state
2367 checking on a healthy but otherwise heavily loaded system. It
2368 is possible to return to the previous behaviour during configure
2369 with --enable-health-write.
2371 Severity : enhancement
2373 Description: 64-bit inode version
2374 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2375 to the inode. This will be required for version-based recovery.
2380 Description: MDS fails to start if a duplicate client export is detected
2381 Details : in some rare cases it was possible for a client to connect to
2382 an MDS multiple times. Upon recovery the MDS would detect this
2383 and fail during startup. Handle this more gracefully.
2385 Severity : enhancement
2387 Description: Add -o localflock option to simulate old noflock
2389 Details : This will achieve local-only flock/fcntl locks
2395 Description: log_commit_thread vs filter_destroy race leads to crash
2396 Details : Take import reference before releasing llog record semaphore
2401 Description: Wrong request locking in request set processing
2402 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2403 imp_delayed_list, in this place should be used imp_lock.
2406 Frequency : when reconnection
2408 Description: Grant Leak when osc reconnect to OST
2409 Details : When osc reconnect ost, OST(filter) should check whether it
2410 should grant more space to client by comparing fed_grant and
2411 cl_avail_grant, and return the granted space to client instead
2412 of "new granted" space, because client will call osc_init_grant
2413 to update the client grant space info.
2416 Frequency : when client reconnect to OST
2418 Description: Grant Leak when osc do resend and replay bulk write
2419 Details : When osc reconnect to OST, OST(filter)should clear grant info of
2420 bulk write request, because the grant info while be sync between
2421 OSC and OST when reconnect, and we should ignore the grant info
2422 these of resend/replay write req.
2427 Description: Grant space more than avaiable left space sometimes.
2428 Details : When then OST is about to be full, if two bulk writing from
2429 different clients came to OST. Accord the avaliable space of the
2430 OST, the first req should be permitted, and the second one
2431 should be denied by ENOSPC. But if the seconde arrived before
2432 the first one is commited. The OST might wrongly permit second
2433 writing, which will cause grant space > avaiable space.
2436 Frequency : when client is evicted
2438 Description: Grant might be wrongly erased when osc is evicted by OST
2439 Details : when the import is evicted by server, it will fork another
2440 thread ptlrpc_invalidate_import_thread to invalidate the
2441 import, where the grant will be set to 0. While the original
2442 thread will update the grant it got when connecting. So if
2443 the former happened latter, the grant will be wrongly errased
2444 because of this race.
2449 Description: Checking Stale with correct fid
2450 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2451 to check whether it is stale, because sometimes, we want the
2452 enqueue happened anyway, and op_data.fid2 will not be initialized.
2454 Severity : enhancement
2456 Description: update patchless client
2457 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2460 Frequency : only with 2.4 kernel
2462 Description: random memory corruption
2463 Details : size of struct ll_inode_info is to big for union inode.u and this
2464 can be cause of random memory corruption.
2469 Description: Memory leak in recovery
2470 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2471 It should also check obd_fail before fsfilt_start, otherwise if
2472 fsfilt_start return -EROFS,(failover mds during mds recovery).
2473 then the req will return with repmsg->transno = 0 and rc = EROFS.
2474 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2475 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2476 be freed no matter whether fsfilt_commit success or not.
2479 Frequency : only with huge count clients
2481 Description: Prevents from taking the superblock lock in llap_from_page for
2483 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2484 for llap_from_page call in ll_removepage prevents from taking the
2485 superblock lock for a soon died page.
2490 Description: Not check open intent error before release open handle
2491 Details : in some rare cases, the open intent error is not checked before
2492 release open handle, which may cause
2493 ASSERTION(open_req->rq_transno != 0), because it tries to release
2494 the failed open handle.
2499 Description: Set cat log bitmap only after create log success.
2500 Details : in some rare cases, the cat log bitmap is set too early. and it
2501 should be set only after create log success.
2505 Description: Accessing a block bevice can re-enable I/O when Lustre is
2506 tearing down a device.
2507 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2512 Description: service threads may hog cpus when there are a lot of requests
2513 Details : Insert cond_resched to give other threads a chance to use some CPU
2518 Description: the cat log was not initialized in recovery
2519 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2520 unlink log on mds will not be initialized until mds post
2521 recovery. And also in mds post recovery, the unlink log will
2522 initialization will be done asynchronausly, so there will be race
2523 between add unlink log and unlink log initialization.
2527 Description: brw_stats were being printed incorrectly
2528 Details : brw_stats were being printed as log2 but all of them were not
2529 recorded as log2. Also remove some code duplication arising from
2530 filter_tally_{read,write}.
2534 Frequency : rare, only in recovery.
2535 Description: ASSERTION(req->rq_type != LI_POISON) failed
2536 Details : imp_lock should be held while iterating over imp_sending_list for
2537 prevent destroy request after get timeout in ptlrpc_queue_wait.
2541 Description: replay-single.sh test 52 fails
2542 Details : A lock's skiplist need to be cleanup when it being unlinked
2543 from its resource list.
2547 Description: Short directio read returns full requested size rather than
2549 Details : Direct I/O operations should return actual amount of bytes
2550 transferred rather than requested size.
2552 Severity : enhancement
2554 Description: metadata RPC reduction (e.g. for rm performance)
2555 Details : decrease the amount of synchronous RPC between clients and servers
2556 by canceling conflicing lock before the operation on the client
2557 and packing thier handles into the main operation RPC to server.
2559 Severity : enhancement
2561 Description: Async OSC create to avoid the blocking unnecessarily.
2562 Details : If a OST has no remain object, system will block on the creating
2563 when need to create a new object on this OST. Now, ways use
2564 pre-created objects when available, instead of blocking on an
2565 empty osc while others are not empty. If we must block, we block
2566 for the shortest possible period of time.
2570 Description: improve handling recoverable errors
2571 Details : if request processig with error which can be recoverable on server
2572 request should be resend, otherwise page released from cache and
2575 Severity : enhancement
2577 Description: refine locking for avoid write wrong info into lov_objid file
2578 Details : fix possible races with add new target and write/update data in
2581 --------------------------------------------------------------------------------
2583 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2588 Frequency : on some architectures
2590 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2591 Details : Uninitialized flags sometimes cause configuration commands to
2595 Frequency : patchless clients only
2597 Description: missing __iget() symbol export
2598 Details : The __iget() symbol export is missing. To avoid the need for
2599 this on patchless clients the deathrow inode reaper is turned
2600 off, and we depend on the VM to clean up old inodes. This
2601 dependency was during via the fix for bug 12181.
2603 --------------------------------------------------------------------------------
2605 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2607 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2608 INTEROPERATE with older versions automatically. In many cases a
2609 special upgrade step is needed. Please read the
2610 user documentation before upgrading any part of a live system.
2611 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
2612 version will not interoperate with 1.6 betas before beta5 (1.5.95).
2613 * WARNING: Lustre configuration and startup changes are required with
2614 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2618 Severity : enhancement
2620 Description: MountConf
2621 Details : Lustre configuration is now managed via mkfs and mount
2622 commands instead of lmc and lconf. New obd types (MGS, MGC)
2623 are added for dynamic configuration management. See
2624 https://mail.clusterfs.com/wikis/lustre/MountConf for
2627 Severity : enhancement
2629 Description: dynamic OST addition
2630 Details : OSTs can now be added to a live filesystem
2632 Severity : enhancement
2634 Description: startup order invariance
2635 Details : MDTs and OSTs can be started in any order. Clients only
2636 require the MDT to complete startup.
2638 Severity : enhancement
2640 Description: parallel, asynchronous orphan cleanup
2641 Details : orphan cleanup is now performed in separate threads for each
2642 OST, allowing parallel non-blocking operation.
2644 Severity : enhancement
2646 Description: optimized stripe assignment
2647 Details : stripe assignments are now made based on ost space available,
2648 ost previous usage, and OSS previous usage, in order to try
2649 to optimize storage space and networking resources.
2651 Severity : enhancement
2653 Description: Permanently set tunables
2654 Details : All writable /proc/fs/lustre tunables can now be permanently
2655 set on a per-server basis, at mkfs time or on a live system.
2657 Severity : enhancement
2659 Description: Lustre message v2
2660 Details : Add lustre message format v2.
2662 Severity : enhancement
2664 Description: client OST exclusion list
2665 Details : Clients can be started with a list of OSTs that should be
2666 declared "inactive" for known non-responsive OSTs.
2670 Description: SPEC SFS validation failure on NFS v2 over lustre.
2671 Details : Changes the blocksize for regular files to be 2x RPC size,
2672 and not depend on stripe size.
2674 Severity : enhancement
2676 Description: Multiple MD RPCs in flight.
2677 Details : Further unserialise some read-only MDS RPCs - learn about intents.
2678 To avoid overly-overloading MDS, introduce a limit on number of
2679 MDS RPCs in flight for a single client and add /proc controls
2680 to adjust this limit.
2682 Severity : enhancement
2684 Description: client read/write statistics
2685 Details : Add client read/write call usage stats for performance
2686 analysis of user processes.
2687 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2688 file access. extents_stats shows chunk size distribution.
2689 extents_stats_per_process show chunk size distribution per
2692 Severity : enhancement
2694 Description: mds statistics
2695 Details : Add detailed mds operations statistics in
2696 /proc/fs/lustre/mds/*/stats.
2700 Description: Failure of copying files with lustre special EAs.
2701 Details : Client side always return success for setxattr call for lustre
2702 special xattr (currently only "trusted.lov").
2707 Description: Refcount LNET uuids
2708 Details : The global LNET uuid list grew linearly with every startup;
2709 refcount repeated list entries instead of always adding to
2712 Severity : enhancement
2714 Description: Dynamic service threads
2715 Details : Within a small range, start extra service threads
2716 automatically when the request queue builds up.
2719 Frequency : mixed-endian client/server environments
2721 Description: mixed-endian crashes
2722 Details : The new msg_v2 system had some failures in mixed-endian
2725 Severity : enhancement
2727 Description: Easy OST removal
2728 Details : OSTs can be permanently deactivated with e.g. 'lctl
2729 conf_param lustre-OST0001.osc.active=0'
2731 Severity : enhancement
2733 Description: MGS proc entries
2734 Details : Added basic proc entries for the MGS showing what filesystems
2737 Severity : enhancement
2739 Description: provide MGS failover
2740 Details : Added config lock reacquisition after MGS server failover.
2742 Severity : enhancement
2744 Description: add Linux 2.4 support
2745 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
2749 Description: a large application tries to do I/O to the same resource and dies
2750 in the middle of it.
2751 Details : Check the req->rq_arrival time after the call to
2752 ost_brw_lock_get(), but before we do anything about
2753 processing it & sending the BULK transfer request. This
2754 should help move old stale pending locks off the queue as
2755 quickly as obd_timeout.
2758 Frequency : when an incorrect nid is specified during startup
2760 Description: ptlrpc connect to non-existant node causes kernel crash
2761 Details : LNET can't be re-entered from an event callback, which
2762 happened when we expire a message after the export has been
2763 cleaned up. Instead, hand the zombie cleanup off to another
2766 Severity : enhancement
2768 Description: plain/inodebits lock performance improvement
2769 Details : Grouping plain/inodebits in granted list by their request modes
2770 and bits policy, thus improving the performance of search through
2774 Frequency : only if OST filesystem is corrupted
2776 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2777 Details : for a short time RPCs with bulk IO are in the replay list,
2778 but replay of bulk IOs is unimplemented. If the OST filesystem
2779 is corrupted due to disk cache incoherency and then replay is
2780 started it is possible to trip an assertion. Avoid putting
2781 committed RPCs into the replay list at all to avoid this issue.
2784 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2786 Description: System hang on startup
2787 Details : This bug allowed the liblustre (e.g. catamount) client to
2788 return to the app before handling all startup RPCs. This
2789 could leave the node unresponsive to lustre network traffic
2790 and manifested as a server ptllnd timeout.
2792 Severity : enhancement
2794 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2795 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
2796 boolean to control whether to print peer debug info when a
2797 client's RPC times out.
2800 Frequency : only for kernels with patches from Lustre below 1.4.3
2802 Description: Remove old rdonly API
2803 Details : Remove old rdonly API which unused from at least lustre 1.4.3
2806 Frequency : only for devices with external journals
2808 Description: Set external device read-only also
2809 Details : During a commanded failover stop, we set the disk device
2810 read-only while the server shuts down. We now also set any
2811 external journal device read-only at the same time.
2814 Frequency : when upgrading from 1.4 while trying to change parameters
2816 Description: The wrong (new) MDC name was used when setting parameters for
2817 upgraded MDT's. Also allows changing of OSC (and MDC)
2818 parameters if --writeconf is specified at tunefs upgrade time.
2821 Frequency : when setting specific ost indicies
2823 Description: QOS code breaks on skipped indicies
2824 Details : Add checks for missing OST indicies in the QOS code, so OSTs
2825 created with --index need not be sequential.
2827 Severity : enhancement
2829 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
2830 Details : The uninit_groups feature works in conjunction with the kernel
2831 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
2832 up the pass1 processing of e2fsck. This is a read-only feature
2833 in ldiskfs2 only, so older kernels and current ldiskfs cannot
2834 mount filesystems that have had this feature enabled.
2836 Severity : enhancement
2838 Description: Improve multi-block allocation algorithm to avoid fragmentation
2839 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
2840 allocation locality and avoid filesystem fragmentation.
2842 ------------------------------------------------------------------------------
2844 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
2846 * Support for kernels:
2847 2.4.21-47.0.1.EL (RHEL 3)
2848 2.6.5-7.283 (SLES 9)
2849 2.6.9-42.0.10.EL (RHEL 4)
2850 2.6.12.6 vanilla (kernel.org)
2851 2.6.16.27-0.9 (SLES 10)
2852 * Recommended e2fsprogs version: 1.39.cfs5
2854 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2858 Frequency : occasional, depends on client load and configuration
2859 Bugzilla : 12181, 12203
2860 Description: data loss for recently-modified files
2862 Details : In some cases it is possible that recently written or created
2863 files may not be written to disk in a timely manner (this should
2864 normally be within 30s unless client IO load is very high).
2865 The problem appears as zero-length files or files that are a
2866 multiple of 1MB in size after a client crash or client eviction
2867 that are missing data at the end of the file.
2869 This problem is more likely to be hit on clients where files are
2870 repeatedly created and unlinked in the same directory, clients
2871 have a large amount of RAM, have many CPUs, the filesystem has
2872 many OSTs, the clients are rebooted frequently, and/or the files
2873 are not accessed by other nodes after being written.
2875 The presence of the problem can be detected by looking at
2876 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
2877 smaller than the second (nr_unused) then dirty files will not
2878 be flushed automatically to disk. "sync; sleep 10" should be
2879 run several times on the node before unmounting it to update
2880 Lustre (this is also safe to run on nodes without this problem).
2882 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
2883 that can cause this same problem, so customers using that kernel
2884 also need to update the kernel in addition to Lustre. In order
2885 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
2887 It is normal that files written just before a client crash (less
2888 than 30s) may not yet have been flushed to disk, even for local
2892 Frequency : frequent on thin XT3 nodes
2894 Description: UUID collision on thin XT3 Linux nodes
2895 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
2896 randomly, since we relied on an insufficiently-seeded PRNG.
2901 Description: OSS hangs after "All ost request buffers busy"
2902 Details : A deadlock between quota and journal operations caused OSS
2903 hangs after printing "All ost request buffers busy."
2906 Frequency : always on liblustre builds
2908 Description: Cleanup compiler warnings on liblustre
2911 Frequency : always on liblustre builds on XT3
2913 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
2917 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
2918 > 32000 subdirectories
2919 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
2920 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
2921 subdirectory count crosses 32000. This will aid e2fsck to
2922 correctly handle more than 32000 subdirectories.
2925 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
2927 Description: System hang on startup
2928 Details : This bug allowed the liblustre (e.g. catamount) client to
2929 return to the app before handling all startup RPCs. This
2930 could leave the node unresponsive to lustre network traffic
2931 and manifested as a server ptllnd timeout.
2933 Severity : enhancement
2935 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
2936 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
2937 boolean to control whether to print peer debug info when a
2938 client's RPC times out.
2943 Description: make O_SYNC working on 2.6 kernels
2944 Details : 2.6 kernels use different method for mark pages for write,
2945 so need add a code to lustre for O_SYNC work.
2950 Description: Failure to close file and release space on NFS
2951 Details : Put inode details into lock acquired in ll_intent_file_open.
2952 Use mdc_intent_lock in ll_intent_open to properly
2953 detect all kind of errors unhandled by mdc_enqueue.
2958 Description: proc file read during shutdown sometimes raced obd removal,
2960 Details : Add lock to prevent obd access after proc file removal.
2963 Frequency : Only for files larger than 4GB on 32-bit clients.
2965 Description: improperly doing page alignment of locks
2966 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
2967 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
2972 Description: RPCs being resent when they shouldn't be
2973 Details : Some RPCs that should not be resent are being resent. This
2974 can cause inconsistencies in the RPC state machine. Do not
2975 resend such requests.
2978 Frequency : rare, only with NFS export
2980 Description: Crash on NFS re-export node
2981 Details : under very unusual load conditions an assertion is hit in
2982 ll_intent_file_open()
2985 Frequency : only if OST filesystem is corrupted
2987 Description: client incorrectly hits assertion in ptlrpc_replay_req()
2988 Details : for a short time RPCs with bulk IO are in the replay list,
2989 but replay of bulk IOs is unimplemented. If the OST filesystem
2990 is corrupted due to disk cache incoherency and then replay is
2991 started it is possible to trip an assertion. Avoid putting
2992 committed RPCs into the replay list at all to avoid this issue.
2997 Description: large O_DIRECT requests fail under memory pressure/fragmentation
2998 Details : Large single O_DIRECT read and write calls can fail to allocate
2999 a sufficiently large buffer to process the request. In case of
3000 allocation failure the allocation is retried with a smaller
3001 buffer and broken into smaller requests.
3003 Severity : enhancement
3005 Description: Add -o localflock option to simulate old noflock behaviour.
3006 Details : This will achieve local-only flock/fcntl locks coherentness.
3011 Description: versioning check is incomplete
3012 Details : Checking the version difference of client vs. server, report
3013 error if the gap is too big.
3018 Description: add support PG_writeback bit
3019 Details : add support for PG_writeback bit for Lustre, for more carefull
3020 work with page cache in 2.6 kernel. This also fix some deadlocks
3021 and remove hack for work O_SYNC with 2.6 kernel.
3023 Severity : enhancement
3025 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3026 Details : The uninit_groups feature works in conjunction with the kernel
3027 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3028 up the pass1 processing of e2fsck. This is a read-only feature
3029 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3030 mount filesystems that have had this feature enabled.
3032 Severity : enhancement
3034 Description: Improve multi-block allocation algorithm to avoid fragmentation
3035 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3036 allocation locality and avoid filesystem fragmentation.
3038 ------------------------------------------------------------------------------
3040 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3042 * Support for kernels:
3043 2.6.9-42.0.3.EL (RHEL 4)
3044 2.6.5-7.276 (SLES 9)
3045 2.4.21-47.0.1.EL (RHEL 3)
3046 2.6.12.6 vanilla (kernel.org)
3047 2.6.16.21-0.8 (SLES10)
3048 * Recommended e2fsprogs version: 1.39.cfs2-0
3050 * The backwards-compatible /proc/sys/portals symlink has been removed
3051 in this release. Before upgrading, please ensure that you change
3052 any configuration scripts or /etc/sysctl.conf files that access
3053 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3054 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3055 in advance of the upgrade on any system running Lustre 1.4.6 or
3056 newer, since /proc/sys/lnet was added in that version.
3057 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3061 Frequency : only when quota is used
3063 Description: avoid scanning export list for quota master
3064 Details : Change the algorithms to avoid scanning export list in order
3065 to improve the efficiency.
3068 Frequency : MDS failover only, very rarely
3070 Description: "went back in time" messages on mds failover
3071 Details : The greatest transno may be lost when the current operation
3072 finishes with an error (transno==0) and the client's last_rcvd
3073 record is over-written. Save the greatest transno in the
3074 mds_last_transno for this case.
3077 Frequency : always for specific kernels and striping counts
3079 Description: client may get "Matching packet too big" without ACL support
3080 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3081 when trying to access files in certain configurations. The
3082 clients should in fact be denied when mounting because they do
3083 not understand ACLs.
3086 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3088 Description: many clients connecting with IO in progress causes connect timeouts
3089 Details : Avoid synchronous journal commits to avoid delays caused by many
3090 clients connecting/disconnecting when bulk IO is in progress.
3091 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3092 OST_IO_PORTAL to avoid delays behind potentially many pending
3096 Frequency : occasionally with multiple writers to a single file
3098 Description: shared writes to file may result in wrong size reported by stat()
3099 Details : Allow growing of kms when extent lock is cancelled
3102 Frequency : always with random mmap IO to multi-striped file
3104 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3105 Details : Only if the hole is at the end of OST object so that kms is too
3106 small. Fix is to increase kms accordingly in ll_nopage.
3109 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3111 Description: writes to a missing object would leak memory on the OST
3112 Details : If there is an inconsistency between the MDS and OST filesystems,
3113 such that the MDS references an object that doesn't exist, writes
3114 to that object will leak memory due to incorrect cleanup in the
3115 error handling path, eventually running out of memory on the OST.
3120 Description: Creating too long symlink causes lustre errors
3121 Details : Check symlink and name lengths before sending requests to MDS.
3124 Frequency : only if flock is enabled (not on by default)
3126 Description: posix locks not released on fd closure on 2.6.9+
3127 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3128 kernels, this caused such locks not to be released on fd close and
3129 then assertions on fs unmount about still used locks.
3132 Frequency : MDS failover only, very rarely
3134 Description: clients may get ASSERTION(granted_lock != NULL)
3135 Details : When request was taking a long time, and a client was resending
3136 a getattr by name lock request. The were multiple lock requests
3137 with the same client lock handle and
3138 mds_getattr_name->fixup_handle_for_resent_request found one of the
3139 lock handles but later failed with ASSERTION(granted_lock != NULL).
3144 Description: handle->h_buffer_credits > 0, assertion failure
3145 Details : h_buffer_credits is zero after truncate, causing assertion
3146 failure. This patch extends the transaction or creates a new
3150 Frequency : NFS re-export or patchless client
3151 Bugzilla : 11179, 10796
3152 Description: Crash on NFS re-export node (__d_move)
3153 Details : We do not want to hash the dentry if we don't have a lock.
3154 But if this dentry is later used in d_move, we'd hit uninitialised
3155 list head d_hash, so we just do this to init d_hash field but
3156 leave dentry unhashed.
3159 Frequency : NFS re-export or patchless client
3161 Description: NFS exports has problem with symbolic link
3162 Details : lustre client didn't properly install dentry when re-exported
3163 to NFS or running patchless client.
3166 Frequency : NFS re-export or patchless client
3168 Description: Various nfs/patchless fixes.
3169 Details : fixes reuse disconected alias for lookup process - this fixes
3170 warning "find_exported_dentry: npd != pd",
3171 fix permission error with open files at nfs.
3172 fix apply umask when do revalidate.
3175 Frequency : occasional
3177 Description: Crash on NFS re-export node
3178 Details : calling clear_page() on the wrong pointer triggered oops in
3179 generic_mapping_read().
3182 Frequency : rarely, using O_DIRECT IO
3184 Description: unaligned directio crashes client with LASSERT
3185 Details : check for unaligned buffers before trying any requests.
3188 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3190 Description: stale data returned from RAID cache
3191 Details : If only a small amount of IO is done to the RAID device before
3192 reading it again it is possible to get stale data from the RAID
3193 cache instead of reading it from disk.
3196 Frequency : always for sles10 kernel
3198 Description: sles10 support
3199 Details : ll_follow_link: compile fixes and using of nd_set_link
3200 under newer kernels.
3203 Frequency : depends on arch, kernel and compiler version, always on sles10
3206 Description: recursive or deep enough symlinks cause stack overflow
3207 Details : getting rid of large stack-allocated variable in
3211 Frequency : depends on hardware
3213 Description: lustre write performance loss in the SLES10 kernel
3214 Details : the performance loss is caused by using of write barriers in the
3215 ext3 code. The SLES10 kernel turns barrier support on by
3216 default. The fix is to undo that change for ldiskfs.
3218 ------------------------------------------------------------------------------
3220 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3222 * Support for kernels:
3223 2.6.9-42.0.3EL (RHEL 4)
3224 2.6.5-7.276 (SLES 9)
3225 2.4.21-47.0.1.EL (RHEL 3)
3226 2.6.12.6 vanilla (kernel.org)
3230 Frequency : quota enabled and large files being deleted
3232 Description: releasing more than 4GB of quota at once hangs OST
3233 Details : If a user deletes more than 4GB of files on a single OST it
3234 will cause the OST to spin in an infinite loop. Release
3235 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3240 Description: statfs data retrieved from /proc may be stale or zero
3241 Details : When reading per-device statfs data from /proc, in the
3242 {kbytes,files}_{total,free,avail} files, it may appear
3243 as zero or be out of date.
3246 Frequency : systems with MD RAID1 external journal devices
3248 Description: lconf's call to blkid is confused by RAID1 journal devices
3249 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3250 of returning all block devices that match the journal UUID.
3253 Frequency : always, for aggregate stripe size over 4GB
3255 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3256 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3257 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3260 Frequency : always if "lfs find" used on a local file/directory
3262 Description: "lfs find" segfaults if used on a local file/directory
3263 Details : The case where a directory component was not specified wasn't
3264 handled correctly. Handle this properly.
3267 Frequency : always on ppc64
3269 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3270 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3271 perform endianness conversion.
3274 Frequency : rarely (truncate to non-zero file size after write under load)
3275 Bugzilla : 10730, 10687
3276 Description: Files padded with zeros to next 4K multiple
3277 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3278 it is possible that files that are truncated to a non-zero size
3279 immediately after being written are filled with zero bytes beyond
3280 the truncated size. No file data is lost.
3282 Severity : enhancement
3284 Description: Allow recovery/failover for liblustre clients.
3285 Details : liblustre clients were unaware of failover configurations until
3288 Severity : enhancement
3290 Description: user file locks should fail when not mounting with flock option
3291 Details : Set up an error-returning stub in ll_file_operations.lock field
3292 to prevent incorrect behaviour when client is mounted without
3293 flock option. Also, set up properly f_op->flock field for
3297 Frequency : always on ia64
3299 Description: "lfs df" loops on printing out MDS statfs information
3300 Details : The obd_ioctl_data was not initialized and in some systems
3301 this caused a failure during the ioctl that did not return
3302 an error. Initialize the struct and return an error on failure.
3305 Frequency : SLES 9 only
3307 Description: Error of copying files with lustre special EAs as root
3308 Details : Client side always return success for setxattr call for lustre
3309 special xattr (currently only "trusted.lov").
3312 Frequency : rarely on clusters with both ia64+i386 clients
3314 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3315 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3316 clients are doing concurrent writes to the same file it is
3317 possible that the ia64 clients may LASSERT because the OST
3318 extent locks are not PAGE_SIZE aligned. Ensure that grown
3319 locks are always aligned on the request boundary.
3322 Frequency : specific use, occasional
3324 Description: Overwriting in use executable truncates on-disk binary image
3325 Details : If one node attempts to overwrite an executable in use by
3326 another node, we now correctly return ETXTBSY instead of
3327 truncating the file.
3329 Severity : enhancement
3331 Description: Async OSC create to avoid the blocking unnecessarily.
3332 Details : If a OST has no remain object, system will block on the creating
3333 when need to create a new object on this OST. Now, ways use
3334 pre-created objects when available, instead of blocking on an
3335 empty osc while others are not empty. If we must block, we block
3336 for the shortest possible period of time.
3341 Description: chmod on Lustre root is propagated to other clients
3342 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3343 invalid by the follow_mount time.
3346 Frequency : liblustre clients only
3348 Description: Race in 'instant cancel' lock handling could lead to such locks
3349 never to be granted in case of SMP MDS
3350 Details : Do not destroy not yet granted but cbpending locks in
3354 Frequency : replay/resend of open
3356 Description: non null lock assetion failure in mds_intent_policy
3357 Details : Trying to replay/resend lockless open requests resulted in
3358 mds_open() returning 0 with no lock. Now it sets a flag if
3359 a lock is going to be returned.
3361 Severity : enhancement
3363 Description: Checksum enhancements
3364 Details : New checksum enhancements allow for resending RPCs that failed
3367 Severity : enhancement
3369 Description: Tunables on number of dirty pages in cacche
3370 Details : Allow to set limit on number of dirty pages cached.
3375 Description: client crash on unmount - lock still has references
3376 Details : In some error handling cases it was possible to leak a lock
3377 reference on a client while accessing a file. This was not
3378 harmful to the client during operation, but would cause the
3379 client to crash when the filesystem is unmounted.
3382 Frequency : specific case, rare
3384 Description: ETXTBSY on mds though file not in use by client
3385 Details : ETXTBSY is no longer incorrectly returned when attempting to
3386 chmod or chown a directory that the user previously tried to
3387 execute or a currently-executing binary.
3390 Frequency : extremely rare except on liblustre-based clients
3392 Description: Lustre space not freed when files are deleted
3393 Details : Clean up open-unlinked files after client eviction. Previously
3394 the unlink was skipped and the files remained as orphans.
3399 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3400 Details : In some cases it was possible to send a blocking callback to
3401 a client doing a glimpse, even though that client didn't get
3402 a lock granted. When the glimpse lock is cancelled on the OST
3403 the freed lock is left on the waiting list and corrupted the list.
3406 Frequency : all core dumps
3408 Description: Broke core dumps to lustre
3409 Details : Negative dentry may be unhashed if parent does not have UPDATE
3410 lock, but some callers, e.g. do_coredump, expect dentry to be
3411 hashed after successful create, hash it in ll_create_it.
3413 ------------------------------------------------------------------------------
3415 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3417 * Support for kernels:
3418 2.6.9-42.0.2.EL (RHEL 4)
3419 2.6.5-7.276 (SLES 9)
3420 2.4.21-40.EL (RHEL 3)
3421 2.6.12.6 vanilla (kernel.org)
3425 Frequency : always on RHEL 3
3427 Description: Number of open files grows over time
3428 Details : The number of open files grows over time, whether or not
3429 Lustre is started. This was due to a filp leak introduced
3430 by one of our kernel patches.
3432 ------------------------------------------------------------------------------
3434 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3436 * Support for kernels:
3437 2.6.9-42.EL (RHEL 4)
3438 2.6.5-7.276 (SLES 9)
3439 2.4.21-40.EL (RHEL 3)
3440 2.6.12.6 vanilla (kernel.org)
3445 Bugzilla : 5719, 9635, 9792, 9684
3446 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3447 Details : If a server is under heavy load and cannot reply to new
3448 connection requests before the client resends the (re)connect,
3449 the connection handling code can behave badly if two service
3450 threads are concurrently handing separate (re)connections from
3451 the same client. Add better locking to the connection handling
3452 code, and ensure that only a single connection will be processed
3453 for a given client UUID, even if the lock is dropped.
3455 Severity : enhancement
3457 Description: add TCP zero-copy support to kernel
3458 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3459 sends if the hardware supports scatter-gather and checksumming.
3460 This allows socklnd to do client-write and server-read more
3461 efficiently and reduce CPU utilization from skbuf copying.
3464 Frequency : only if NFS exporting from client
3466 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3467 Details : If setting attributes on a file created under NFS that had
3468 never been opened it would be possible to oops the client
3469 if the file had no objects.
3472 Frequency : always for liblustre
3474 Description: liblustre client does MDS+OSTs setattr RPC for each write
3475 Details : When doing a write from a liblustre client, the client
3476 incorrectly issued an RPC to the MDS and each OST the file was
3477 striped over in order to update the timestamps. When writing
3478 with small chunks and many clients this could overwhelm the MDS
3479 with RPCs. In all cases it would slow down the write because
3480 these RPCs are unnecessary.
3482 Severity : enhancement
3484 Description: allow number of MDS service threads to be changed at module load
3485 Details : It is now possible to change the number of MDS service threads
3486 running. Adding "options mds mds_num_threads={N}" to the MDS's
3487 /etc/modprobe.conf will set the number of threads for the next
3488 time Lustre is restarted (assuming the "mds" module is also
3489 reloaded at that time). The default number of threads will
3490 stay the same, 32 for most systems.
3495 Description: OST crash if filesystem is unformatted or corrupt
3496 Details : If an OST is started on a device that has never been formatted
3497 or if the filesystem is corrupt and cannot even mount then the
3498 error handling cleanup routines would dereference a NULL pointer.
3503 Description: NULL pointer deref in llap_from_page.
3504 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3505 incorrect) mapping if the page was truncated/reclaimed while it was
3506 searched for. Check for this condition and skip such pages when
3507 doing readahead. Introduce extra check to llap_from_page() to
3508 verify page->mapping->host is non-NULL (so page is not anonymous).
3511 Frequency : Sometimes when using sys_sendfile
3513 Description: "page not covered by a lock" warnings from ll_readpage
3514 Details : sendfile called ll_readpage without right page locks present.
3515 Now we introduced ll_file_sendfile that does necessary locking
3516 around call to generic_file_sendfile() much like we do in
3520 Frequency : with certain MDS communication failures at client mount time
3522 Description: NULL pointer deref after failed client mount
3523 Details : a client connection request may delayed by the network layer
3524 and not be sent until after the PTLRPC layer has timed out the
3525 request. If the client fails the mount immediately it will try
3526 to clean up before the network times out the request. Add a
3527 reference from the request import to the obd device and delay
3528 the cleanup until the network drops the request.
3531 Frequency : occasionally during client (re)connect
3533 Description: assertion failure during client (re)connect
3534 Details : processing a client connection request may be delayed by the
3535 client or server longer than the client connect timeout. This
3536 causes the client to resend the connection request. If the
3537 original connection request is replied in this interval, the
3538 client may trip an assertion failure in ptlrpc_connect_interpret()
3539 which thought it would be the only running connect process.
3542 Frequency : only with obd_echo servers and clients that are rebooted
3544 Description: kernel BUG accessing uninitialized data structure
3545 Details : When running an obd_echo server it did not start the ping_evictor
3546 thread, and when a client was evicted an uninitialized data
3547 structure was accessed. Start the ping_evictor in the RPC
3548 service startup instead of the OBD startup.
3550 Severity : enhancement
3551 Bugzilla : 10193 (patchless)
3552 Description: Remove dependency on various unexported kernel interfaces.
3553 Details : No longer need reparent_to_init, exit_mm, exit_files,
3554 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3557 Frequency : rare (only users of deprecated and unsupported LDAP config)
3559 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3560 Details : LDAP apparently contains 'lustreName' attributes instead of
3561 'name'. A simple remapping of the name is sufficient.
3564 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3566 Description: waiting_locks_callback trips kernel BUG if client is evicted
3567 Details : Running with the dump_on_timeout debug flag turned on makes
3568 it possible that the waiting_locks_callback() can try to dump
3569 the Lustre kernel debug logs from an interrupt handler. Defer
3570 this log dumping to the expired_lock_main() thread.
3572 Severity : enhancement
3574 Description: Support NFS exporting on 2.6 kernels.
3575 Details : Implement non-rawops metadata methods for NFS server to use without
3576 changing NFS server code.
3579 Frequency : very rare (synthetic metadata workload only)
3581 Description: two racing renames might cause an MDS thread to deadlock
3582 Details : Running the "racer" program may cause one MDS thread to rename
3583 a file from being the source of a rename to being the target of
3584 a rename at exactly the same time that another thread is doing
3585 so, and the second thread has already enqueued these locks after
3586 doing a lookup of the target and is trying to relock them in
3587 order. Ensure that we don't try to re-lock the same resource.
3590 Frequency : only very large systems with liblustre clients
3592 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3593 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3594 all OSTs in parallel. This allows the eviction of stale liblustre
3595 clients to proceed much faster than if they were done in series,
3596 and also offers similar improvements for other set_info RPCs.
3601 Description: excessive CPU usage during initial read phase on client
3602 Details : During the initial read phase on a client, it would agressively
3603 retry readahead on the file, consuming too much CPU and impacting
3604 performance (since 1.4.5.8). Improve the readahead algorithm
3605 to avoid this, and also improve some other common cases (read
3606 of small files in particular, where "small" is files smaller than
3607 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3612 Description: MDS crash when receiving packet with unknown intent.
3613 Details : Do not LBUG in unknown intent case, just return -EFAULT
3615 Severity : enhancement
3616 Bugzilla : 9293, 9385
3617 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3618 Details : Do not serialize getattr (non-intent version) and statfs.
3621 Frequency : occasional, when OST network is overloaded/intermittent
3623 Description: client evicted by OST after bulk IO timeout
3624 Details : If a client sends a bulk IO request (read or write) the OST
3625 may evict the client if it is unresposive to its data GET/PUT
3626 request. This is incorrect if the network is overloaded (takes
3627 too long to transfer the RPC data) or dropped the OST GET/PUT
3628 request. There is no need to evict the client at all, since
3629 the pinger and/or lock callbacks will handle this, and the
3630 client can restart the bulk request.
3633 Frequency : Always when mmapping file with no objects
3635 Description: client crashes when mmapping file with no objects
3636 Details : Check that we actually have objects in a file before doing any
3637 operations on objects in ll_vm_open, ll_vm_close and
3643 Description: Request leak when working with deleted CWD
3644 Details : Introduce advanced request refcount tracking for requests
3645 referenced from lustre intent.
3647 Severity : Enhancement
3649 Description: Cache open file handles on client.
3650 Details : MDS now will return special lock along with openhandle, if
3651 requested and client is allowed to hold openhandle, even if unused,
3652 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3653 closing files for every read/write openration.
3655 Severity : Enhancement
3657 Description: Cache open negative dentries on client when possible.
3658 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3659 negative dentries on lock revocation.
3664 Description: Remounting a client read-only wasn't possible with a zconf mount
3665 Details : It wasn't possible to remount a client read-only with llmount.
3667 Severity : enhancement
3668 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3669 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3670 MPICH is included to add Lustre-specific ADIO interfaces.
3671 This is based closely on the UFS ADIO layer and only differs
3672 in file creation, in order to allow the OST striping to be set.
3673 This is user-contributed code and not supported by CFS.
3678 Description: extended inode attributes (immutable, append-only) work improperly
3679 when 2.4 and 2.6 kernels are used on client/server or vice versa
3680 Details : Introduce kernel-independent values for these flags.
3682 Severity : enhancement
3685 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3686 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3687 granularity. Now, Fractional MB granularity is be supported,
3688 this is very useful for low memory system.
3690 Severity : enhancement
3692 Description: Getattr by fid
3693 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3694 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3695 have locks and avoids getattr rpc requests for it on every stat.
3698 Frequency : Always, for filesystems larger than 2TB
3700 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3701 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
3702 filesystems larger than 2TB to be mounted with mballoc enabled.
3705 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
3707 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3708 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
3709 filesystems larger than 2TB is mounted then the kernel will
3710 silently corrupt the start of the filesystem. CONFIG_LBD is
3711 enabled for all CFS-supported kernels, but the possibility of
3712 this happening with a modified kernel config exists.
3714 Severity : enhancement
3716 Description: add client O_DIRECT support for 2.6 kernels
3717 Details : It is now possible to do O_DIRECT reads and writes to files
3718 in the Lustre client mountpoint on 2.6 kernel clients.
3720 Severity : enhancement
3722 Description: parallel glimpse, setattr, statfs, punch, destroy requests
3723 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
3724 parallel, not waiting for response from every OST before sending
3725 a rpc to the next OST.
3730 Description: setattr vs write race when updating file timestamps
3731 Details : Client processes that update a file timestamp into the past
3732 right after writing to the file (e.g. tar) it is possible that
3733 the updated file modification time can be reset to the current
3734 time due to a race between processing the setattr and write RPC.
3736 Severity : enhancement
3738 Description: Bring 'lfs find' closer in line with regular Linux find.
3739 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
3740 -print0 options and obtains all the needed info through the lustre
3743 Severity : enhancement
3745 Description: support up to 1024 configured devices on one node
3746 Details : change obd_dev array from statically allocated to dynamically
3747 allocated structs as they are first used to reduce memory usage
3752 Description: Flush dirty partially truncated pages during truncate
3753 Details : Immediatelly flush partially truncated pages in filter_setattr,
3754 this way we completely avoid having any pages in page cache on OST
3755 and can retire ugly workarounds during writes to flush such pages.
3760 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
3761 Details : protect inode from truncation within vfs_unlink() context
3762 just take a reference before calling vfs_unlink() and release it
3763 when parent's i_sem is free.
3768 Description: last_id value checked outside lock on OST caused LASSERT failure
3769 Details : If there were multiple MDS->OST object precreate requests in
3770 flight, it was possible that the OST's last object id was checked
3771 outside a lock and incorrectly tripped an assertion. Move checks
3772 inside locks, and discard old precreate requests.
3775 Frequency : always, if extents are used on OSTs
3777 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
3778 Details : OSTs using the extents format would not zero the high 16 bits of
3779 the index physical block number. This is not a problem for any
3780 OST filesystems smaller than 16TB, and no kernels support ext3
3781 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
3782 new/modified files) and can be fixed for existing filesystems
3783 with e2fsprogs-1.39-cfs1.
3788 Description: import connection selection may be incorrect if timer wraps
3789 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
3790 import connections to be ignored if the 32-bit jiffies counter
3791 wraps. Use a 64-bit jiffies counter.
3794 Frequency : very large clusters immediately after boot
3796 Description: LNET request buffers exhausted under heavy short-term load
3797 Details : If a large number of client requests are generated on a service
3798 that has previously never seen so many requests it is possible
3799 that the request buffer growth cannot keep up with the spike in
3800 demand. Instead of dropping incoming requests, they are held in
3801 the LND until the RPC service can accept more requests.
3804 Frequency : Sometimes during replay
3806 Description: Assertion failure in ll_local_open after replay.
3807 Details : If replay happened on an open request reply before we were able
3808 to set replay handler, reply will become not swabbed tripping the
3809 assertion in ll_local_open. Now we set the handler right after
3810 recognising of open request
3813 Frequency : very rare
3815 Description: kernel reports "badness in vsnprintf"
3816 Details : Reading from the "recovery_status" /proc file in small chunks
3817 may cause a negative length in lprocfs_obd_rd_recovery_status()
3818 call to vsnprintf() (which is otherwise harmless). Exit early
3819 if there is no more space in the output buffer.
3821 Severity : enhancement
3823 Description: clear OBD RPC statistics by writing to them
3824 Details : It is now possible to clear the OBD RPC statistics by writing
3825 to the "stats" file.
3830 Description: Client mtime is not the same on different clients after utimes
3831 Details : In some cases, the client was using the utimes() syscall on
3832 a file cached on another node. The clients now validate the
3833 ctime from the MDS + OSTs to determine which one is right.
3838 Description: Inability to activate failout mode
3839 Details : lconf script incorrectly assumed that in pythong string's numeric
3840 value is used in comparisons.
3843 Frequency : always with multiple stripes per file
3845 Description: Inefficient object allocation for mutli-stripe files
3846 Details : When selecting which OSTs to stripe files over, for files with
3847 a stripe count that divides evenly into the number of OSTs,
3848 the MDS is always picking the same starting OST for each file.
3849 Return the OST selection heuristic to the original design.
3854 Description: mount failures may take full timeout to return an error
3855 Details : Under some heavy load conditions it is possible that a
3856 failed mount can wait for the full obd_timeout interval,
3857 possibly several minutes, before reporting an error.
3858 Instead return an error as soon as the status is known.
3860 Frequency : quota enabled and large files being deleted
3862 Description: releasing more than 4GB of quota at once hangs OST
3863 Details : If a user deletes more than 4GB of files on a single OST it
3864 will cause the OST to spin in an infinite loop. Release
3865 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3870 Description: statfs data retrieved from /proc may be stale or zero
3871 Details : When reading per-device statfs data from /proc, in the
3872 {kbytes,files}_{total,free,avail} files, it may appear
3873 as zero or be out of date.
3876 Frequency : systems with MD RAID1 external journal devices
3878 Description: lconf's call to blkid is confused by RAID1 journal devices
3879 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3880 of returning all block devices that match the journal UUID.
3883 Frequency : always, for aggregate stripe size over 4GB
3885 Description: assertion fails when trying to use 4GB stripe size
3886 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
3887 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
3890 Frequency : always on ppc64
3892 Description: the first write on an ext3 filesystem with mballoc got stuck
3893 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3894 perform endianness conversion.
3896 ------------------------------------------------------------------------------
3898 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
3900 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
3901 INTEROPERATE with older versions automatically. Please read the
3902 user documentation before upgrading any part of a live system.
3903 * WARNING: Lustre networking configuration changes are required with
3904 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
3907 * Support for kernels:
3908 2.6.9-22.0.2.EL (RHEL 4)
3909 2.6.5-7.244 (SLES 9)
3910 2.6.12.6 vanilla (kernel.org)
3913 Severity : enhancement
3914 Bugzilla : 7981/8208
3915 Description: Introduced Lustre Networking (LNET)
3916 Details : LNET is new networking infrastructure for Lustre, it includes
3917 a reorganized network configuration mode (see the user
3918 documentation for full details) as well as support for routing
3919 between different network fabrics. Lustre Networking Devices
3920 (LNDs) for the supported network fabrics have also been
3921 created for this new infrastructure.
3923 Severity : enhancement
3924 Description: Introduced Access control lists
3925 Details : clients can set ACLs on files and directories in order to have
3926 more fine-grained permissions than the standard Unix UGO+RWX.
3927 The MDS must be started with the "-o acl" mount option.
3929 Severity : enhancement
3930 Description: Introduced filesystem quotas
3931 Details : Administrators may now establish per-user quotas on the
3934 Severity : enhancement
3936 Description: Configuration change for the XT3
3937 The PTLLND is now used to run Lustre over Portals on the XT3
3938 The configure option(s) --with-cray-portals are no longer used.
3939 Rather --with-portals=<path-to-portals-includes> is used to
3940 enable building on the XT3. In addition to enable XT3 specific
3941 features the option --enable-cray-xt3 must be used.
3946 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
3947 Details : A race between allocating a new llcd and re-getting the llcd_lock
3948 allowed another thread to grab newly-allocated llcd.
3950 Severity : enhancement
3952 Description: 2.6 OST async journal commit and locking fix to improve performance
3953 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
3954 2.6 kernels are now async as they already were in 2.4 kernels so
3955 that they can commit concurrently with the network bulk transfer.
3956 For block-allocated files the filter allocation semaphore is held
3957 to avoid filesystem fragmentation during allocation. BKL lock
3958 removed for 2.6 xattr operations where it is no longer needed.
3963 Description: lconf incorrectly determined whether two IP networks could talk
3964 Details : In some more complicated routing and multiple-network
3965 configurations, lconf will avoid trying to make a network
3966 connection to a disjoint part of the IP space. It was doing the
3967 math incorrectly for one set of cases.
3972 Description: Fix for potential infinite loop processing records in an llog.
3973 Details : If an llog record is corrupted/zeroed, it is possible to loop
3974 forever in llog_process(). Validate the llog record length
3975 and skip the remainder of the block on error.
3978 Frequency : occasional (liblustre only)
3980 Description: liblustre could not open files whose last component is a symlink
3981 Details : sysio_path_walk() would incorrectly pass the open intent to
3982 intermediate path components.
3985 Frequency : rare (liblustre only with non-standard tuning)
3986 Bugzilla : 7201 (7350)
3987 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
3988 Details : llu_lookup_finish_locks() tries to set lock data on a lock
3989 after it has been released, only do this for referenced locks
3991 Severity : enhancement
3993 Description: specifying an (invalid) directory default stripe_size of -1
3994 would reset the directory default striping
3995 Details : stripe_size -1 was used internally to signal directory stripe
3996 removal, now use "all default" to signal dir stripe removal
3997 as a directory striping of "all default" is not useful
4000 Frequency : common for large clusters running liblustre clients
4002 Description: doing an ls when liblustre clients are running is slow
4003 Details : sending a glimpse AST to a liblustre client waits for every AST
4004 to time out, as liblustre clients will not respond. Since they
4005 cannot cache data we refresh the OST lock LVB from disk instead.
4007 Severity : enhancement
4009 Description: doing an ls at the same time as file IO can be slow
4010 Details : enqueue and other "small" requests can be blocked behind many
4011 large IO requests. Create a new OST IO portal for non-IO
4012 requests so they can be processed faster.
4015 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4017 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4018 created file with mode 000
4019 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4020 allow an ftruncate() as a non-root user to a file with mode 000.
4021 Lustre now respects this flag to disable mode checks when
4022 truncating a file owned by the user
4025 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4028 Description: Revoking locks from clients that went dead or catatonic might take
4030 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4031 cancellation of such locks instant on servers without waiting for
4032 any reply from clients. Clients drops these locks when cancel
4033 notification from server is received without replying.
4036 Frequency : liblustre-only, when liblustre client dies or becomes busy
4038 Description: Doing ls on Linux clients can take a long time with active
4040 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4041 granting such locks to it in the first place if possible. Locks
4042 are taken by proxy on the OST during the read or write and
4043 dropped immediately afterward. Add connect flags handling, do
4044 not grant locks to liblustre clients for glimpse ASTs.
4046 Severity : enhancement
4048 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4049 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4050 sequential reads and behaves badly with applications doing random
4051 reads. Improve it by reading ahead at least read region, and
4052 avoiding excessive large RPC for small reads.
4054 Severity : enhancement
4056 Description: Creating more than 1000 files for a single job may cause a load
4057 imbalance on the OSTs if there are also a large number of OSTs.
4058 Details : qos_prep_create() uses an OST index reseed value that is an
4059 even multiple of the number of available OSTs so that if the
4060 reseed happens in the middle of the object allocation it will
4061 still utilize the OSTs as uniformly as possible.
4066 Description: OST or MDS may oops in ping_evictor_main()
4067 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4068 but doesn't restart at beginning of obd_exports_timed list
4071 Severity : enhancement
4073 Description: improve by-nid export eviction on the MDS and OST
4074 Details : allow multiple exports with the same NID to be evicted at one
4075 time without re-searching the exports list.
4078 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4080 Description: MDS may oops in groups_free()
4081 Details : in rare race conditions a newly allocated group_info struct is
4082 freed again, and this can be NULL. The 2.4 compatibility code
4083 for groups_free() checked for a NULL pointer, but 2.6 did not.
4086 Frequency : common for liblustre clients doing little filesystem IO
4087 Bugzilla : 9352, 7313
4088 Description: server may evict liblustre clients accessing contended locks
4089 Details : if a client is granted a lock or receives a completion AST
4090 with a blocking AST already set it would not reply to the AST
4091 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4094 Frequency : lfs setstripe, only systems with more than 160 OSTs
4096 Description: unable to set striping with a starting offset beyond OST 160
4097 Details : llapi_create_file() incorrectly limited the starting stripe
4098 index to the maximum single-file stripe count.
4101 Frequency : LDAP users only
4103 Description: lconf did not handle in-kernel recovery with LDAP properly
4104 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4106 Severity : enhancement
4108 Description: bind OST threads to NUMA nodes to improve performance
4109 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4110 node and do their allocations there to localize memory access
4112 Severity : enhancement
4114 Description: llmount can determine client NID directly from Myrinet (GM)
4115 Details : the client NID code from gmnalnid was moved directly into
4116 llmount, removing the need to use this or specifying the
4117 client NID explicitly when mounting GM clients with zeroconf
4120 Frequency : if client is started with down MDS
4122 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4123 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4124 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4125 ptlrpc_import_delayed_req() because we didn't check if the
4126 request was interrupted, and we also didn't break out of the
4127 event loop if there was no timeout
4132 Description: data loss during non-page-aligned writes to a single file from
4133 both multiple nodes and multiple threads on one node at same time
4134 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4135 inconsistency led to false short-reads, that were cached and later
4136 used by ->prepare_write() to fill in partially written page,
4137 leading to data loss.
4140 Frequency : always, if lconf --abort_recovery used
4142 Description: lconf --abort_recovery fails with 'Operation not supported'
4143 Details : lconf was attempting to abort recovery on the MDT device and not
4146 Severity : enhancement
4148 Description: remove cleanup logs
4149 Details : replace lconf-generated cleanup logs with lustre internal
4150 cleanup routines. Eliminates the need for client-cleanup and
4153 Severity : enhancement
4155 Description: add support for EAs (user and system) on lustre filesystems
4156 Details : it is now possible to store extended attributes in the Lustre
4157 client filesystem, and with the user_xattr mount option it
4158 is possible to allow users to store EAs on their files also
4160 Severity : enhancement
4162 Description: Add possibility (config option) to show minimal available OST free
4164 Details : When compiled with --enable-mindf configure option, statfs(2)
4165 (and so, df) will return least minimal free space available from
4166 all OSTs as amount of free space on FS, instead of summary of
4167 free spaces of all OSTs.
4169 Severity : enhancement
4171 Description: do not expand extent locks acquired on OST-side
4172 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4173 by server: they are not cached anyway.
4176 Frequency : when mmap is used/binaries executed from Lustre
4178 Description: Unmmap pages before throwing them away from read cache.
4179 Details : llap_shrink cache now attempts to unmap pages before discarding
4180 them (if unmapping failed - do not discard). SLES9 kernel has
4181 extra checks that trigger if this unmapping is not done first.
4186 Description: lconf didn't resolve symlinks before checking to see whether a
4187 given mountpoint was already in use
4190 Frequency : when migrating failover services
4191 Bugzilla : 6395, 9514
4192 Description: When migrating a subset of services from a node (e.g. failback
4193 from a failover service node) the remaining services would
4194 time out and evict clients.
4195 Details : lconf --force (implied by --failover) sets the global obd_timeout
4196 to 5 seconds in order to quickly disconnect, but this caused
4197 other RPCs to time out too quickly. Do not change the global
4198 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4200 Severity : enhancement
4201 Frequency : if MDS is started with down OST
4202 Bugzilla : 9439,5706
4203 Description: Allow startup/shutdown of an MDS without depending on the
4204 availability of the OSTs.
4205 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4206 Add appropriate locking and lov-osc refcounts for safe
4207 cleaning. Add osc abort_inflight calls in case the
4208 synchronize never started.
4211 Frequency : occasional (Cray XT3 only)
4213 Description: root not authorized to access files in CRAY_PORTALS environment
4214 Details : The client process capabilities were not honoured on the MDS in
4215 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4216 been accessed by an authorized user then root was able to access
4217 the file on the local client also. The root user capabilities
4218 are now allowed on the MDS, as this environment has secure UID.
4221 Frequency : occasional
4223 Description: ldiskfs "too long searching" message happens too often
4224 Details : A debugging message (otherwise harmless) prints too often on
4225 the OST console. This has been reduced to only happen when
4226 there are fragmentation problems on the filesystem.
4231 Description: Division by zero in statfs when all OSCs are inactive
4232 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4233 lov_statfs divides by value returned by lov_get_stripecnt().
4237 Bugzilla : 9489, 3273
4238 Description: First write from each client to each OST was only 4kB in size,
4239 to initialize client writeback cache, which caused sub-optimal
4240 RPCs and poor layout on disk for the first writen file.
4241 Details : Clients now request an initial cache grant at (re)connect time
4242 and so that they can start streaming writes to the cache right
4243 away and always do full-sized RPCs if there is enough data.
4244 If the OST is rebooted the client also re-establishes its grant
4245 so that client cached writes will be honoured under the grant.
4250 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4251 Details : Now I/O RPCs go to different portal number and (presumably) fast
4252 lock requests (and glimses) and other RPCs get their own service
4253 threads pool that should be able to service those RPCs
4256 Severity : enhancement
4258 Description: Ability to exchange lustre version between client and servers and
4259 issue warnings at client side if client is too old. Also for
4260 liblustre clients there is ability to refuse connection of too old
4262 Details : New 'version' field is added to connect data structure that is
4263 filled with version info. That info is later checked by server and
4267 Frequency : rare, liblustre only.
4268 Bugzilla : 9296, 9581
4269 Description: Two simultaneous writes from liblustre at offset within same page
4270 might proceed at the same time overwriting eachother with stale
4272 Details : I/O lock withing llu_file_prwv was released too early, before data
4273 actually was hitting the wire. Extended lock-holding time until
4274 server acknowledges receiving data.
4277 Frequency : extremely rare. Never observed in practice.
4279 Description: avoid generating lustre_handle cookie of 0.
4280 Details : class_handle_hash() generates handle cookies by incrementing
4281 global counter, and can hit 0 occasionaly (this is unlikely, but
4282 not impossible, because initial value of cookie counter is
4283 selected randonly). Value of 0 is used as a sentinel meaning
4284 "unassigned handle" --- avoid it. Also coalesce two critical
4285 sections in this function into one.
4287 Severity : enhancement
4289 Description: allow liblustre clients to delegate truncate locking to OST
4290 Details : To avoid overhead of locking, liblustre client instructs OST to
4291 take extent lock in ost_punch() on client's behalf. New connection
4292 flag is added to handle backward compatibility.
4294 Severity : enhancement
4295 Bugzilla : 4928, 7341, 9758
4296 Description: allow number of OST service threads to be specified
4297 Details : a module parameter allows the number of OST service threads
4298 to be specified via "options ost ost_num_threads={N}" in the
4299 OSS's /etc/modules.conf or /etc/modprobe.conf.
4303 Bugzilla : 6146, 9635, 9895
4304 Description: servers crash with bad pointer in target_handle_connect()
4305 Details : In rare cases when a client is reconnecting it was possible that
4306 the connection request was the last reference for that export.
4307 We would temporarily drop the export reference and get a new
4308 one, but this may have been the last reference and the export
4309 was just destroyed. Get new reference before dropping old one.
4311 Severity : enhancement
4312 Frequency : if client is started with failover MDS
4314 Description: Allow multiple MDS hostnames in the mount command
4315 Details : Try to read the configuration from all specified MDS
4316 hostnames during a client mount in case the "primary"
4319 Severity : enhancement
4321 Description: Stop sending data to evicted clients as soon as possible.
4322 Details : Check if the client we are about to send or are sending data to
4323 was evicted already. (Check is done every second of waiting,
4324 for which l_wait_event interface was extended to allow checking
4325 of exit condition at specified intervals).
4328 Frequency : rare, normally only when NFS exporting is done from client
4330 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4332 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4335 Frequency : always, if srand() is called before liblustre initialization
4337 Description: Liblustre uses system PRNG disturbing its usage by user application
4338 Details : Introduce internal to lustre fast and high-quality PRNG for
4339 lustre usage and make liblustre and some other places in generic
4340 lustre code to use it.
4342 Severity : enhancement
4343 Bugzilla : 9477, 9557, 9870
4344 Description: Verify that the MDS configuration logs are updated when xml is
4345 Details : Check if the .xml configuration logs are newer than the config
4346 logs stored on the MDS and report an error if this is the case.
4347 Request --write-conf, or allow starting with --old_conf.
4349 Severity : enhancement
4351 Description: Handle symlinks in the path when checking if Lustre is mounted.
4352 Details : Resolve intermediate symlinks when checking if a client has
4353 mounted a filesystem to avoid duplicate client mounts.
4358 Description: lconf can hit an error exception but still return success.
4359 Details : The lconf command catches the Command error exception at the top
4360 level script context and will exit with the associated exit
4361 status, but doesn't ensure that this exit status is non-zero.
4366 Description: failure of ptlrpc thread startup can cause oops
4367 Details : Starting a ptlrpc service thread can fail if there are a large
4368 number of threads or the server memory is very fragmented.
4369 Handle this without oopsing.
4372 Frequency : always, only if liblustre and non-default acceptor port was used
4374 Description: liblustre cannot connect to servers with non-default acceptor port
4375 Details : tcpnal_set_default_params() was not called and was therefore
4376 ignoring the environment varaible TCPNAL_PORT, as well as other
4377 TCPNAL_ environment variables
4382 Description: two objects could be created on the same OST for a single file
4383 Details : If an OST is down, in some cases it was possible to create two
4384 objects on a single OST for a single file. No problems other
4385 than potential performance impact and spurious error messages.
4389 Bugzilla : 5681, 9562
4390 Description: Client may oops in ll_unhash_aliases
4391 Details : Client dcache may become inconsistent in race condition.
4392 In some cases "getcwd" can fail if the current directory is
4398 Description: Inode refcounting problems in NFS export code
4399 Details : link_raw functions used to call d_instantiate without obtaining
4400 extra inode reference first.
4404 Bugzilla : 9942, 9903
4405 Description: Referencing freed requests leading to crash, memleaks with NFS.
4406 Details : We used to require that call to ll_revalidate_it was always
4407 followed by ll_lookup_it. Also with revalidate_special() it is
4408 possible to call ll_revalidate_it() twice for the same dentry
4409 even if first occurence returned success. This fix changes semantic
4410 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4411 reference on a request referred from the intent.
4412 ll_intent_release() then releases such a request.
4415 Frequency : rare, normally benchmark loads only
4417 Description: unlinked inodes were kept in memory on the client
4418 Details : If a client is repeatedly creating and unlinking files it
4419 can accumulate a lot of stale inodes in the inode slab cache.
4420 If there is no other client load running this can cause the
4421 client node to run out of memory. Instead flush old inodes
4422 from client cache that have the same inode number as a new inode.
4425 Frequency : SLES9 2.6.5 kernel and long filenames only
4426 Bugzilla : 9969, 10379
4427 Description: utime reports stale NFS file handle
4428 Details : SLES9 uses out-of-dentry names in some cases, which confused
4429 the lustre dentry revalidation. Change it to always use the
4433 Frequency : rare, unless heavy write-truncate concurrency is continuous
4434 Bugzilla : 4180, 6984, 7171, 9963, 9331
4435 Description: OST becomes very slow and/or deadlocked during object unlink
4436 Details : filter_destroy() was holding onto the parent directory lock
4437 while truncating+unlinking objects. For very large objects this
4438 may block other threads for a long time and slow overall OST
4439 responsiveness. It may also be possible to get a lock ordering
4440 deadlock in this case, or run out of journal credits because of
4441 the combined truncate+unlink. Solution is to do object truncate
4442 first in one transaction without parent lock, and then do the
4443 final unlink in a new transaction with the parent lock. This
4444 reduces the lock hold time dramatically.
4447 Frequency : rare, 2.4 kernels only
4449 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4450 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4451 before unmounting their filesystem in order to avoid deadlock.
4452 The kernel_locked() function in 2.4 kernels only checks whether
4453 the kernel lock is held, not whether it is this process that is
4454 holding it as 2.6 kernels do.
4459 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4460 Details : The client ptlrpc code may be trying to reconnect to a down
4461 server before a previous connection attempt has timed out.
4462 Increase the reconnect interval to be longer than the connection
4463 timeout interval to avoid sending duplicate connections to
4467 Frequency : echo_client brw_test command
4469 Description: fix echo_client to work with OST preallocated code
4470 Details : OST preallocation code (5137) didn't take echo_client IO path
4471 into account: echo_client calls filter methods outside of any
4472 OST thread and, hence, there is no per-thread preallocated
4473 pages and buffers to use. Solution: hijack pga pages for IO. As
4474 a byproduct, this avoids unnecessary data copying.
4478 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4479 Description: Client can oops in mdc_commit_close() after open replay
4480 Details : It was possible for the MDS to return an open request with no
4481 transaction number in mds_finish_transno() if the client was
4482 evicted, but without actually returning an error. Clients
4483 would later try to replay that open and may trip an assertion
4484 Simplify the client close codepath, and always return an error
4485 from the MDS in case the open is not successful.
4488 Frequency : rare, 2.6 OSTs only
4490 Description: OST may deadlock under high load on fragmented files
4491 Details : If there was a heavy load and highly-fragmented OST filesystems
4492 it was possible to have all the OST threads deadlock waiting on
4493 allocation of biovecs, because the biovecs were not released
4494 until the entire RPC IO was completed. Instead, release biovecs
4495 as soon as they are complete to ensure forward IO progress.
4497 Severity : enhancement
4499 Description: Support for specifying external journal device at mount
4500 Details : If an OST or MDS device is formatted with an external journal
4501 device, this device major/minor is stored in the ext3 superblock
4502 and may not be valid for failover. Allow detecting and
4503 specifying the external journal at mount time.
4508 Description: Mounting an MDS with pending unlinked files may cause oops
4509 Details : target_finish_recovery() calls mds_postrecov() which returned
4510 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4511 considers this an error and immediately begins cleaning up the
4512 lov, just after starting the mds_lov process
4514 Severity : enhancement
4516 Description: Implement 'lfs df' to report actual free space on per-OST basis
4517 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4518 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4519 usage of objects; '-h' to report in human readable format.
4521 ------------------------------------------------------------------------------
4523 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
4530 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4531 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4532 at startup during orphan unlinking. Instead, skip buddy update
4533 before setup, as it will be regenerated after recovery is complete.
4538 Description: If an OST is inactive, its locks might reference stale inodes.
4539 Details : lov_change_cbdata() must iterate over all namespaces, even if
4540 they are inactive to clear inode references from the lock.
4542 Severity : enhancement
4543 Frequency : occasional, if non-standard max_dirty_mb used
4545 Description: Client will block write RPCs if not enough grant
4546 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4547 then the client will block writes while waiting for another RPC
4548 to complete instead of consuming its dirty limit. With change
4549 we get improved performance when max_dirty_mb is small.
4551 Severity : enhancement
4552 Bugzilla : 3389, 6253
4553 Description: Add support for supplementary groups on the MDS.
4554 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4555 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4556 lookups for user supplementary groups into a cache.
4560 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4561 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4562 for liblustre O_EXCL is enforced by MDS.
4566 Description: readdir never returns NULL in liblustre.
4567 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4568 dentry in struct dirent.
4572 Frequency : liblustre only, depends on application IO pattern
4573 Description: liblustre clients evicted if not contacting servers
4574 Details : Don't put liblustre clients into the ping_evictor list, so
4575 they will not be evicted by the pinger ever.
4577 Severity : enhancement
4579 Description: Add ability to evict clients by NID from MDS.
4580 Details : By echoing "nid:$NID" string into
4581 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4582 $NID would be instantly evicted from this MDS and from all active
4583 OSTs connected to it.
4587 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4588 Details : lookup_it_finish() used to query file size from OSTs that was not
4593 Description: service threads change working directory to that of init
4594 Details : Starting lustre service threads may pin the working directory
4595 of the parent thread, making that filesystem busy. Threads
4596 now change to the working directory of init to avoid this.
4600 Frequency : during shutdown only
4601 Description: shutdown with a failed MDS or OST can cause unmount to hang
4602 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4607 Frequency : 2.6 only
4608 Description: chmod/chown may include an extra supplementary group
4609 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4610 supplementary group and if none is specified this is used.
4613 Bugzilla : 5479 (6816)
4615 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4616 Details : If lookup is in progress on a file that is unlinked we might try
4617 to revalidate the inode and fail in revalidate after lookup is
4618 complete and ll_file_open() enqueues the open again but
4619 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4622 Frequency : always, if lconf --abort_recovery used
4624 Description: lconf --abort_recovery fails with 'Operation not supported'
4625 Details : lconf was attempting to abort recovery on the MDT device and not
4628 ------------------------------------------------------------------------------
4630 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4635 Frequency : rare (only unsupported configurations with a node running as an
4637 Bugzilla : 6514, 5137
4638 Description: Mounting a Lustre file system on a node running as an OST could
4640 Details : OSTs now preallocates memory needed to write out data at
4641 startup, instead of when needed, to avoid having to
4642 allocate memory in possibly low memory situations.
4643 Specifically, if the file system is mounted on on OST,
4644 memory pressure could force it to try to write out data,
4645 which it needed to allocate memory to do. Due to the low
4646 memory, it would be unable to do so and the node would
4647 become unresponsive.
4649 Severity : enhancement
4651 Description: Addition of lconf --service command line option
4652 Details : lconf now accepts a '--service <arg>' option, which is
4653 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4655 Severity : enhancement
4657 Description: Failover mode is now the default for OSTs.
4658 Details : By default, OSTs will now run in failover mode. To return to
4659 the old behaviour, add '--failout' to the lmc line for OSTs.
4661 Severity : enhancement
4663 Description: Health checks are now provided for MDS and OSTs
4664 Details : Additional detailed health check information on MSD and OSTs
4665 is now provided through the procfs health_check value.
4668 Frequency : occasional, depends on IO load
4670 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4671 after numerous create/delete cycles
4672 Details : The ext3 inode allocation policy would not allocate new inodes
4673 very well on the OSTs because there are no new directories
4674 being created. Instead we look for groups with free space if
4675 the parent directories are nearly full.
4680 Description: Network or server problems during mount may cause partially
4681 mounted clients instead of returning an error.
4682 Details : The config llog parsing code may overwrite the error return
4683 code during mount error handling, returning success instead
4689 Description: MDS can fail to allocate large reply buffers
4690 Details : After long uptimes the MDS can fail to allocate large reply
4691 buffers (e.g. zconf client mount config records) due to memory
4692 fragmentation or consumption by the buffer cache. Preallocate
4693 some large reply buffers so that these replies can be sent even
4694 under memory pressure.
4698 Frequency : rare (liblustre)
4699 Description: fsx running with liblustre complained that using truncate() to
4700 extend the file doesn't work. This patch corrects that issue.
4701 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4702 ATTR_SIZE and lsm use in llu_setattr_raw.
4706 Frequency : rare, only 2.6 kernels
4707 Description: Unusual file access patterns on the MDS may result in inode
4708 data being lost in very rare circumstances.
4709 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4710 code in the 2.6 kernel caused the inode and/or EA data not to
4711 be read from disk, causing single-file corruption.
4715 Frequency : rare, only 2.6 filesystems using extents
4716 Description: Heavy concurrent write and delete load may cause data corruption.
4717 Details : It was possible under high-load situations to have an extent
4718 metadata block in the block device cache from a just-unlinked
4719 file overwrite a newly-allocated data block. We now unmap any
4720 metadata buffers that alias just-allocated data blocks.
4724 Frequency : filesystems with default stripe_count larger than 77
4725 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4726 Details : lconf specifies an inode size of 4096 bytes when the default
4727 stripe_count is larger than 77. This conflicts with the default
4728 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4729 this case to avoid pinning too much memory for large EAs.
4731 ------------------------------------------------------------------------------
4733 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4738 Frequency : rare (extremely heavy IO load with hundreds of clients)
4740 Description: Client is evicted, gets IO error writing to file
4741 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
4742 caused clients to be evicted for AST timeouts. The fixes in
4743 bug 5192 mean we no longer need to have such short AST timeouts
4744 so ldlm_timeout has been increased.
4747 Frequency : occasional during --force or --failover shutdown under load
4748 Bugzilla : 5949, 4834
4749 Description: Server oops/LBUG if stopped with --force or --failover under load
4750 Details : a collection of import/export refcount and cleanup ordering
4751 issues fixed for safer force cleanup
4754 Frequency : only filesystems larger than 120 OSTs
4755 Bugzilla : 5990, 6223
4756 Description: lfs getstripe would oops on a very large filesystem
4757 Details : lov_getconfig used kfree on vmalloc'd memory
4760 Frequency : only filesystems exporting via NFS to Solaris 10 clients
4761 Bugzilla : 6242, 6243
4762 Description: reading from files that had been truncated to a non-zero size
4763 but never opened returned no data
4764 Details : ll_file_read() reads zeros from no-object files to EOF
4769 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
4770 Details : There's little harm in aborting MDS/OSS recovery and letting it
4771 try again, so I removed the LASSERT and return an error instead.
4773 Severity : enhancement
4775 Description: New debugging infrastructure for tracking down data corruption
4776 Details : The I/O checksum code was replaced to: (a) control it at runtime,
4777 (b) cover more of the client-side code path, and (c) try to narrow
4778 down where problems occurred
4782 Bugzilla : 3819, 4364, 4397, 6313
4783 Description: Racing close and eviction MDS could cause assertion in mds_close
4784 Details : It was possible to get multiple mfd references during close and
4785 client eviction, leading to one thread referencing a freed mfd.
4787 Severity: : enhancement
4788 Bugzilla : 3262, 6359
4789 Description: Attempts to reconnect to servers are now more aggressive.
4790 Details : This builds on the enhanced upcall-less recovery that was added
4791 in 1.4.2. When trying to reconnect to servers, clients will
4792 now try each server in the failover group every 10 seconds. By
4793 default, clients would previously try one server every 25 seconds.
4798 Description: After recovery, certain operations trigger a failed
4799 assertion on a client.
4800 Details : Failing over an mds, using lconf -d --failover, while a
4801 client was doing a readdir() call would cause the client to
4802 LBUG after recovery completed and the readdir() was resent.
4804 Severity : enhancement
4806 Description: Default groups are now added by lconf
4807 Details : You can now run lconf --group <servicename> without having to
4808 manually add groups with lmc.
4811 Frequency : occasional
4813 Description: Nodes with an elan id of 0 trigger a failed assertion
4816 Frequency : always when accessing e.g. tty/console device nodes
4818 Description: tty and some other devices nodes cannot be used on lustre
4819 Details : file's private_data field is used by device data and lustre
4820 values in there got lost. New field was added to struct file to
4821 store fs-specific private data.
4824 Frequency : when exporting Lustre via NFS
4826 Description: NFSD failed occasionally when looking up a path component
4827 Details : NFSD is looking up ".." which was broken in ext3 directories
4828 that had grown large enough to become hashed.
4831 Frequency : Clusters with multiple interfaces not on the same subnet
4833 Description: Nodes will repeatedly try to reconnect to an interface which it
4834 cannot reach and report an error to the log.
4835 Details : Extra peer list entries will be created by lconf with some peers
4836 unreachable. lconf now validates the peer before adding it.
4839 Frequency : Only if a default stripe is set on the filesystem root.
4841 Description: Setting a default stripe on the filesystem root prevented the
4842 filesystem from being remounted.
4843 Details : The client was sending extra request flags in the root getattr
4844 request and did not allocate a reply buffer for the dir EA.
4847 Frequency : occasional, higher if lots of files are accessed by one client
4848 Bugzilla : 6159, 6097
4849 Description: Client trips assertion regarding lsm mismatch/magic
4850 Details : While revalidating inodes the VFS looks up inodes with ifind()
4851 and in rare cases can find an inode that is being freed.
4852 The ll_test_inode() code will free the lsm during ifind()
4853 when it finds an existing inode and then the VFS later attaches
4854 this free lsm to a new inode.
4858 Bugzilla : 6422, 7030
4859 Description: MDS deadlock between mkdir and client eviction
4860 Details : Creating a new file via mkdir or mknod (starting a transaction
4861 and getting the ns lock) can deadlock with client eviction
4862 (gets ns lock and trying to finish a synchronous transaction).
4865 Frequency : occasional
4866 Description: While starting a server, the fsfilt_ext3 module could not be
4868 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
4869 kernels. Previously, lconf would still use the ext3 name
4870 when trying to load modules. Now, it will correctly use
4871 ext3 on 2.4 and ldiskfs on 2.6.
4873 Severity : enhancement
4874 Description: The default stripe count has been changed to 1
4875 Details : The interpretation of the default stripe count (0, to lfs
4876 or lmc) has been changed to mean striping across a single
4877 OST, rather than all available. For general usage we have
4878 found a stripe count of 1 or 2 works best.
4880 Severity : enhancement
4881 Description: Add support for compiling against Cray portals.
4882 Details : Conditional compiling for some areas that are different
4886 Frequency : occasional
4887 Bugzilla : 6409, 6834
4888 Description: Creating files with an explicit stripe count may lead to
4889 a failed assertion on the MDS
4890 Details : If some OSTs are full or unavailable, creating files may
4891 trigger a failed assertion on the MDS. Now, Lustre will
4892 try to use other servers or return an error to the
4896 Frequency : occasional
4898 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
4899 caused lock timeout during readahead (since 1.4.2).
4900 Details : Processes doing readahead might match a lock that hasn't been
4901 granted yet if there are overlapping and conflicting lock
4902 requests. The readahead process waits on ungranted lock
4903 (original lock is CBPENDING), while OST waits for that process
4904 to cancel CBPENDING read lock and eventually evicts client.
4906 Severity : enhancement
4908 Description: Initial enabling of flock support for clients
4909 Details : Implements fcntl advisory locking and file status functions.
4910 This feature is provided as an optional mount flag (default
4911 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
4912 locking are implemented yet, and those that are are not guaranteed
4913 to be completely correct in production environments.
4914 mount -t lustre -o [flock|noflock] ...
4917 Frequency : occasional
4919 Description: OSTs running 2.4 kernels but with extents enabled might trip an
4920 assertion in the ext3 JBD (journaling) layer.
4921 Details : The b_committed_data struct is protected by the big kernel lock
4922 in 2.4 kernels, serializing journal_commit_transaction() and
4923 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
4924 access to this struct. In 2.6 kernels there is finer grained
4925 locking to improve SMP performance of the JBD layer.
4929 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
4934 Description: merge and cleanup kernel patches.
4937 -----------------------------------------------------------------------------
4939 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
4941 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
4942 versions of Lustre. Please update all servers and clients to
4943 version 1.4.2 or later at the same time. You must also run
4944 "lconf --write-conf {config}.xml" on the MDS while it is stopped
4945 to update the configuration logs.
4947 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
4948 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
4949 - swab llog records of type '0' so we get proper header size/idx (5861)
4950 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
4951 - fix rename of one directory over another leaking an inode (5953)
4952 - avoid SetPageDirty on 2.6 (5981)
4953 - don't re-add just-being-destroyed locks to the waiting list (5653)
4954 - when creating new directories, inherit the parent's custom
4955 striping settings if present parent (3048)
4956 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
4957 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
4958 - don't hold client locks on temporary worklist from l_lru (5666)
4959 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
4960 - automatically evict dead clients (5921)
4961 - Update file size properly in create+truncate+fstat case (6196)
4962 - Do not unhash mountpoint dentries, do not allow removal of
4964 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
4965 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
4966 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
4967 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
4968 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
4969 - Correctly handle reads of files with no objects (6243)
4970 - lctl recover will also mark a device active if deactivate used (5933)
4972 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
4973 - allow --write-conf on an MDS with different nettype than client (5619)
4974 - don't write config llogs to MDS for mounts not from that MDS (5617)
4975 - lconf should create multiple TCP connections from a client (5201)
4976 - init scripts are now turned off by default; run chkconfig --on
4977 lustre and chkconfig --on lustrefs to use them
4978 - upcalls are no longer needed for clients to recover to failover
4980 - add --abort-recovery option to lconf to abort recovery on device
4982 - add support for an arbitrary number of OSTs (3026)
4983 - Quota support protocol changes.
4984 - forward compatibility changes to wire structs (6007)
4985 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
4986 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
4987 - improved "lustre status" script
4988 - initialize blocksize for non-regular files (6062)
4989 - added --disable-server and --disable-client configure options (5782)
4990 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
4991 - Vanilla 2.4.29 support
4992 - increase maximum number of obd devices to 520 (6242)
4993 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
4994 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
4995 SLES 9 2.6.5 kernels
4996 - limit stripes per file to 160 (the maximum EA size) (6093)
4998 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5001 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5002 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5003 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5004 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5005 - small ext3 extents cleanups and fixes (5733)
5006 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5007 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5008 - avoid needless client->OST connect, fix handle mismatch (5317)
5009 - fix DLM error path that led to out-of-sync client, long delays (5779)
5010 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5011 - fix several locking issues related to i_size (5492,5624,5654,5672)
5012 - don't move pending lock onto export if it is already evicted (5683)
5013 - fix kernel oops when creating .foo in unlinked directory (5548)
5014 - fix deadlock in obdfilter statistics vs. object create (5811)
5015 - use time_{before,after} to avoid timer jiffies wrap (5882)
5016 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5017 - Do not leak request if server was not able to process it (5154)
5018 - If mds_open unable to find parent dir, make that negative lookup(5154)
5019 - don't create new directories with extent-mapping (5909, 5936)
5021 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5022 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5023 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5024 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5025 - support for RHEL4 kernel on i686 (5773)
5026 - provide error messages when incompatible logs are encountered (5898)
5028 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5029 * version 1.4.0.10 (1.4.1 release candidate 1)
5031 - don't keep a lock reference when lock is not granted (4238)
5032 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5033 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5034 - drop import inflight refcount on signal_completed_replay error (5255)
5035 - unlock page after async write error during send (3677)
5036 - handle missing objects in filter_preprw_read properly (5265)
5037 - no transno return for symlink open, don't save no-trasno open (3440)
5038 - don't try to complete elan receive that already failed (4012)
5039 - free RPC server reply state on error (5406)
5040 - clean up thread from ptlrpc_start_thread() on error (5160)
5041 - readahead could read extra page into cache that wasn't ejected (5388)
5042 - prevent races in class_attach/setup/cleanup/detach (5260)
5043 - don't dereference de->d_inode after l_dput of de (5458)
5044 - use "int" for stripe value returned from lock_to_stripe (5544)
5045 - mballoc allocation and error-checking fixes in 2.6 (5504)
5046 - block device patches to fix I/O request sizes in 2.6 (5482)
5047 - look up hostnames for IB nals (5602)
5048 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5049 - don't start multiple acceptors for the same port (5277)
5050 - fix incorrect LASSERT in mds_getattr_name (5635)
5051 - export a proc file for general "ping" checking (5628)
5052 - fix "lfs check" to not block when the MDS is down (5628)
5054 - service request history (4965)
5055 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5056 - create an "evict_client" /proc entry on OSTs, like the MDS has
5057 - fix mount usage message, return errors per mount(8) (5168)
5058 - change grep [] to grep "[]" in tests so they work in more UMLs
5059 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5060 - remove ancient LOV_MAGIC_V0 EA support (5047)
5061 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5062 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5063 - allow readahead allocations to fail when low on memory (5383)
5064 - mmap locking landed again, after considerable improvement (2828)
5065 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5067 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5070 - send OST transaction number in read/write reply to free req (4966)
5071 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5072 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5073 - the watchdog thread now runs as interruptible (5246)
5074 - flock/lockf fixes (but it's still disabled, pending 5135)
5075 - don't use EXT3 constants in llite code (5094)
5076 - memory shortage at startup could cause assertion (5176)
5078 - reorganization of lov code
5079 - single portals codebase
5081 - add extents/mballoc support (5025)
5082 - direct I/O reads in the obdfilter (4048)
5083 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5085 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5088 - send OST transaction number in read/write reply to free req (4966)
5089 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5090 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5091 - the watchdog thread now runs as interruptible (5246)
5092 - handle missing objects in filter_preprw_read properly (5265)
5093 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5094 - drop import inflight refcount on signal_completed_replay error (5255)
5095 - unlock page after async write error during send (3677)
5096 - return original error code on reconstructed replies (3761)
5097 - no transno return for symlink open, don't save no-trasno open (3440)
5099 - add pid to ldlm debugging output (4922)
5100 - bump the watchdog timeouts -- we can't handle 30sec yet
5101 - extra debugging for orphan dentry/inode bug (5259)
5103 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5106 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5107 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5108 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5109 - fix NULL dereference in /proc/sys/portals/routes (4827)
5110 - allow failed mdc_close() operations to be interrupted (4561)
5111 - stop precreate on OST before MDS would time out on it (4778)
5112 - don't send partial-page writes before EOF from client (4410)
5113 - discard client grant for sub-page writes on large-page clients (4520)
5114 - don't free dentries not owned by NFS code, check generation (4806)
5115 - fix lsm leak if mds_create_objects() fails (4801)
5116 - limit debug_daemon file size, always print CERROR messages (4789)
5117 - use transno after validating reply (3892)
5118 - process timed out requests if import state changes (3754)
5119 - update mtime on OST during writes, return in glimpse (4829)
5120 - add mkfsoptions to LDAP (4679)
5121 - use ->max_readahead method instead of zapping global ra (5039)
5122 - don't interrupt __l_wait_event() during strace
5124 - add software watchdogs to catch hung threads quickly (4941)
5125 - make lustrefs init script start after nfs is mounted
5126 - fix CWARN/ERROR duplication (4930)
5127 - return async write errors to application if possible (2248)
5128 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5129 - print NAL number in %x format (4645)
5130 - update barely-supported suse-2.4.21-171 series (4842)
5131 - support for sles 9 %post scripts
5132 - support for building 2.6 kernel-source packages
5133 - support for sles km_* packages
5135 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5138 - ignore -ENOENT errors in osc_destroy (3639)
5139 - notify osc create thread that OSC is being cleaned up (4600)
5140 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5141 - reconstruct ost_handle() like mds_handle() (4657)
5142 - create a new thread to do import eviction to avoid deadlock (3969)
5143 - let lconf resolve symlinked-to devices (4629)
5144 - don't unlink "objects" from directory with default EA (4554)
5145 - hold socknal file ref over connect in case target is down (4394)
5146 - allow more than 32000 subdirectories in a single directory (3244)
5147 - fix blocks count for O_DIRECT writes (3751)
5148 - OST returns ENOSPC from object create when no space left (4539)
5149 - don't send truncate RPC if file size isn't changing (4410)
5150 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5151 - bind to privileged port in socknal and tcpnal (3689)
5153 - rate limit CERROR/CWARN console message to avoid overload (4519)
5154 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5155 - basic mmap support (3918)
5156 - kernel patch series update from b1_4 (4711)
5158 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5161 - avoid crash during MDS cleanup with OST shut down (2775)
5162 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5163 - don't use bad inodes on the MDS (3744)
5164 - dynamic object preallocation to improve recovery speed (4236)
5165 - don't hold spinlock over lock dumping or change debug flags (4401)
5166 - don't zero obd_dev when it is force cleaned (3651)
5167 - print grants to console if they go negative (4431)
5168 - "lctl deactivate" will stop automatic recovery attempts (3406)
5169 - look for existing locks in ldlm_handle_enqueue() (3764)
5170 - don't resolve lock handle twice in recovery avoiding race (4401)
5171 - revalidate should check working dir is a directory (4134)
5173 - don't always mark "slow" obdfilter messages as errors (4418)
5175 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5178 - don't close LustreDB during write_conf until it is done (3860)
5179 - fix typo in lconf for_each_profile (3821)
5180 - allow dumping logs from multiple threads at one time (3820)
5181 - don't allow multiple threads in OSC recovery (3812)
5182 - fix debug_size parameters (3864)
5183 - fix mds_postrecov to initialize import for llog ctxt (3121)
5184 - replace config semaphore with spinlock (3306)
5185 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5186 - don't allow enqueue to complete on a destroyed export (3822)
5187 - down write_lock before checking llog header bitmap (3825)
5188 - recover from lock replay timeout (3764)
5189 - up llog sem before sending rpc (3652)
5190 - reduce ns lock hold times when setting kms (3267)
5191 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5192 - fix NULL deref and obd_dev leak on setup error (3312)
5193 - replace some LBUG about llog ops with error handling (3841)
5194 - don't match INVALID dentries from d_lookup and spin (3784)
5195 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5196 - fix invalid assertion in ptlrpc_set_wait (3880)
5198 - add libwrap support for the TCP acceptor (3996)
5199 - add /proc/sys/portals/routes for non-root route listing (3994)
5200 - allow setting MDS UUID in .xml (2580)
5201 - print the stack of a process that LBUGs (4228)
5203 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5206 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5207 - make sure to unset replay flag from failed open requests (3440)
5208 - if default stripe count is 0, use OST count for inode size (3636)
5209 - update parent mtime/ctime on client for create/unlink (2611)
5210 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5211 - free recovery state on server during a forced cleanup (3571)
5212 - unregister_reply for resent reqs (3063)
5213 - loop back devices mounting and status check on 2.6 (3563)
5214 - fix resource-creation race that can provoke i_size == 0 (3513)
5215 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5216 - more debugging for page-accounting assertion (3746)
5217 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5218 - avoid deadlock after precreation failure (3758)
5219 - fix race and lock order deadlock in orphan handling (3450, 3750)
5220 - add validity checks when grabbing inodes from l_ast_data (3599)
5222 - add /proc/.../recovery_status to obdfilter (3428)
5223 - lightweight CDEBUG infrastructure, debug daemon (3668)
5224 - change default OSC RPC parameters to be better on small clusters
5225 - turn off OST read cache for files smaller than 32MB
5226 - install man pages and include them in rpms (3100)
5227 - add new init script for (un)mounting lustre filesystems (2593)
5228 - run chkconfig in %post for init scripts (3701)
5229 - drop scimac NAL (unmaintained)
5231 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5234 - clean kiobufs before and after use (3485)
5235 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5236 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5237 - change init script to fail more clearly if not run as root (1528)
5238 - allow clients to reconnect during replay (1742)
5239 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5240 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5241 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5242 - fix cond_resched (3554)
5243 - only evict unfinished clients after recovery (3515)
5244 - allow bulk resend, prevent data loss (3570)
5245 - dynamic ptlrpc request buffer allocation (2102)
5246 - don't allow unlinking open directory if it isn't empty (2904)
5247 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5248 - remove extraneous obd dereference causing LASSERT failure (3334)
5249 - don't use get_cycles() when creating temp. files on the mds (3156)
5250 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5251 - handle EEXIST for set-stripe, set proper directory name (3336)
5253 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5254 - fix ksocknal_fmb_callback() error messages (2918)
5256 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5259 - don't copy lvb into (possibly NULL) reply on error (2983)
5260 - don't deref dentry after dput, don't free lvb on error (2922)
5261 - use the kms to determine writeback rpc length (2947)
5262 - increment oti_logcookies when osc is inactive (2948)
5263 - update client's i_blocks count via lvb messages (2543)
5264 - handle intent open/close of special files properly (1557)
5265 - mount MDS with errors=remount-ro, like obdfilter (2009)
5266 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5267 - don't use cancelling-locks' kms values (2947)
5268 - use highest lock extent for kms, not last one (2925)
5269 - don't dereference ERR_PTR() dentry in error handling path (3107)
5270 - fix thread race in portals_debug_dumplog() (3122)
5271 - create lprocfs device entries at setup instead of at attach (1519)
5272 - common AST error handler, don't evict client on completion race (3145)
5273 - zero nameidata in detach_mnt in 2.6 (3118)
5274 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5275 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5276 - initialize RPC timeout timer earlier for 2.6 (3219)
5277 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5278 - print nal/nid for unknown nid (3258)
5279 - additional checks for oscc recovery before doing precreate (3284)
5280 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5281 - don't crash in mdc_close for bad permissions on open (3285)
5282 - zero i_rdev for non-device files (3147)
5283 - clear page->private before handing to FS, better assertion (3119)
5284 - tune the read pipeline (3236)
5285 - fix incorrect decref of invalidated dentry (2350)
5286 - provide read-ahead stats and refine rpc in flight stats (3328)
5287 - don't hold journal transaction open across create RPC (3313)
5288 - update atime on MDS at close time (3265)
5289 - close LDAP connection when recovering to avoid server load (3315)
5290 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5291 - don't leak open file on MDS after open resend (3325)
5292 - serialize filter_precreate and filter_destroy_precreated (3329)
5293 - loop device shouldn't call sync_dev() for nul device (3092)
5294 - clear page cache after eviction (2766)
5295 - resynchronize MDS->OST in background (2824)
5296 - refuse to mount the same filesystem twice on same mountpoint (3394)
5297 - allow llmount to create routes for mounting behind routers (3320)
5298 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5299 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5300 - fix rare problem with rename on htree directories (3417)
5302 - allow default OST striping configuration per directory (1414)
5303 - fix compilation for qswnal for 2.6 kernels (3125)
5304 - increase maximum number of MDS request buffers for large systems
5305 - change liblustreapi to be useful for external progs like lfsck (3098)
5306 - increase local configuration timeout for slow disks (3353)
5307 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5309 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5312 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5313 - don't overwrite extent policy data in reply if lock was blocked (2901)
5314 - drop filter export grants atomically with removal from device (2663)
5315 - del obd_self_export from work_list in class_disconnect_exports (2908)
5316 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5317 - swab reply message in mdc_close, other PPC fixes (2464)
5318 - fix destroying of named logs (2325)
5319 - overwrite old logs when running lconf --write_conf (2264)
5320 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5321 - fix race in target_handle_connect (2898)
5322 - mds_reint_create() should take same inode create lock (2926)
5323 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5324 - don't close files for self_export to avoid uninitialized obd (2936)
5325 - allow MDS with the same name as client node (2939)
5326 - hold dentry reference for closed log files for unlink (2325)
5327 - reserve space for all logs during transactions (2059)
5328 - don't evict page beyond end of stripe extent (2925)
5329 - don't oops on a deleted current working directory (2399)
5330 - handle hard links to targets without a parent properly (2517)
5331 - don't dereference NULL lock when racing during eviction (2867)
5332 - don't grow lock extents when lots of conflicting locks (2919)
5334 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5337 - account for cache space usage on clients to avoid data loss (974)
5338 - lfsck support in lustre kernel code (2349)
5339 - reduce journal credits needed for BRW writes (2370)
5340 - orphan handling to avoid losing space on client/server crashes
5341 - ptlrpcd can be blocked, stopping ALL progress (2477)
5342 - use lock value blocks to assist in proper KMS, faster stat (1021)
5343 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5344 - recovery for initial connections (2355)
5345 - fixes for mds_cleanup_orphans (1934)
5346 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5347 - block all file creations until orphan recovery completes (1901)
5348 - client remove rq_connection from request struct (2423)
5349 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5350 - recovery timer race (2670)
5351 - mdc_close recovey bug (2532)
5352 - ptlrpc cleanup bug (2710)
5353 - mds timeout on local locks (2588)
5354 - namespace lock held during RPCs (2431)
5355 - handle interrupted sync write properly (2503)
5356 - don't try to handle a message that hasn't been replied to (2699)
5357 - client assert failure during cleanup after abort recovery (2701)
5358 - leak mdc device after failed mount (2712)
5359 - ptlrpc_check_set allows timedout requests to complete (2714)
5360 - wait for inflight reqs when ptlrpcd finishes (2710)
5361 - make sure unregistered services are removed from the srv_list
5362 - reset bulk XID's when resending them (caught by 1138 test)
5363 - unregister_bulk after timeout
5364 - fix lconf error (2694)
5365 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5366 - readahead locks pages, leaves pending causing memory pressure (2673)
5367 - increase OST request buffers to 4096 on large machines (2729)
5368 - fix up permission of existing directories in simple_mkdir (2661)
5369 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5370 - don't assign transno to errored transactions (2742)
5371 - don't delete objects on OST if given a bogus objid from MDS (2751)
5372 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5373 - if rq_no_resend, then timeout request after recovery (2432)
5374 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5375 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5376 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5377 - MDS recovery shouldn't delete objects that it has given out (2730)
5378 - if enqueue arrives after completion, don't clobber LVB (2819)
5379 - don't unlock pages twice when trigger_group_io returns error (2814)
5380 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5381 - don't write pages to disk if there was an error (1450)
5382 - don't ping imports that have recovery disabled (2676)
5383 - take buffered bytes into account when balancing socknal conn (2817)
5384 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5385 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5386 - remove little-endian swabbing of llog records (1987)
5387 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5388 - retry reposting request buffers if they fail (1191)
5389 - grow extent at grant time to avoid granting a revoked lock (2809)
5390 - lock revoke doesn't evict page if covered by a second lock (2765)
5391 - disable VM readahead to avoid reading outside lock extents (2805)
5393 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5394 - updated kernel patches for hp-2.4.20 kernel (2681)
5396 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5399 - fix truncated write corruption (2366)
5400 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5402 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5403 - ensure len(uuid) < 37 in lmc (1171)
5404 - fix ia64 OOPS in llog_test (2255)
5405 - zero end of page at obdfilter for partial page writes (2648)
5406 - don't leave stale dentries around after renames (bug 2428)
5407 - fix timeouts when evicting a client with a single lock held (2642)
5408 - set deadline for the initial HELLO message to drain (2634)
5409 - print out dotted-quad IP addresses in the socknal (2302)
5411 - additional debugging for MDS client eviction problem (2443)
5412 - fix mkfsoptions support for osts (2603, 2604)
5414 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5417 - add series for the vanilla 2.6.0 kernel
5418 - add series for the vanilla 2.4.24 kernel
5419 - add series for a cray x86/64 UL kernel drop
5420 - fix xattr patches for the vanilla 2.4.19 series
5422 - generate true UUIDs in lmc (1171)
5423 - have portals stack dumping break in UML (2466)
5424 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5425 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5426 - wake callback waiting threads on client eviction (2460)
5427 - Add --ptldebug and --subsystem to lmc (1719)
5428 - update assertion to allow safe interrupt allocation
5429 - set rq_no_resend for cancel requests (2432)
5430 - recalculate ptlrpcd timeout after resend (2494)
5431 - call vfs_rmdir when removing pending directories (2368)
5432 - fix renaming a file to itself (2429)
5433 - lmc creates a default one-stripe lov (2454)
5434 - expand procfs space to handle large clusters (2326)
5435 - increase UML stack to avoid overflow
5436 - update lconf's list of debug and subsystem masks
5437 - fix lfs find --obd (2510)
5438 - /proc tunable for disabling filter read caching (2591)
5439 - stop rpm packages from altering slapd.conf (2301)
5440 - disable nagle in the socknal under 0conf (2578)
5441 - choose mds inode size based on stripe count (2572)
5442 - fix kernel-source rpm problems (2516)
5444 - add --disable-doc to avoid pdf generation (2421)
5445 - update documentation, tests, type-os, comments
5446 - avoid format warnings on ia64
5447 - remove the TOE NAL
5448 - tiny code cleanups by removing unused fields
5450 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5453 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5454 - remove the most blatant lies from BUILDING file (2371)
5455 - change default debug level to reasonable production setting
5456 - reduce client side cache size to reduce cache flush time
5457 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5458 - make TCP zerocopy and pinger support enabled by default (2476)
5459 - sync writes completed after process exits caused crashes (2319)
5460 - maintain correct mount count on the MDS (2356)
5461 - backout 1557, because 2316 wasn't really fixed
5462 - better file I/O statistics gathering in /proc
5463 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5464 - another kernel patch to fix zero-copy TCP function export
5465 - don't take duplicate lock when processing re-sent getattr (2420)
5466 - lctl uses obd_self_export instead of creating new conn (2353)
5467 - MDS/OST recovery case which requires object creation asserted (2425)
5468 - move lfs from /usr/sbin to /usr/bin in packages
5469 - fix race between mds_client_add and mds_client_free (2417)
5470 - use kmalloc instead of slabs in portals (2430)
5471 - don't create duplicate records when a failover MDS is present (2442)
5472 - remove unnecessary mount age check (2332)
5473 - don't remove directory inodes from locks prematurely (2451)
5474 - don't break if MDS service name is the same as hostname (2103)
5475 - fix races in client write RPC generation when cache full (2482)
5477 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5480 - remove now-unused request->rq_obd (278)
5481 - if an allocation fails, print out how much memory we've used (1933)
5482 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5483 - add the "configurable stack size" patch to most series files (1256)
5484 - ability to write large log records, for 100+ OST configs (2306)
5485 - fix NULL deref when filter_prep fails (2314)
5486 - fix operator precedence error in filter_sync
5487 - dynamic allocation of socknal TX descriptors (2315)
5488 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5489 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5490 - ensure that CWARN messages in Portals always get to the syslog
5491 - __init/__exit are not for prototype decls (ldlm_init/exit)
5492 - x86-64 compile warning fixes
5493 - fix gateway LMC keyword conflict (2318)
5494 - fix MDS lock inversions in getattr/reint paths (1844)
5495 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5496 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5497 - DLM race condition prevented some lock evictions (2328)
5498 - ENOMEM detection and retry on socknal sends (2230)
5499 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5500 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5501 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5502 - better error messages when a client is rejected during recovery (1505)
5503 - avoid cancelling locks which were never granted, after failure (2330)
5504 - fix i_sem/journal inversion in mds_client_add (2333)
5505 - fix truncate/getattr lock cycle deadlock (2334)
5506 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5507 - fix two rare exit paths which could leak an l_lock() ref (2321)
5508 - fencepost error in MDS/OST orphan recovery (2226)
5509 - make log record alignment 8 bytes (1988)
5510 - lstripe now fails when requested offset > ost_count (2237)
5511 - ensure that all kernel series have a complete list.h (1607)
5512 - fix crashes in special-file operations (2316)
5513 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5515 - allow configurable automake binary, for testing new versions
5516 - small update to the lfs documentation
5518 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5520 * fix negative export reference count in fsfilt_sync (2312)
5522 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5523 * release candidate 0.9.1
5525 - orphans are moved into the PENDING directory for possible recovery
5526 - replayed opens now open by fid for orphan/rename safety (1042)
5527 - last close of an orphan inode generates a transno (683)
5528 - chdir() and mount() now pin the directory entry (1020)
5529 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5530 - discard very old requests without processing them (1502)
5531 - remove some common, well-understood CERRORs (1505)
5532 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5533 - clear "grant" flags in OST replies until OST grant code lands (1644)
5534 - fix read performance by not clobbering i_blksize on client (1598)
5535 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5536 - make LRU size a /proc tunable, clears locks when reduced (707)
5537 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5538 - server locks take references on exports now (1558)
5539 - build fixes for 2.4.20-rh trees (1663)
5540 - return an error from lov_create if all OSCs are inactive (1751)
5541 - fix import levels when a reconnect happens without a timeout (1597)
5542 - exit early from mds_open if we get a lookup error (1749)
5543 - partial page read at EOF wouldn't wait for disk before sending (1642)
5544 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5545 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5546 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5547 - write/truncate lock inversion (1639)
5548 - Don't auto-load obdclass, portals modules during cleanup (1495)
5549 - fix timestamps from jumping to "now" (1763)
5550 - extra journal assertions (1648)
5551 - add an extra multiunlink test (1771)
5552 - fix read_record/write_record API (1776)
5553 - fix leak of offset_extent, possible incorrect i_size later (1772)
5554 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5555 - Debugging for the kqswnal_get_idle_tx problems (1820)
5556 - Allow recovery to be attempted multiple times (1536)
5557 - Write out MDS last_rcvd file after it is first created (1600)
5558 - Fix tx_descriptor leak in failed transmit situations (1827)
5559 - ext3 journaling fixes for assertion failure after IO error (1871)
5560 - class_export_put() on freed export after completion AST error (1896)
5561 - Fix revalidate looping in VFS (1322)
5562 - Don't access a freed export during MDS_REINT timeout (1521)
5563 - Add open-unlink recovery support on the MDS (1673,1764)
5564 - Return an error if no MDS data was read from last_rcvd (1946)
5565 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5566 - Don't setup a disk device that doesn't match exported UUID (317)
5567 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5568 - avoid committing NULL handle in force close
5569 - local.sh is now a one-stripe LOV configuration
5570 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5571 - fix ext3 htree duplicate directory entry corruption (1516)
5572 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5573 - update to new LOV EA format (2097)
5574 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5576 2003-06-15 Phil Schwan <phil@clusterfs.com>
5579 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5580 - per-import/export recovery handling (958, 931, 959)
5581 - multiple last-rcvd slots, for serving multiple FSes (949)
5582 - connections are again shared between multiple imp/exports (963, 964)
5583 - "umount -f" would hang if any requests needed to be sent (393, 978)
5584 - avoid pinning large req buffer by copying for queued messages (989)
5585 - add "uuid" to "lctl device" command to help upcalls (991)
5586 - "open" RPCs with transnos would confuse recovery counters (1037)
5587 - do proper endian conversion of all wire messages (288, 340, 891)
5588 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5589 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5590 - fix ptlrpc_connection leak in target_handle_connect (1174)
5591 - fix import refcounting bug in OST and MDS cleanup (1134)
5592 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5593 - fix very unlikely obd_types race condition (501)
5594 - remove osc_open hack for echo_client (1187)
5595 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5596 - a failure in read_inode2 leads to deadlock (1139)
5597 - cancel ack-locks as soon as transaction is committed (1072)
5598 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5599 - make sure to commitrw after any preprw to avoid deadlock (1162)
5600 - failing to execute a file in a lustre FS would lock inode (1203)
5601 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5602 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5603 - obd_uuid2tgt would walk off the end of the list (1255)
5604 - on IA64 the osc would give portals incorrect bulk size (1258)
5605 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5606 - fix lock inversion caused by new llite matching code (1282)
5607 - limit the number of dirty pages on a client to 10MB (1286)
5608 - timed out locks were not being corrected cancelled (1289)
5609 - fix O_DIRECT above 4GB on IA-32 (1292)
5610 * major user-visible changes
5611 - fail out/fail over policy now controlled by the upcall (993)
5613 - add OBD_PING to check server availability and failure (954)
5614 - lustre messages are now sent in sending host order (288, 340, 891)
5615 - add eadatalen to MDS getattr reply (340)
5616 - OST read replies may contain second buffer, with per-page status (593)
5618 2003-03-11 Phil Schwan <phil@clusterfs.com>
5621 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5622 - failed open()s could cause deadlock; fixed (867, 869)
5623 - stop cancelling OST locks when files are closed (481)
5624 - overlapping XID spaces caused network corruption (851, 853)
5625 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5626 - setattr_raw added, to do single-RPC, server-side setattrs
5627 - lmc/lconf syntax change for OST UUIDs
5628 - fix crashy race condition between ptlrpc_free_req and osc_close
5629 - don't use request in mdc_enqueue if we hit a timeout (889)
5630 - don't set the inode i_size for regular files from the MDS (896)
5631 - handle out of order completion AST (842)
5632 - don't LBUG if a lock request times out after receiving AST (913)
5633 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5634 - if a bad lock AST arrives, send an error instead of dropping entirely
5635 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5636 - fix leak in bulk IO when only partially completed (899, 900, 926)
5637 - fix O_DIRECT for ia64 (55)
5638 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5639 - C-z after timeout could hang a process forever; fixed (977)
5641 - client-side I/O cache (678, 924, 929, 941, 970)
5643 - READPAGE and SETATTRs which don't take server-side locks get
5646 2003-02-11 Phil Schwan <phil@clusterfs.com>
5649 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5650 - Fix fsfilt_extN_readpage() to read a full page of directory
5651 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5652 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5653 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5654 - Most intent operations were removed, in favour of a new RPC mode
5655 that does a single RPC to the server and bypasses most of the VFS
5656 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5657 - Aggressively cancel local locks on DLM servers
5658 - mds_reint_unlink sends EA to the client if it's the last nlink.
5659 client uses that EA to unlink OST objects.
5660 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5661 - recursive symlinks were fixed (439)
5662 - fixed NULL deref in DEBUG_REQ
5663 - filter_update_lastobjid no longer calls sync, which annoyed extN
5664 - fixed multi-client small-writes to a single file problem (445)
5665 - fixed mtime updates during file writes (607)
5666 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5667 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5668 - fixed Linux OST size reporting problem (444, 656)
5669 - OST now updates object mtime with writes or setattr (607, 619)
5670 - client verifies file size before zeroing page past EOF (445)
5671 - OST now writes last allocated objid to disk with allocation (108)
5672 - LOV on echo now works (409)
5674 - mds_reint_unlink sends a new buffer, with the EA included. this
5675 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5676 set if a regular file was being unlinked, and it was the last link
5677 - use PtlGet from the target for bulk writes (315)
5678 - OST now updates object mtime with writes or setattr (607, 619)
5679 - LDLM now has a grant-time callback to revalidate locked items, if
5681 - Many MDS operations were reorganized to combat race conditions
5683 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5684 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5686 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5689 - Fully reactivate OST imports after reconnection (512, others)
5690 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5691 - More graceful error handling for truncating on dead OST (515)
5692 - Don't error out unless we're actually accessing dead stripes (474)
5693 - Fix garbage sizes when stripes are missing (410)
5694 - LRU counters were broken, causing constant lock purge (433, 432)
5695 - garbage on read from stripes with failed OSTs (441)
5696 - mark OSCs as active before reconnecting during recovery (438)
5697 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5698 - lfind did not preserve OST order in output (443)
5699 - symlinks cause hung clients, incorrect data (439)
5700 - stop dereferencing request after dropping refcount (457)
5701 - don't LASSERT(spin_is_locked) on non-SMP (455)
5702 - fixes for many rename() bugs
5703 - fstat didn't correctly synchronize attributes (399)
5704 - server must handle lock cancellation during blocking AST prep (487)
5705 - bulk descriptors were free()d too soon (511)
5706 - fix paths in lconf, which would load incorrect modules (451, 507)
5707 - fix confusing lconf 'host not found' error message (386)
5708 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5709 - fix race condition in mdc_blocking_ast() for inode access (526)
5710 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5711 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5712 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5713 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5714 - fix lstripe in several ways (kernel side) (527)
5715 - fix request leak in ldlm_cli_enqueue (262)
5716 - incorrect OSC was marked inactive after OST failure
5717 - call mds_fs_cleanup before unmounting filesystem (524)
5718 - fix races between taking ns_lock and ldlm_lock_change_resource
5719 - fix races updating LOV export open file list
5720 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5721 - fix recovery NULL deref in ldlm_cli_cancel_unused
5722 - fix some DLM races by using new hash table for lock handles (419)
5723 - permit the client to specify desired inodes, at replay
5724 - duplicate requests when we queue them for replay reintegration
5725 - fix last_rcvd offset calculation
5726 - sync after each recovered transaction, so we always make progress
5727 - never, not always, ERESTART requests without transnos
5728 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5730 - skip replay if the MDS says that the client is already connected
5731 - don't check for a recovery-enabled export to match lctl's UUID
5732 - don't INC_USE_COUNT for phantom exports
5733 - don't crash when cleaning up phantom exports (567)
5734 - don't double-finish or set replay data for errored mdc_open requests
5735 - abort requests when they time out, so we don't get old replies
5736 - send/receive replies for AST messages again
5737 - if the client says that it doesn't have the lock, cancel it on the
5739 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5740 mark it as destroyed, it will all work out when decref is called
5741 - fix module use counts (22, 581)
5743 - ASTs now expect a reply (server cancels lock on error reply)
5745 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
5748 - fix many simultaneous client startup (392)
5749 - fix dentry->d_it clobbering
5750 - credentials weren't being shipped for readdir/getattr operations
5751 - remove invalid assertions triggered during some concurrent MD
5753 - proper Lustre versions added (336, 389)
5754 - fix memory leak for create error case (398)
5755 - fix LOV locking bug that would get cli/srv out of sync
5756 - fix echo client over LOV (409)
5757 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
5758 - fix double-O_EXCL intent crash (424)
5759 - avoid sending multiple lock CANCELs (352)
5761 - MDS can do multi-client recovery (modulo bugs in new code)
5763 - many updates, edits, cleanups
5765 2002-11-18 Phil Schwan <phil@clusterfs.com>
5768 - fix null d_it dereference (346)
5769 - fix full OST/dbench hang (333)
5770 - fix permission problem with file removal (286)
5771 - fix removal of OSCs from LOV when they fail
5772 - fix NULL deref during bulk timeout (214)
5773 - fix problems related to multiple filesystems on one MDS (241)
5774 - fixed serious subtle metadata locking bugs
5775 - free locks on clients when inodes are removed due to memory
5777 - fix inode pointer in lock data (285)
5778 - partial support for multiple MDS on a single host (241)
5779 - data locks weren't cancelled at clear_inode time (290, 311)
5780 - intent locks could lead to unbounded lock growth (205)
5781 - added a maximum lock count, an LRU list, and a flusher
5782 - fix multiple rename (365)
5783 - properly abstracted the echo client
5784 - OSC locked 1 byte too many; fixed
5785 - rewrote brw callback code:
5786 - fixed recovery bugs related to LOVs (306)
5787 - fixed too-many-pages-in-one-write crash (191)
5788 - fixed (again) crash in sync_io_timeout (214)
5789 - probably fixed callback-related race (385)
5791 - Add capability to MDS protocol
5792 - LDLM cancellations and callbacks on different portals
5794 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
5797 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
5798 - defer open object destruction to close time (601981, 138)
5799 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
5800 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
5801 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
5802 - don't call obd_setattr in ll_file_release for destroyed objects
5803 * protocol change to lustre_msg: move |version| and add |flags|
5804 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
5805 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
5806 * added replay of create, unlink, link and rename operations during
5807 MDS failover; recovery should be much more robust now
5808 * remove failed OSCs from LOVs (only lov_create uses this so far)
5809 * the lustre-HOWTO was brought (more) up to date (582544)
5811 2002-10-23 Phil Schwan <phil@clusterfs.com>
5814 - in-use dentries weren't being reused properly (617851)
5815 - prevent multiple LDLM setup (599178)
5816 - fix LOV size calculations for truncate (617853)
5817 - fix client handling of MDS intent errors (POSIX)
5818 - fix permission bug in lovstripe.c test (624321)
5819 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
5820 - truncate past end of file could corrupt data
5821 - proper cleanup after timeouts, crashes, etc (592524, 550815)
5822 - a race in recovery could return ETIMEDOUT to apps (623947)
5823 - building outside the source directory was fixed
5824 * the lustre-HOWTO was brought (more) up to date (582544)
5825 * major progress was made on recovery functionality
5827 2002-10-10 Phil Schwan <phil@clusterfs.com>
5830 - recovery deadlock fix
5831 - rm -rf causes LBUG fix (617817)
5832 - file open by multiple tasks fix (618962)
5833 - directory permissions bugs (602707 and 620007)
5834 - journal_stop fixed with locking (611313)
5835 - O_APPEND failures resolved (618273, perhaps 614459)
5836 - lconf PATH fix (619770)
5837 - IA64 build fix (621450)
5838 - RPC buffer sizes scale with amount of memory
5840 2002-10-01 Phil Schwan <phil@clusterfs.com>
5843 - locks would be cancelled without throwing away data pages,
5844 resulting in inconsistent data (605627)
5845 - inode attributes were not always being refreshed (605627, 612449)
5846 - lconf now continues to cleanup after lctl reports an error
5847 - MDS now enforces user permissions (602707)
5848 - lprocfs cleanup fixed, but not yet enabled (614157)
5849 - fixed infinite server hang, should a client not respond to an AST
5850 - avoid going into recovery if user calls readlink() with a buffer
5851 that's too small (613941)
5852 - AST RPCs no longer require replies (614867) -- this may be changed
5853 - don't crash server if client sends an IOV that's too big (611336)
5854 - fixed lock conversion deadlock (611892)
5855 - fixed the following of symlinks (614622)
5856 * recovery: the server can remove locks from a client that dies, other
5857 clients can make progress
5858 * more extN patch fixes
5859 * compile-time configurable ptlrpc buffer allocations
5861 - collaborative read cache document
5862 - Lustre Lite Performance CDR document-in-progress
5864 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5867 - fix typo in patch-2.4.18
5869 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
5872 - clear ptlrpc request each time in handle_incoming_request()
5873 - unlink of files now destroys the object on the OST
5875 2002-09-19 Peter Braam <braam@clusterfs.com>
5877 * add hard link support
5878 * change obdfile creation method
5879 * kernel patch changed
5881 2002-09-19 Peter Braam <braam@clusterfs.com>
5884 - stack overflow bug in extN fixed
5886 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
5888 * documentation updates
5889 - add man pages for config tools
5890 - update tests/README to describe testing with new config tools
5891 - finish metadata API descriptions
5892 * bug fixes and cleanups
5893 - statfs workaround for 16TB limit
5894 - LOV stripe allocation improved, can stripe on subset of OSTs
5895 - LOV file size/IO offset was wrong for files > 4GB in size
5896 - object EA data was being dropped, caused files to be unreadable
5897 - memory overflow with non-LOV OST caused memory corruption
5898 - fixed regression tests to work with new config tools, obdfilter
5899 - fixed bug when directory size became larger than 1 block
5900 - fixed bug (for single client case) when PWD was deleted
5901 - invalidate local directory pages when doing intent-based ops
5902 - avoid LDLM oops when lock callback contained bad data
5904 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
5906 * documentation updates
5907 * bug fixes and cleanups
5908 - configuration tools
5911 - 64-bit compile warnings
5912 - 64-bit internal statfs data
5914 * test_brw on persistent OST devices
5916 * lprocfs (disabled)
5918 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
5920 * documentation updates
5921 * bug fixes and cleanups
5922 * configuration tools
5924 2002-08-30 Peter J. Braam <braam@clusterfs.com>
5927 * many small fixes to 0_5_4
5928 * io/network handling
5929 * thinkos in MDS operations
5931 2002-08-24 Peter J. Braam <braam@clusterfs.com>
5934 * crucial basic fixes to 0.5.3
5935 * IOR, Iozone work over Elan
5938 2002-08-07 Phil Schwan <phil@clusterfs.com>
5939 * version 0_5_3, our first alpha
5940 * we use the new Portals iovs
5941 * documentation updates
5942 * bug fixes and cleanups
5943 * small changes in the DLM wire protocol
5945 2002-07-25 Peter J. Braam <braam@clusterfs.com>
5946 * version 0_5_1 with some initial stability,
5947 * locking on MD and file I/O.
5948 * documentation updates
5949 * several bug fixes since 0.5.0
5950 * small changes in wire protocol
5952 2002-07-18 Phil Schwan <phil@clusterfs.com>
5954 * delivered as Lustre Light Alpha
5955 * fixed a crash after handling invalid MDS requests
5956 * fixed directory pages for architectures with non-4k pages sizes
5958 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
5959 * release version v0_4_4
5960 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
5962 * Fixes a number of interruption problems with OST operations.
5963 * Update documentation for portals header changes
5964 * Move all wire protocol structs/defines to lustre_idl.h
5965 * Fixes symlink length bug.
5966 * Add tcpdump to repository.
5968 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
5969 * release version v0_4_3
5970 * Fixes statfs for inodes on extN.
5971 * Fixes bug in runtests which would delete /etc/hosts.
5972 * Use 64-bit object IDs wherever possible (not into VFS though)
5973 Remove ost_get_info, which is unused by lustre, and out of date.
5975 2002-07-03 Peter Braam <braam@clusterfs.com>
5976 * release version v0_4_2 Fixes a lookup error (type not passed)
5977 * move forward to head of Portals
5978 * move forward to latest Lustre kernel
5980 2002-06-25 Peter Braam <braam@clusterfs.com>
5981 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob